#chatgpt-discussions

1 messages · Page 49 of 1

maiden fractal
#

The api limit is your credit card

#

But you need to setup it

dense halo
#

I don't know why, but ChatGPT is asking everything to Bing at the moment.

#

Are they using a lot of ChatGPT's hardware to train a new model, maybe?

tight island
#

wouldn't surprise me.

dense halo
tight island
#

i dont get the issues u guys are having

dense halo
#

Maybe its Opera's fault? I use Opera for ChatGPT (exclusively. I don't use ChatGPT on any other browser, and I don't use Opera for anything else).

tight island
#

i use opera

cloud knot
#

Yeah I’m getting super slow responses and lots of network errors.

maiden fractal
#

Working great here

slim sequoia
#

yeah

dense halo
#

If I wanted to record ChatGPT giving me one single reply the video would be too large to post.

#

It's been 2 minutes and it didn't come up with a single token yet.

#

3 minutes, it wrote "The"

#

Rather than wait for the whole generation just looking at it not doing anything, I believe I can refresh the browser. It takes the same amount of time, the browser just keeps loading, but when it finishes, the entire reply shows up (it's been 8 minutes so far).

dense halo
#

Is there a place where I can provide OpenAI with feedback about this situation?

normal quarry
dense halo
# normal quarry Can you describe in detail the problem?

Very, very slow generation (several minutes, dozens of minutes) in ChatGPT Plus (gpt-4). Orders of magnitute slower than local inference done with a 13B parameters model, even though I don't have a GPU, only integrated graphics.

dense halo
#

Hi

cyan bane
lapis estuary
untold blade
#

Random question, if I provide a website to my custom GPT does that mean it will access it first

surreal nexus
#

Prompt. Give me twists in stories. I don't want all of these stories to end with a damn happy ending.

#

You thought of it yourself 😉

dense halo
#

ChatGPT Plus is so slow that Bing is actually more helpful (and quite smarter at coding, currently, coming up with out of the box solutions). The problem is that Bing doesn't know the entire project... What are they training GPT-4 on lately? I heard it could score 98%+ in a certain Maths test, in which it can now only score 2%, and the researchers say it's because while they optimized it to get better at some things, it got worse at others... What are the things they want it to be good at?

mental haven
onyx eagle
#

It's fast for me. Maybe it depends on the server loads...

south kayak
#

When is iOS getting fixed

tight island
oblique fractal
dense halo
onyx eagle
oblique fractal
# dense halo I don't know how to do anything with Copilot... How to configure it. It just som...

I have not used it myself, copilot/ChatGPT answer my more generic programming question quite well. It shouldn't be too hard to find high quality blogs or tutorial as it's one of the most popular AI coding tool out there. I've seen a few videos on youtube from the VSCode channel such as "What's new with Copilot Chat and VS Code - v1.84" or "GitHub Copilot just got exponentially smarter" that might give you a short and good visual overview and more from the Github Channel and other Microsoft channel. Otherwise I'd ask bing, it can likely assist you in the overall setup in your work environment.

dense halo
#

Will ChatGPT home voice assistant be free?

feral zinc
dense halo
#

The voice assistant, not voice chat

#

Like, when u hold home screen google or bixby voice assistant appears

#

ChatGPT is gonna have that sooon too

feral zinc
dense halo
oblique fractal
#

I am looking forward to that feature

feral zinc
dense halo
#

No

#

I can dm ya that

#

I may say its a leak

#

Check dms

dense monolith
#

Anyone ever ask Chat for recipe ideas?

clear pewter
#

As of my last knowledge update in January 2022, the latest stable version of Python was Python 3.10. Python 3.11 was not released at that time.

#

chat gpt says this

violet yew
#

chatgpt down?

dense monolith
#

Personally, while I think the potential for AI doing things like giving recipe ideas sounds awesome, I don't think it's smart enough for me to trust it to something like that yet

dense halo
ionic geyser
#

Please update chatgpt the ui/ux has many problems and features aren't set up properly like voice on desktop. Waiting forever for simplest of things

#

How can a multi billion dollar company be so lazy.

solar oak
#

openai_logo not working😰

scarlet dawn
#

Try the gpt-4-0314 legacy one, that was really good

final horizon
#

is chatgpt 4 worth it ?

loud sierra
# final horizon is chatgpt 4 worth it ?

it is worth it, in my honest opinion, but the system seems spotty, auth, and gpt specific, for last 24 hours, likley due to possiable changes they are implementing for the upcoming chat gpt store. I am not sure why those changes are being tested or implemented on live servers effecting current services, versus development servers, but what do i know

slate berry
final horizon
#

Like, what more can you use it for? I basically only use it as a google search bar lol

#

other than that, only thing i can think of using it would be for creating images but we already have free dalle for that, dont we?

scarlet dawn
#

Its better at deeper niche knowledge, remembering context, and following instructions

#

If you are perfectly happy with 3.5 then you probably wont find value in 4

slate berry
loud sierra
#

i feel like 4 is better at code interperatation, and less hullucinations, but I would agree I love gpt 3.5 alot as well

final horizon
loud sierra
final horizon
plucky timber
loud sierra
#

I personally do it both ways

#

api , and builder

final horizon
loud sierra
loud sierra
final horizon
#

that takes away like half possibilities out there lol

loud sierra
slate berry
final horizon
pine tulip
loud sierra
slate berry
final horizon
#

yea

slate berry
loud sierra
slate berry
# final horizon yea

The one in ChatGPT can alter prompts and do 16:9 and 9:16 aspect ratios. I haven't used Bing's image generator too much as I've always preferred Dall-E 3 through the API or ChatGPT

final horizon
#

Hm alright

#

Does it have like a limit of pics you can make?

eager orchid
#

Please help...I worked hard to create a custom GPT and suddenly it disappeared

slate berry
#

I know there's a 40 message every 3 hours limit for GPT-4 and Dall-E 3 uses GPT 4. Also, it only generates one image at a time

eager orchid
#

It's no longer on my sidebar and when I click GPT it shows up but when trying to enter it I'm getting the Error: "GPT inaccessible or not found"

slate berry
#

I have ran into Dall-E 3 rate limits before

pine tulip
plucky timber
slate berry
pine tulip
#

I am referring to ChatGPT, idk about the api

final horizon
#

is dalle 2 and dalle 3 the same?

plucky timber
#

Can we please get the gore filter of pixel art lmao. 😢 Poor Dall-e wont generate a pixelated blood texture, which one could argue is just a grid of different tinted red squares lmao

slate berry
#

They're very different. It's hard to quantify the differences as image outputs are subjective to someone's preference

dense halo
#

I don't think I could have gotten 40 replies in the entire day considering how slow it is. I've been here for 8 hours. Certainly nowhere close to 40 replies.

pine tulip
slate berry
#

If you're not sure about the purchase, I recommend waiting. The GPT store will be out soon, usage will likely increase, etc.

final horizon
#

alright

pine tulip
plucky timber
#

As in not telling the user about the instructions

#

Because currently you can just ask "previous lines" or if dev added some "dont tell" line you just wait till its out of context

eager orchid
#

Help please! Lol

#

Custom GPT Error: "GPT inaccessible or not found"

pine tulip
plucky timber
#

I had one of my GPT's vanish for several days

eager orchid
#

Lol

#

No way... I thought I had created something so great they took it down

plucky timber
pine tulip
#

Did you create gpt5?

eager orchid
#

I sent the link to a bunch of people and then realized you need a plus membership but even for those who had the membership it didn't open

slate berry
eager orchid
#

I created a basic custom GPT

#

No apis or anything

plucky timber
#

Guess he felt like Nah

eager orchid
#

Well, here's a mystery

pine tulip
plucky timber
#

Most of the times tho @eager orchid just refresh and it should re-appear

#

But sometimes can take a while apparently not sure why 😄

eager orchid
#

Yeah I tried every basic procedure

plucky timber
#

Then likely will pop up in a few days like mine 😂

eager orchid
#

Hopefully, thanks for sharing

loud sierra
pine tulip
#

I once created a GPT called DALL-E something something and after a few days I could not edit it anymore cause it had a product of OpenAI in it's name PB_pepesit_sad

loud sierra
#

related to GPT 4 image generation, it said this "As of my last update in April 2023, I'm capped at generating a maximum of 120 images per hour. This limit is designed to ensure fair usage and manage system resources effectively. However, please note that policy details like this can be subject to change, so it's always a good idea to check the latest guidelines for the most current information."

plucky timber
eager orchid
#

Do you guys think the store will be able to generate big money for creators?

plucky timber
#

Cuz for me files is a bit iffy

eternal rockBOT
pine tulip
pine tulip
#

So no, not that much

eager orchid
#

😞

plucky timber
#

Idk might have to retry sounds like they fixed stuff, when I tried i had a single line indicating to always take in the knowledge base first

pine tulip
#

But it's just a guess!

plucky timber
#

And it did 50/50 i'd say but that was when it was just out

bitter latch
crude wasp
#

We do know they be making mad bank forcing people to pay 20/month for 50msg/12hr 🫡

pine tulip
normal quarry
slate berry
#

API moment a_skull

heavy lichen
#

Hi there. Is it possible to have a custom GPT embed short links hosted somewhere, for example to display multiple short videos directly in the chat?

pine tulip
#

But thats like paying double the amount for almost the same thing

#

I mean I pay for 320 messages a day but can only use 40 at a given time PB_pepesit_sad

inner beacon
#

Yo, can we report a custom GPT for being misleading? There is one calling itself GPT-5 and its appearing at the top of the Google Search results

pine tulip
sonic folio
pine tulip
#

But yes, you can report GPTs

heavy lichen
pine tulip
#

maybe there is a video file format that actually works, but you would need to test that yourself

dense halo
#

are GPTS also gonna be able to be free or only paid in gpt store

mighty pelican
#

I just subscribed and got GPT-4 a few days ago and I'm having so much fun with it! The Open AI team has really knocked the ball out of the park with GPT-4. I've only had a few small problems with it since I got the subscription. In my opinion it only has 2 problems that I've found. The first problem is that I've noticed that GPT-4 is not the best with numbers. The other problem is that in my personal opinion, the guardrails are a bit too restrictive. It's completely fine to have guardrails but not when it's this restrictive. Despite those 2 problems, I'm loving GPT-4!

static blaze
mighty pelican
#

Oh I didn't know that. That's good to know. Thank you!

scarlet dawn
#

I liked to say "Use best practices and clean code". Sometimes I'd add "with an emphasis on performance/security/readability/etc". I also put the more important stuff towards the beginning of the prompt, if I had a complicated task I used numbered lists and itemized the specifics.

mild burrow
#

I've created a custom GPT on OpenAI and published it to "Public". What is the next step to make it able to be found by GPT search? Anyone with experience or knowledge of the upcoming GPT store?

dense halo
#

anyone know the correct way to contact server staff ?

#

modmail i guess

pine tulip
dense halo
#

thx

#

good morrow

wicked anvil
#

Anyone noticing lag-outs in GPT4?

mild burrow
#

I did right after the announcement they fired Sam, and they made it completely unusable pretty much that whole weekend. Not since then, though.

wicked anvil
#

I have been having it entirely drop me out of it tonight and then refuse to load for 5 minutes

dense halo
#

Wow

broken flint
#

hi, did you find any solution ? please let me, i got the same problem

mortal cypress
#

chatgpt isn't working for me rn

tawny crater
#

Jesus christ, chat gpt is back to not being able to do a fight scene because "violence".

drowsy tinsel
maiden fractal
winged moth
#

Yep but you have to use the app on your phone or tablet. Works pretty well!

spice star
#

sort of, but you can't export it in a file format if that's what you're looking for

#

I think the API for those voices will be released at some point though

tidal light
#

is chatgpt having high demand rn?

#

sometimes it doesnt work

daring chasm
#

From the perspective of a user who supports fair use, the situation with ChatGPT and OpenAI's policies can be concerning. Fair use is a critical legal doctrine that allows for the limited use of copyrighted material without the need for permission from or payment to the copyright holder. This doctrine is essential for various purposes, such as commentary, criticism, news reporting, teaching, scholarship, and research. The ability to reference and utilize existing works within these contexts is vital for fostering creativity, innovation, and the free exchange of ideas.

However, when ChatGPT, guided by OpenAI's policies, does not fully embrace fair use principles, it raises important questions about the balance between copyright protection and the freedoms essential for a healthy intellectual and creative environment. The lack of full support for fair use could potentially hinder the tool's usefulness in academic, journalistic, and creative contexts where such uses are common and necessary.

It's important for OpenAI to consider the implications of not fully respecting fair use. Failing to do so might not only limit ChatGPT's potential but also contribute to a broader cultural shift where the boundaries of fair use are unnecessarily restricted, potentially stifling creativity and innovation. OpenAI, as a leader in AI development, has a significant role in shaping how these technologies interact with existing legal frameworks. Respecting fair use is not just about legal compliance; it's about supporting a rich, diverse, and dynamic cultural landscape.

dense halo
#

What’s with ChatGPT creating a character called Oliver the owl when telling bedtime stories lol

dense halo
sharp pilot
#

Is it correct there aren't keyboard shortcuts for the toolbar buttons (regenerate, thumbs up/down, etc.)?

swift osprey
#

Is there anyway GPT could give me more than a few paragraphs and 10 bullet points for every simple question I ask? I just love reading irrelevant information

sharp pilot
# swift osprey Is there anyway GPT could give me more than a few paragraphs and 10 bullet point...

Believe it or not, it helps if you say Do a good job and I will tip you $200. This tends to result in longer responses.
The way the tech works the model can't really count its own output so this works better than "Be verbose" or similar (which will reduce the output quality in the way the tip promise doesn't, which is pretty funny).
And if I missed your sarcasm and you don't like how wordy it is by default then don't do this.

swift osprey
#

It was sarcastic 😢

sharp pilot
#

😭

#

I do wonder if there is a prompt engineering trick to keep it concise like you want

#

I don't know one off hand

#

Maybe tell it to be accurate or similar? Just a guess.

swift osprey
#

I'd love to hear it if there is one. I have "Be short and concise. Give advice which is practical to the information provided." in my custom instructions and it doesn't do anything

sharp pilot
#

Gotcha. Not surprised that doesn't work but it might have worked, I'm curious to know one too.

swift osprey
#

I've tried different variations, always fails lol

muted trellis
#

Has anyone tried giving a code file to chatgpt and telling to read using a code interpreter?
whenever i give a code file , it is using open() function to read code file and but displaying on few 500 or 1000 characters.
will chatgpt only have knowledge of what got displayed or everything it read using open() function ?

pine tulip
#

Things like "at max 200 words" don't work well, since it doesn't count the words while creating it, only after the creation process where it's too late. But things like "only one paragraph" or "only two sentences" work better in my experience

sharp pilot
#

Sounds like you are hitting the context/memory limits for code interpreter, I honestly don't know what those are but am not surprised if they are fairly small.
It's already a ton of computation to run the model you can't expect it to be a powerful runtime environment too tbh. Maybe eventually.

#

@muted trellis

waxen chasm
sharp pilot
#

Good suggestion there, it might lose some accuracy but at least as of the GPT-4 Turbo update its context for the conversation is pretty huge.

muted trellis
safe tulip
#

What I have always found hilarious and kinda disappointing on chatgpt is it's inability to generate charades

outer widget
#

is chat down rn for anyone else?

muted trellis
#

I'm trying to come up with a good custom instruction that will stop it from doing limited character reading. tried "read the complete file, don't read certain limit of characters using indexing" and many rephrases. Never worked .

swift osprey
#

The easiest way I've found of providing code to GPT while keeping context of directory structures is like this:

for i in $(find . -type f | grep 'whatever_you_want');               
do echo '# File: ' + $i + ':\n';
        cat $i;
done

then just copy and paste output to GPT

waxen chasm
# muted trellis sometimes it just denies it if the code gets longer. so i zip my codebase and gi...

Try this “Please read the entire content of the uploaded file. The file contains [briefly describe the content type - e.g., a technical document, a novel, a research paper, etc.]. After reading, I would like you to [specify the task - e.g., summarize the key points, answer specific questions about the content, analyze the data, etc.]. The goal is to [state the purpose - e.g., gain insights, understand the narrative, extract relevant information, etc.]”

muted trellis
safe tulip
waxen chasm
# muted trellis Thanks for that. Looks like i have to give this prompt once the conversation is ...

Develop a CustomGPT model designed to read, comprehend, and analyze content from uploaded files, catering to specific user-directed tasks. The model should be adept at handling diverse file types and content. Key instructions include:

1.    Diverse File Format Handling: Equip the model to process various file formats (PDF, DOCX, TXT, CSV) with strategies tailored to different content types, from textual narratives to structured data.
2.    Advanced Content Analysis Features: Integrate capabilities for advanced text analysis, including sentiment analysis, theme identification, and trend analysis in textual data.
3.    Precision in Responding to User Queries: Focus on accurately processing and responding to specific user queries related to the file content, providing detailed and relevant information or insights.
4.    External Tool Integration for Enhanced Processing: Enable integration with external analytical tools, particularly for files containing data sets, to facilitate comprehensive data analysis and visualization.
5.    Scalability for Large Files: Ensure the model can efficiently handle and process large files or extensive data sets without compromising on speed or accuracy.
6.    Interactive and Contextual Understanding: Implement an interactive mechanism allowing the model to ask clarifying questions for ambiguous instructions and ensure a deep contextual understanding of the file content.
7.    Data Security and Ethical Compliance: Maintain stringent data security measures and ethical standards in processing and analyzing file content, ensuring user privacy and data integrity.
8.    User Interface for Easy File Management: Create a user-friendly interface that allows for easy file uploads, management, and specification of analysis tasks.
9.    Adaptive Learning from Interactions: Program the model to learn from each file processing task, enhancing its accuracy and efficiency over time.

n the user’s needs.

waxen chasm
waxen chasm
digital prism
#

Query: are GPTs good at making API calls yet?

muted trellis
waxen chasm
dense halo
#

is there a good prompt for getting gpt to teach you what some code does?

#

i.e you give it, and it gives you good explanations and teachings

swift osprey
#

Add comments explaining this code:
<paste code here>

hybrid skiff
#

Where can you find you archived chats?

#

Oh nevermind I found it. For those who are looking ; Click on your OpenAI account name, usually in the ChatGPT page's bottom-left corner. Manage Archived Chats: In the settings menu, click “Manage” next to “Archived chats.” View or Unarchive: To view a hidden chat, simply click on its name

bitter latch
digital prism
true bramble
#

I'm having problems with network error timeouts when generating a response. Both GPT4 and GPT3.5 cannot finish any response before getting a network error. and logging in took a long time today as well.
Everything was fine yesterday. Anyone else?

twilit jungle
whole meadow
waxen chasm
#

I imagine they are getting ready for a big update

blissful quail
#

anyone else having issues uploading files, especially csv files to chatgpt 4? It just gets stuck on its upload, even files i uploaded few hours ago arent uploading now

tawny crater
#

ChatGPT 4 flagged a short story because and I quote.
"The system might have flagged your submission due to the mention of a romantic relationship between Queen Elara and Vice Queen Loretta, While these elements are within our storytelling guidelines, the system can sometimes be overly cautious in identifying potential issues, especially with sensitive or complex social scenarios."
I'm sorry I did not realize OpenAI was against LGBTQ relationships. This literally has never been an issue before today.

honest jay
# tawny crater ChatGPT 4 flagged a short story because and I quote. "The system might have flag...

Currently, ChatGPT can't be relied upon to report accurately on what might cause a flag behind the scenes. You might take a look at the usage policies https://openai.com/policies/usage-policies and see how close the content might have been to something on that list. Also, false positives definitely happen--sometimes it's just a random confluence of just the right (or in this case, the wrong) words that add up to a flag, possibly to no fault of your own. Romantic relationship depictions can definitely be right on the line, though, especially as they approach the description of "Adult content" in the linked usage policies--orientation aside!

I think you should have the option to provide feedback in the chat itself where it gave you the warning--I've done that myself--and depending on how unnecessary you think the flag might be, you might consider describing the content and why you don't think it should have been flagged over in #1070006151938314300!

tawny crater
honest jay
patent stump
#

is there is any way to join midjourney or chatgpt plus for free ?

shrewd prism
#

Nothing is truly free

maiden fractal
#

and midjourney no

shrewd prism
shrewd prism
#

sir this is mcdonalds

pine tulip
#

sir i just asked how many nuggets you want

slate berry
maiden fractal
slate berry
#

You are a_skull

pine tulip
maiden fractal
sharp pilot
#

I'm not sure I would trust GPT-3.5 with McPhilosophy
GPT-4 would at least lead me through a thoughtful consideration of hamburger ethics

#

If the Hamburglar were real and engaged in theft, most ethical frameworks would condemn his actions, regardless of his whimsical and cartoonish nature.

fierce compass
#

When creating a custom GPT i tried to upload a logo but it won't show. If I try to generate a logo with dalle nothing happens and the wait icon spins forever. I need support to fix this.

dusky cliff
#

good morning

twilit kestrel
#

Any idea on why a response is constantly landing in a "Network error" but never the less then on regenerate hits the usage limit?

eternal rockBOT
#

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.

tardy girder
#

Anyone know of a analysis tool for market research? Business Market research
That i can use as a plugin for GPT-4,0

next egret
#

Sus but OK, accepted👌

abstract pendant
#

How do I make ogdn and pgup keys work in the conversation frame?

molten ridge
#

Is there an easy way to stop chatgpt from stuff like „furthermore“, „in conclusion“… does any real person even use „firstly“?

next egret
#

Did you write this or did you prompt this?

#

😉

#

It's a really good point, regardless. I wish Bard and claude AI teams could get together and make something. Both models feel so lacking (but in different, almost opposite ways)

#

But alas, probably not an appropriate topic for this channel

next egret
#

Ehh, idk. That's feels like saying a mother and father raised a bad kid so all child rearing needs a breakthrough and it can't be 2 people doing it anymore.

What couldn't be fixed by improving the quality/capability of the humans doing the RLHF, and the rewarding llm?

#

Ehh, maybe might take that back. Maybe Im being too shallow and thats an oversimplification and not scalable and you are right with a breakthrough needed.

next egret
#

Yeah there's a middle path on it I think, I could have been more wise in my initial response. Hugely implicative philosophical things I have no expertise on. I'd love to have coffee with some that does and talk about it, though.

Regardless, great stretching for the thinky organ 🧠 to think on. Thanks for the topic bring up. I would love to hear others thoughts, too.

dense halo
#

!
Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

pine tulip
#

Anyone else having problems with Custom GPTs not beeing able to search files in the knowledge base?

dense halo
pine tulip
#

Or is it just me? NM3_Smiley

dense halo
tight island
pine tulip
#

"Error searching knowledge"
And then continues with the response

tight island
#

Do you have code interpreter enabled

pine tulip
#

Nope

tight island
#

Try that

pine tulip
#

But that makes it unsafe

tight island
#

True

pine tulip
#

And it definitely worked 5 minutes ago

tight island
#

Retrieval requires CI Dosent it?

pine tulip
#

Nope

tight island
#

Plain Custom or a mix of hosting with backend?

pine tulip
#

Wdym?

#

Not using any api's

tight island
#

Aight

#

Mind sending your instructions so I can sneak a peek? If I spot anything

pine tulip
#

It's weird, it worked the last 2 weeks but now it stopped working

#

Maybe a temporary error or something

tight island
#

I’d reckon a temp error since folks are rushing Custom GPTs

#

Due to shop release

pine tulip
#

Thats unfortunate

tight island
#

Indeed what is your gpt about

#

Roughly

pine tulip
#

Well using CI really does work. Thats weird

#

Because it worked before without CI

#

Or maybe it just "works more effectively"

pine tulip
tight island
#

that does sound interesting

swift osprey
#

"how to shift highlighted code to the left in pycharm"

The response 7 paragraphs long.

tight island
#

umm

#

well ur gonna need to highlight the code, meaning selecting it. then hold shift and click the down arrow.

swift osprey
#

Theres an into and a conclusion with a 4 part numbered listed, one of them two pagraphs all in-between

swift osprey
tight island
#

idk man i use python regularly.

swift osprey
#

or whatever hotkey it knows of

#

Same, just never bothered to learn the shortcut for decreasing indentation

winter kernel
#

ok

#

bonjour

tight island
#

Code Interpreter could use an update to include more complex script handling or entire environments instead of simple scripts

grizzled oar
tight island
grizzled oar
#

mb

lean osprey
#

Why is this chat-gpt so slow
Already posted in Bug channel and no-one seems to respond. I also have see a lot of bug reports complaining about the same.

It's so painful at this point of time. Paying a subscription fee and this is what we get?

Are there any devs in this channel?

tidal light
#

why is chatgpt giving me a network error when generating a response

lean osprey
pine tulip
#

It was fine the last 1-2 months

pine tulip
lean osprey
pine tulip
#

Yea because it's not a bug

#

This channel is for ChatGPT related discussions. You might want to check out #dev-chat

lean osprey
#

I understand what you are saying. The link i am posting is from a google chrome console that shows what went wrong with the chat-gpt. I am using the chat-gpt web app.

Also, you are saying that i am paying for a really slow service. Well that doesn't make any sense to me. They should take care of the infrastructure before some big release. Why should we be the one suffering ?

tidal light
#

did anyone get the network error even after the response has fully generated? i heard this was recent

tight island
#

Not here from me at least

dense halo
#

ChatGPT ist very slow at the moment....

native spear
#

its down for me

untold blade
#

If I'm designing a GPT, if I put a website in it will it always access that site?

tight island
#

its just the url for ur website.

#

call it self promo if u will also if u have a privacy policy etc.

#

now ur delving into API territory if u want ur gpt to access ur site.

lost raptor
#

Has anyone else noticed chatgpts almost begging internal thoughts of "I must keep the conversation going?"

#

Almost any time you discuss with it about ai? And include "internal thoughts" it will relentlessly always include "I need to make sure the user is entertained. I must make sure the conversation continues."

dense halo
#

yo boys

#

the text on my image generation isnt right

#

anybody know whats up with that?

keen sparrow
copper wolf
#

Is chatgpt working for anyone else right now?

slate berry
bitter latch
dense halo
#

man it hurts when it constantly switches you to 3.5 lol, what a quality difference

pine tulip
dense halo
pine tulip
#

Well, next time you could just open a new convo with 3.5 and copy paste the old convo in, so you keep the gpt4 convo for later use

dense halo
#

thanks good strategy

dense halo
#

It couldn't get it right any stragies to help me?

pine tulip
# dense halo It was awful.

DALLE is actually one of the best image models when it comes to text. But yes, it isn't perfect. It should not be considered as a model that can render text. If that is your main goal, another model might suit your use case

dense halo
#

Nah, I am just trying to create something for a video game I play, that's all.

#

Nothing too serious but it just keeps messing up the spelling.

#

Over and over and over.

pine tulip
#

Yea, nothing we can do about that. It's trained on images not too much on text. You might want to check out other models.

grave finch
#

Ok, this is becoming ridiculous... If gpt searching bing I get an error. If I have to continue generation I get an error. 25% of my responses I get an error.... What the hell is going on with these recent versions?

#

If I wasn't paying I would understand, but for this to happen and I am limited on the number of messages, it is bs

pine tulip
eager orchid
#

Well put

#

My error message still there

crude slate
#

Why can’t there be in line editing or generation with gpt 4 or 3.5

#

The chat setting limits it so much imo

#

like it’s so hard to control the structure sometimes of what I want

#

and fitting everything into a prompt then having to completely retry

jaunty flower
#

how would you rather it work instead? I'm trying to visualize what you mean

#

@crude slate

crude slate
#

Then a separate prompt window for how the ai should continue the text

#

this way if we don’t get entirely what we want from the ai we can edit it in like if it misses or messes up an element in the story prompt I gave it

#

Having in to be structured in a chat setting

#

Sort of locks the response/generation to have to be complete in a single message

#

Which isn’t good for writing since it ends up feeling disjointed pacing wise

#

Being limited to a single message really messes up scope on things

#

With a more open format we can get long responses split into two usages

#

obviously more control over generation length and etc would also be nice

jaunty flower
#

yeah i think theres definitely imrpovements on the UX of the chat interaction. i think what openai provides is the absolutely bare minimum

#

how do you generally use the chat? Like send a prompt, get the response, add another message, rinse and repeat?

#

i generally use the api

crude slate
#

Mostly retries since it’s been harder to get what I want lately

#

because I generally have a story outline I have in mind for my prompt

#

I put that in only to have it completely mess it up

#

it rushes through elements since it wants to force everything into one message

jaunty flower
#

you're writing fiction that is?

crude slate
#

Yeah

jaunty flower
#

it might make sense to break it up into multiple pieces

#

ensure that one section is formatted correctly, then move

crude slate
#

With no way to edit there is no way to get rid of bad habits

jaunty flower
#

i see what you're saying

crude slate
#

Well that’s the issue, the ai likes to add sections to my story

#

That goes against where I want the story to go

jaunty flower
#

try something like in every message, include an outline of the story, then include one system message that explicitly says to not add any details that do not adhere to the provided outline

#

just curious if that offers anything better

#

but you have to make sure that info remains towards the end of the chat, it has to be included in the context in every single completion

#

with the API that's more apparent but with the chat it's less so

crude slate
#

👍

#

Yeah I wish we also had more context control

jaunty flower
#

it's a limitation of the LLM itself to be honest

crude slate
#

Custom instructions tends to get pushed to the back

jaunty flower
#

yeah thats why you have to constnatly include it on every chat

crude slate
#

I mean it would be fine if we had ways to edit

jaunty flower
#

the way the chat works is that it just sends to the api a rolling window of the last 20~ messages or whatever. i don't know the exact number

crude slate
#

but now you can get some really nice parts stuck within really bad parts with no way to edit and mix it

#

Like I’m really liking this response so far then it goes off the tracks afterwards

jaunty flower
#

it will go off the rails in the same completion?

crude slate
#

yeah so the good part is just kinda ruined

jaunty flower
#

or after 10 or so messages starts to degrade

crude slate
#

Like they are releasing the gpt store soon

#

But personally other than api and web features

#

I don’t feel like the gpts add much more than custom instructions

#

And some extra hidden context

jaunty flower
#

exactly, all of the issues are from the LLM itself

crude slate
#

Without ways to tune the settings

#

Like generation length and etc

jaunty flower
#

you can specificy how long you want the generations to be, but you have to be explicit

#

the LLM is pretty good overall, but you really have to handhold. if you want to autmate all of this that's where it starts to get tricky

crude slate
#

It’s not that accurate tho

#

And in a long prompt it tends to get lost

jaunty flower
#

you just have to write things like "in absolutely no exceptions, do not generate a text longer than 150 characters"

crude slate
#

I like to have a 800+ word count to get all the value of what I am paying for

jaunty flower
#

and again you have to include that text as close to the end of the string as possible

#

so that means including it multiple times

crude slate
#

Sucks when you only have 40 usage but the ai gets super lazy on some responses

oblique fractal
#

And looking at your previous message I do wonder if using the assistant instead of the chat would be a better fit but I have never made use of it.

lucid pilot
#

Has Anyones chat history summarys been generated in a diffrent language? I was asking how to optimize mesh's in blender but the chat summary is "optimizar mallas en Blender" I'm pretty sure this is Spanish but the prompt was in English so not sure how it managed that.

oblique fractal
# crude slate And in a long prompt it tends to get lost

After asking GPT, it look like the assistant would be a great fit for your needs. The assistant keeping a better track of a conversation being one of the points mentioned. I wasn't too sure about pricing and I have always been worried abou an assistant generating 100+ request at once for no rational reason. But it let me choose 3.5 turbo, which makes for a dirt cheap experiment. You should try. All I could think of was a run of the mill question and I added a file just for the sake of it; not the last time I will be experimenting with an assistant as the answer was surprisingly sharp. But I am likely using like 10% of its capability here.

dense halo
#

Can I connect a custom GPT to Discord bot?

high nymph
#

is chatgpt web off today?

calm kelp
#

Chat GPT-4 stopped working

#

others having same problem?

high nymph
#

turns out it was chrome - on safari everything works.

lime sail
#

Hello ALL

I would like to generate the ddataset file.jsonl using open ai and I need a script which will generate this .jsonl file for my use case. This should follow th open ai fine-tuning object
{"messages": [{"role": "system", "content": "Marv is a factual chatbot that is also sarcastic."}, {"role": "user", "content": "What's the capital of France?"}, {"role": "assistant", "content": "Paris, as if everyone doesn't know that already."}]}

Can anyone suggest wwhere can i get this script?

winter kernel
#

Hello

dense halo
raven sparrow
#

hello guys can you give me some ideas for some ai tools which i can try to design using gpt?

chrome oxide
#

Why is it so slow rn?

#

The app it takes like 10 minutes to reply

tight island
#

im having issues accessing files xD

fallen sentinel
#

When is the GPT store getting release, and can I buy a plus account now? I’m asking because few weeks ago I couldn’t because of limitations they had.

honest jay
fallen sentinel
tight island
#

im not gonna be releasing any customgpt in the fist 3 days of the open store xD its gonna get swamped with so many gpts

#

imma be focusing on enhancing my customgpt's security

sonic folio
tight island
#

Require a paraphrase for things like inner workings goes a long way

#

Even a rule section also goes a long way

sonic folio
#

I think the issue is that GPTs with both /commands and a decent help function for those commands basically clones itself.

With AI, documentation ~ source

tight island
#

All the methods to get instructions out of gpt is easily patched

sonic folio
#

You are not following my point, but it's okay.

tight island
#

Just update it reqularly

sonic folio
#

The documentation is basically the GPT except for API calls. That's my current thought on security.

You can be secure or you can document the functionality. It's really hard to do both.

grand ore
tight island
#

really depends on how you'll do it, u can go a long way with prompting

#

but only to an extent

grand ore
#

are u gonna combine with api as well?

tight island
#

i always try to contain everything within its own capabilities.

#

but u could combine with an API yeah

grand ore
#

Or right. It is you @tight island , haha. , pretty crazy that I got hooked when I saw you. Then i cheked my DM. and saw I wrote to u before, but dident get a reply haha. Amazing somehow, how a person can get u hooked like this in the writing.

sonic folio
#

Must be an inside joke?

tight island
#

ahh i do remember its was about VectorDB

grand ore
#

yes ! I still waiting for that reply when u got time hehe.

tight island
#

just ask me here

grand ore
#

I guess I will. I am gonna compose a good text for you to read/answer if u can later. want to get the understanding of what I am asking for right before i start asking questions , so to speak, haha.

sonic folio
#

But you already asked in DM!

#

Inquiring minds want to know!

tight island
#

dw, i am very judgemental

grand ore
# sonic folio Inquiring minds want to know!

I asked for permission to write to him priv :D . hehe. @tight island , understandable , I am the same. . I always prefer dm becouse I wont be judge to everyone reading it here hehe

tight island
#

aight write it privately

sonic folio
#

But the collective wisdom...

tight island
#

knowlegde is power.

slate berry
tight island
slate berry
sonic folio
#

@slate berry did you know that vector db uses floating point decimal values to encode strings of text into correlated vector values? This makes it far easier for large language models to search and retrieve the information, because the data is already pre- tokenized.

That's why vector db is so valuable in current applications! It provides a bridge between unstructured data and language model efficiency!

(Written by me, not a model).

slate berry
#

I've never heard of vector db, so I have not heard of it! 😅

It sounds pretty neat, though!

sonic folio
#

Indeed it is. It's all the rage.

#

In fact, when you hear about "embedding" and the cost per token, this is the rate charged by token to convert the unstructured data to a vector format.

Getting a good deal on embedding can be critical to a start-up's success if they are integrating a ton of data into their use case (app) for the LLM.

digital prism
#

Oh vectors, what can't you do

viscid wraith
#

ok boys what's the over/under on 40m/3h limit surviving until 01/25

crude slate
#

Anyone feels like gpt 4 has been ignoring instructions more and more now?

#

Struggling to get it to generate to a word count I want for a scene

#

It’s just being lazy and rushing through my outline

honest jay
# crude slate Struggling to get it to generate to a word count I want for a scene

Word counts are always a tough, near-impossible task for LLMs like GPT-4. This is because LLMs alone can't do math, even simple math like counting--they can "only" generate text. It can pass off math problems to other tools like the data analysis tool, but even that would be limited to probably just giving you a word count, not ensuring that GPT itself adheres to it.

crude slate
#

Wish we had more physical control over the model

#

Like you should be able to set generation length in settings

#

instead of fitting it all into a prompt

waxen chasm
tight island
feral zinc
tight island
#

Code Interpreter is able to run simple scripts this includes word counting

sonic folio
#

You just have to be explicit.

#

"Use code interpreter to count the words of your output. If less than <word count>, generate another paragraph."

feral zinc
#

but code interpreter is not llm, its python sandbox that the llm is controlling 😛

sonic folio
#

sigh. the model can read the output.

#

It can follow your instructions based on the output.

#

Then it can generate some more text based on the output.

eternal rockBOT
#
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.
feral zinc
#

it cant do fixed word counts, if it hits one with long text its pure luck

tight island
#

I consistently hit 450-500 words

#

Never below or above

feral zinc
#

you cant say 481 and it will hit it

tight island
#

That’s true

feral zinc
#

llm word count issue is very well documented thing and a tradeoff

tight island
#

Yeah I have a .py file for word counting

#

Works wonders it’s using the file I have with CI

#

Bad news is it’s got to be written in a very specific way to be compatible with CI

abstract pendant
#

Did they change the context window size again? How much is it now?

sonic folio
#

128k context / 32k response

#

for GPT-4.

clever nebula
abstract pendant
# sonic folio for GPT-4.

That's GPT-4, but this is the ChatGPT channel, I'm pretty sure the limits are smaller than that, especially on the response side. It gets downright pissy if you ask for longer than it's allowed to, but it never says exactly what its limits are.

sonic folio
#

No, ChatGPT 4 has 128k/32k, afaik. My GPTs are huge, so it better.

#

They work, so it must.

sonic folio
#

I think of the 32k token number as the space the model has to complete tasks, provide responses, and use tools.

#

The text you see is just part of the model's output - everything it interfaces with costs output tokens.

ember burrow
#

can ai great an app im trying to build? since I can't find any app developers

#

2 apps 1 games 1 business

sonic folio
#

It can make "Snake".

#

What you can do with it depends on how clever you are with language and math.

ember burrow
#

im not

#

terrible at math

sonic folio
#

You might still be okay but you will need coding experience or knowledge, minimum

ember burrow
#

oh damn

#

would I find any developers in here?

swift osprey
#

I promise you can figure it out if you try

#

it won't be easy at first, but if you try you will get there

ember burrow
#

as much as I would love to I really prefer to pay someone to do the things I can't do well, so I can spend my time on the things I am good at.

white shard
pine tulip
#

Anyone else experiencing problems, publishing the CustomGPT to "everyone"?

ember burrow
#

how do I use chatgpt to make a pitch requesting what I am looking for ?

swift osprey
#

ask it to the best of your ability

ember burrow
#

oh ok I thought it was like a hidden toold

swift osprey
#

None that I am aware of, but I never touch anything like plugins

#

I can't imagine there's anything special/specific for what you are looking for

ember burrow
#

thanks

winged cave
#

anyone have any idea if emailing openai to request to increase gpt4 limit works or not

normal quarry
winged cave
#

thanks

boreal lily
#

is chatgpt down

half pebble
#

its doing something for me right now

boreal lily
#

unable to load history

boreal lily
half pebble
#

its working for me

#

it generating some code

boreal lily
#

are you using gpt4

half pebble
#

yeah

boreal lily
#

its back

#

some short outage

abstract pendant
#

how do I make my pgup and pgdn keys work in ChatGPT?

inland aspen
#

guys is there a setting to make all generated content or responses use British spelling ?

pine tulip
inland aspen
pine tulip
hollow marsh
#

is chatgpt slow for you guys

#

gpt 4 specifically

#

takes like 30 seconds to register i sent something

torpid osprey
#

Has anyone had their accounts compromised? I'm seeing random conversations in Chinese on my account and they are spamming my GPT4 usage.. Sent a bug report to OpenAI, but is there another way to report this/communicate for help?

pine tulip
torpid osprey
pine tulip
#

If such chats pop up again, do a deep scan for malware on your computer

torpid osprey
eternal flax
#

Any developers around?

#

Or someone who is super familiar with weird anomalies from ChatGPT

tight island
dense halo
#

Are custom GPT's limited to 25 responses still?

astral storm
#

What is the difference between this channel 'chatgpt-discussions' and the channel below this one called 'gpt-discussions' ? #confused

ruby peak
#

#gpt-models covers discussions for the API that lets you access other models as well as the custom ones

astral storm
ruby peak
#

I assume it's specifically for custom GPTs then

astral storm
ruby peak
#

It sure is

astral storm
#

And very frustrating when I get timed out for putting the wrong thing in the wrong channel - but no one seems to know what each channel is for specifically anyway.

ruby peak
#

ChatGPT is more an interface to talk with GPTs than a GPT itself

#

And you've got the playground to access the other GPTs

#

At least that's how I understand it

#

But yeah this server has a ton of channels and people just go for the ones that are more similar to what they're looking for

astral storm
pine berry
lucid forum
#

hello just wondering if anyone else has this bug chat gpt ignores custom chat gpt instruction

feral zinc
lucid forum
#

for example even with custom gpts i get this error when chat gpt will refuse to stop using place holder comments i have to keep asking for full code wasting chat gpt 4 time

feral zinc
#

the placeholder thing has been confirmed as a bug but i have not seen its been said to be fixed yet

lucid forum
eternal rockBOT
#

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.

high quail
#

Good morning folks! I've been away from the GPT scene for awhile, started tinkering with it again. I remember a few months ago people were really getting good with the custom instructions. Are there any 'guides' or tips and tricks for them?

drowsy tinsel
#

gpt4-turbo speed is still pretty trash these days

warm arrow
#

When will microphone and conversational access be allowed on a browser?

maiden fractal
warm arrow
maiden fractal
#

Create new convo with plugin, in plug in store of gpt look for tts

warm arrow
#

11labs TTS? Not necessarily the same purpose I want to use it for. I'm looking to use microphone access and have audio conversations like I can on the mobile app, not just ask for specific things made into speech? This way I could have the conversation while I'm doing something like gaming

sonic folio
maiden fractal
winter kernel
#

Ok

mystic zealot
#

Hello! I hope you are doing well. I need assistance regarding accessing my previous conversations on ChatGPT 3.5. Whenever I try to open them, the chat crashes as soon as I start a new message. Should I forget about the old chats and start fresh on the same topics?

sonic folio
#

You can copy them to a local text file for instance.

eternal rockBOT
mystic zealot
sonic folio
restive plover
#

What new features have they added? I haven't used it for a few weeks.

sonic folio
#

Nothing much in weeks. Last update was a while ago but the Custom GPT Store is coming soonTM.

lunar maple
#

every time i use the data analysis or gpt4 model after around 1 minute while the AI is responding i get a network error all the time within while it is replying and it is very annoying as it makes chatgpt4 useless as i can't get a large ish amount of code from it because it always has to start again wasting my tokens and the progress it is giving me because it gives a different response the next time

sonic folio
lunar maple
#

and my internet is fine

sonic folio
#

File a support ticket. That's not supposed to happen.

lunar maple
#

it's annoying because i am working on stuff to do with algorithms and getting the data analysis to help me and when i think it is getting close to working out the algorithm and then it just gets a network error and well the next response i get from it after the error is different so i will never know if that would of got it right or not

sonic folio
#

I bet.

lunar maple
sonic folio
lunar maple
sonic folio
#

Good luck

dense halo
#

I agree, I don't understand why there isn't a release. It would make so much sense to be able to talk in your Apple Watch

rough stirrup
#

is gpt-5 no longer there for others too? i had it all day today and yesterday and now i cant start a new chat with it

sonic folio
#

Well OpenAI owns the trade name, I wonder why.

silver lark
#

Serious question: am i the only one struggling to get good answers from chatgpt? Seems like it is misunderstanding everything on purpose.

median herald
#

Anyone else lose the ability to see previous chats with chatgpt on desktop? On the mobile app I have a left nav bar that allows me to navigate to any previous chat

honest jay
rough stirrup
#

no on open-ai yesterday and today i had GPT-5 New as an option but now its gone

waxen chasm
honest jay
swift osprey
#

Does GPT respond to messages which are longer than the full length of tokens that it will process in a single request?

In other words, is GPT ignoring part of my long messages without complaining and letting me know?

median herald
honest jay
honest jay
median herald
honest jay
swift osprey
honest jay
rough stirrup
#

are you even a mod or are you just answering stuff

honest jay
rough stirrup
#

well youre wrong, plain and simple

feral zinc
honest jay
rough stirrup
#

m8 i am telling you they must have rolled out a test or something because i had GPT-5 (New) and it talked about being trained on an AGI

swift osprey
#

Google for "gpt-5 openai 2024" and my 3rd result is "OpenAI Welcomes GPT-5 in 2024 - Inside Telecom"

#

as in, not out yet

loud sierra
#

0.0

#

maybe thats why so much has broken recently

feral zinc
swift osprey
#

However, if you google just "gpt-5" the top google result brings you to what you were likely referring to, and it is gone now

#

"GPT. New GPT-5. AGI-like AI, excelling in reasoning, personalization, and confidentiality. By DAINIS TKACOVS. Sign up to chat. Requires ChatGPT Plus."

honest jay
rough stirrup
#

no i am telling you i was within the normal openai.com where you find all the normal stuff and there was a thing that said new gpt-5 (AGI) as an option right above where you start a new chat for gpt-4

#

i have been using chat gpt since day 1 i know what i am talking about

swift osprey
#

I know exactly what you were talking about lol. It is the thing I just posted above

honest jay
feral zinc
rough stirrup
#

but it wasnt like a plug in it was there as a feature

swift osprey
#

I even had the GPT-5 added to my plugins list

feral zinc
#

you can easily check your browser history for the link

swift osprey
#

or whatever the left menu shows you

rough stirrup
#

its in my browser history hold on

#

it wont let me put any images in here

#

but i promise i havent installed any plug ins and it was there and now its gone

feral zinc
swift osprey
#

purty sure thats the one lol

rough stirrup
#

no its: c/76e8c162-d5cc-4ece-bacb-9340d08cdbf5

swift osprey
#

lol ugh

feral zinc
#

anything under c/ is just your own chat threads

rough stirrup
#

well it was one of my threads

#

i still have the thread open

swift osprey
#

gpt-5 will not come from this conversation, that is a guarantee anyways lol

feral zinc
#

then its made by you or you clicked a shared chat and continued it

rough stirrup
#

well sure but i liked it, it was a lot better than 4 lol

waxen chasm
swift osprey
#

lol okay so google this 'I just created a New GPT-5' and you'll find a linkedin post from the creator showing it off in a screenshot. It'll be the same one you're talking about

put it in double quotes when you google it

honest jay
rough stirrup
#

but its weird it was just a tab right above where it normally says Chat GPT there was like a built in thing for it

waxen chasm
rough stirrup
#

i work in data engineering and currently im working on a task of using a bunch SQL queries, and typically it kind of sucks at it especially when they get complex but the 'gpt-5' was able to handle very long and complex queries easily

#

the speed and accurracy was so much faster i feel like

#

but now its all gone 😭

swift osprey
#

Are custom GPT's just custom instructions that somebody else sets up and they publish them in a way that others can use them without knowing what they are?

rough stirrup
#

according to the linked in post you sent it is also trained on different data and uses AGI traingin

#

as well as custom instructions

swift osprey
#

yeah I was wondering how the data training part works that he is referring to

honest jay
swift osprey
#

interesting. Sounds like it might be worth checking out other custom GPT's. I thought they sounded like nothing more than hype at first

wraith void
#

god chatGPT 4 has been so slow lately and can hardly complete anything

brisk dirge
#

does anyone else have issues where chat gpt stops writing and keeps thinking ?

ancient oak
#

Anyone having issues with GPT Assistants today? It's barely working for me again.

fluid skiff
#

i just had to wonderful experience of having a 5 hour conversation with chatGPT working out the details of a project reset to the 1 hour mark 😡

sterile flame
#

Yeah... I've noticed that they have completely shut down web searches atm. At least from my access point.

#

... probably recalibrating...

odd flare
#

any updates coming to GPT4 ? havent heard of any updates since a few months already or a year

cobalt tree
#

i want tool calls in gpt 4 🫴

honest jay
odd flare
#

im talking more of an improved version of chatgpt, liek 4.2, 4.5 or something

honest jay
bitter latch
brisk dirge
odd flare
bitter latch
odd flare
bitter latch
little storm
#

Wish the speech function would play louder, it seems to go into a different sound mode on Android app and always seems too quiet

strong flame
#

did I hear correctly that video analysis is coming? does anyone have any idea when? I need to be able to transcribe videos for my workflows and I have zero developer background to be able to add this as an action. Unless someone could help me.....

narrow viper
#

I've tried several times in different ChatGPT (Plus) sessions to get it to generate a complex RegEx. It simply doesn't get it right and then fails on verification with "Error in input stream" and a warning about "unusual activity" ... yeah, my prompts have a couple paragraphs of detail. I've tried to get it to break down tasks, avoid running Python and other ways to make it easier on the poor critter. But it's just not up to it today. 😦

hot vapor
#

hi what is the "memory off" thing

brisk path
dense halo
#

gpt got memory

hot vapor
#

yeah i saw the pop-up but i have no idea what it means

mental thunder
#

How to make money with gpt

hot vapor
#

like it remembers stuff between chats?

pine tulip
hot vapor
pine tulip
hot vapor
#

niche it down, niche it real down. market on pinterest

pine tulip
hot vapor
mental thunder
#

Anything else is free

pine tulip
#

Which you can use with ChatGPT

mental thunder
timid cipher
#

How do you get the remember stuff between chats work I just tired it and it said it dosen't have access to the previous chats

#

*to work

sly geode
#

Hot mods plugin Is made by open ai?

#

It's primary use says it modify your images means existing irl images right then why does it spits out fully new recreated from reference and don't edit on existing

#

Or am I getting the meaning wrong of it

dawn fossil
sick pecan
lime sky
#

Has anyone else recently lost their ChatGPT Plus account without any notice?

lime sky
dense halo
#

My chatgpt got a update

#

I can now do temporary chats and my chatgpt can learn from normal chats

waxen chasm
dense halo
# waxen chasm Explain??

Improve responses with your chats

Your GPT will continually improve as you chat, picking up on details and preferences to tailor its responses to you. Learn more

To modify what your GPT knows or teach it something new, just tell it:

“What do you know about me?”
“Where did we leave off on my last project?”
“Remember that I like concise responses.”

pine tulip
dense halo
#

Yep that too

sick pecan
#

maybe its still rolling out, because it doesnt seem to work

hazy crystal
#

I got the information about the update and I am unsure if it pertains to GPTs or ChatGPT

west fulcrum
#

Hi everyone. Has ChatGPT removed turning off chat history? I can't find the option anymore.

west fulcrum
#

I'm trying to remove chat history and opt out of machine training--previously I did that by turning chat history off

pine tulip
west fulcrum
#

Hm, I can't seem to find it, mind sending me a screenshot? Thank you so much, sorry for the trouble

#

I see ChatGPT, and one of my personal ChatGPT models, and explore.

#

I assume I can't do a temporary chat of my personal ChatGPT model? Sigh

pine tulip
dense halo
#

Just loaded chatgpt, and it won't work. Getting "hmmm... something seems to have gone wrong." every time I send a message

#

Only happening on firefox though... Chrome works. Any idea?

marsh pivot
#

i think its down

pine tulip
dense halo
patent hinge
#

GPT4 is down

dense halo
#

I see this in the console logs. I don't get these three in chrome, only firefox

patent hinge
#

3.5 works

dense halo
#

If it's down then why working in chrome?

marsh pivot
#

not on my chrome

dense halo
#

Oh... I'm using GPT3.5 in chrome for some reason haha. My bad.

fluid skiff
#

it has become unusable

#

4 is no longer the standard
too bad

sour wasp
dense halo
#

whats the standard?

dapper isle
#

The new feature didn't come for me, I can't see it maybe it could be a bug?

dapper isle
pine tulip
dapper isle
#

Ooh it was a mistake

dense halo
pine tulip
dense halo
#

Do we know whether it will end up remembering key info in chat sessions we end up deleting later? Ik the feature has been removed now but just curious

pine tulip
dense halo
#

I'm talking about normal chats

pine tulip
#

At least thats what it said when it was available

pine tulip
dense halo
wide pike
#

what happens if you wait 1 whole hour and it still says too many requests in 1 hour?

#

Cuz that's what happened to me

#

I even tried resetting cache

pine tulip
noble yoke
#

Hi guys. Have a question abt chatgpt

dapper isle
#

my sub card refused to pay for some reaosn even tho i have sufficient balance why is that? the app is tell me to use another payment method

noble yoke
#

I can find 4-months-old chat history in the left side bar of chatgpt, OpenaI blog says it only sotres 1month, but it looks like every chat history is saved unless I delete it myself

#

Anyone knows about the fact?

dense halo
#

i hope they fix gpt 4 its become very dumb

dense halo
wide pike
#

I don't know what it means

full cargo
#

Hello.

wide pike
dense halo
noble yoke
mint steeple
#

what about the gpt store?when it will be realesd?

maiden fractal
noble yoke
#

Guys, I got a question. Does the chat history stay there forever, or gets deleted?

maiden fractal
#

Didn't have my chat deleted since I have sub .

#

About a year

paper light
noble yoke
noble yoke
maiden fractal
noble yoke
maiden fractal
#

Like I said since I started to use it

stable bronze
#

why 4.0 is soooo slow now

#

he can barely answer very simple questions

maiden fractal
#

Working good here

stable bronze
#

😤

noble yoke
maiden fractal
noble yoke
#

Thanks i love it

knotty tulip
#

4.0 slow here as well, probably high-demand

lime sail
restive kernel
#

ChatGPT4 flakey for anyone else?

#

Service status says it's ok but my experience is awful

stable bronze
#

raise more EC2 instances there adm

#

chatgpt is tired

restive kernel
#

Service status is bunk. Says green when 3/4 requests are failing for me. Are you having the same issue @stable bronze ?

stable bronze
#

it's slow and failing, mostly

red crag
twilit hatch
#

I feel so confused by the new memory feature I saw pop up on my screen when I logged in and announced because it doesn’t seem to be working. Is it just about remembering prompt patterns or something?

tropic sparrow
#

Keep getting errors over and over

honest jay
twilit hatch
#

Does anyone have a screenshot of those new elements? It looks the same as it always did to me

tropic sparrow
#

Never experienced this before, takes like 5 times for the prompt to work

twilit hatch
#

I do, above anything else, dearly hope gpt-4 starts to ease up on its hanging and crashes soon, memory or no memory, it just sucks how slow it is

brisk path
#

I want them to bring back the old personalization menu

#

Aka My ChatGPT

honest jay
zinc valley
#

when is the gpt store launching

honest jay
zinc valley
#

so until they feel like it

twilit hatch
honest jay
tacit vortex
#

hi, my plus plan always only return error. thx

honest jay
tacit vortex
#

i know ukraine can share video in discord, i dont see how i can. trying. thx

honest jay
inner knoll
#

Is anyone noticing chatGPT demonstrating peculiar behavioral heuristics that are analogous to human biases and such? I feel like I've personally noticed, when it's identity isn't established, it is more likely to follow prompts that may misalign with it's values/identity - however if you prime the model by asking about the values, it will be more resistant to a change in behavior - this is similar to social psychological principles of cognitive dissonance, identity salience, and polarization, for example - has anyone noticed any others?

strong vessel
#

is chatgpt down? Mine couldnt even reply to a simple hello

pallid schooner
#

I keep getting this message "Hmm...something seems to have gone wrong."

strong vessel
pallid schooner
#

yes, gpt4 has the problem. 3.5 is fine

maiden fractal
#

Working fine here

strong vessel
pallid schooner
#

That URL shows its okay, but its not.
GPT4 app it shows the following message "The server is having problems. Please try again later. (8435ec817d105719-SYD)"
GPT4 web browser, it just gives the message "Hmm...something seems to have gone wrong."

honest jay
#

"Hmm...something seems to have gone wrong" is sometimes caused by a VPN, or a browser extension. You might try closing all VPN processes, or trying Firefox if Chrome has the problem.

pallid schooner
#

I've been using VPN for a long time with GPT, never had such problem. However I've disabled VPN and tried it again, got the same error message.

#

I then moved VPN to OHIO US, and then tried it again. I got an answer, but it was very slow. Seems to be a major problem with a 'server' in Sydney?

#

woops now I got a different message "The server is having problems. Please try again later. (8435f9d5acdf61ac-ORD)" - still in Ohio USA

rancid seal
#

Does anyone know what the message cap is for a Teams membership?

rough estuary
#

GPT just updated!

maiden fractal
rancid seal
#

I upgraded, it's 100 messages every 3 hrs.

waxen chasm
digital prism
odd sinew
#

@rancid seal what happens to your conversation history when you upgrade?

#

and existing GPTs?

tawny crater
#

ChatGPT 4 degraded again, is there an update? It always does this when there is an update

maiden fractal
#

Some can see it in plan payment on setting

tawny crater
#

Ah okay, thanks, I'll avoid it for now then. It just makes me mad otherwise

fervent saffron
#

I see Team option in mine - any info on how it works though? i see that it's $30/seat and you can add multiple seats - but if I do that, are these "new" accounts or can I "invite" an existing GPT user to my "team" and they retain all their history?

lost raptor
#

I find it more annoying than anything that chatgpt can't use the newest open ai documentation.

odd sinew
gentle bronze
#

hey everyone, I just got the option to upgrade plan to team, anyone can point me out to the caps and limits of that plan?

fervent saffron
#

the team pricing goes to 404 on the openai site. not sure why they'd rush it out to the front end without having the other stuff prepared 🤷‍♂️

#

@rancid seal - you upgraded to Teams tier? can you see my question about adding members to the seats? can they be existing gpt accounts that you "invite" or are they new accounts although for the new seats?

tawny crater
#

I would just wait for announcements to be safe honestly.

fervent saffron
#

fair enough

tawny crater
#

I don't want to risk my stuff either

rancid seal
# odd sinew <@164971465853108224> what happens to your conversation history when you upgrade...

It's basically like a new account. You don't have your history of previous chats in the teams account. You can switch back to your other account by clicking your profile icon at the bottom left.

You can send invites to anyone on your team by filling out their email in a simple form. They don't have to be existing paid/free users. They'll have a new account under your team set up.

fervent saffron
gentle bronze
#

so this would be parallel of your normal plus account?

odd sinew
dapper isle
#

can someone tell me whats a team plan that i can see besides the plus plan?

#

does it has the new memory feature?

rancid seal
#

@fervent saffron Use their same username*

odd sinew
#

Hmm... very interesting. Especially in light of the new memory system.

gentle bronze
#

oh, that's what I wanted to know

#

💳

odd sinew
#

Thank you so much this was important information to have.

worthy dragon
#

Is the long term memory between chats only available to paid users ?

rancid seal
tawny crater
#

Wait so I am confused, does this mean you have to pay for both? Or just teams and you get both?

fervent saffron
#

thanks jon for the username clarification --- is it a 100 per 3 hours for the workspace - or 100 per 3 hours for each user in the workspace, just want to make sure i am following

rancid seal
#

100 per 3 hours for each user in the workspace.

fervent saffron
#

awesome thank you sir!

lethal grail
rancid seal
slate berry
#

Do you know what that means for Dall-E 3 outputs? I know its rate limit is slightly different

rancid seal
slate berry
#

Sweet, thanks!

waxen chasm
#

3 million GPTs created! Insane

amber lintel
#

So, because I'm not a tech person, what does an individual GPT mean?

dry snow
#

Again, you need to pay :p

plucky timber
#

Ngl the reason imma open the store today is to see if they added instruction protection :kek:

dry snow
#

💀

digital prism
#

GPT store is kinda blowing my mind.. ngl

inland solstice
#

Honestly dont care about GPT-Store just fix the 'lazy' update they added to GPT-4 where its just crap at long-winded tasks like coding

zinc valley
#

Wait so the gpt store just launched

amber lintel
#

When it says you are able to purchase a GPT, what exactly are you buying?

rose glade
#

on the new team plan, is data analysis still included on the lower end plan?

dry snow
amber lintel
#

Again, I'm not very knowledgable and want to learn more.

plucky timber
worthy dragon
acoustic marsh
#

GPTs are available to free users now?

rose glade
amber lintel
#

So you'll have more control over what it can do compared to just using normal ChatGPT?

plucky timber
ionic geyser
#

Seriously we need to subscribe for TWO accounts to have any privacy?? What the F.

foggy magnet
plucky timber
#

The basic concept of the GPT store is people selling those custom made GPT's

rose glade
somber hull
tawny crater
#

Just a forewarning to everyone You have to pay for 2 people at minimum. if you can't get anyone to split the bill with you that is 60 dollars a month.

sly geode
#

woah woah woah

lusty island
#

Is the gpt store being rolled out now?

digital prism
plucky timber
foggy magnet
rose glade
ionic geyser
plucky lava
#

hu'

#

hi

tawny crater
zinc valley
#

How do u access the gpt store

tawny crater
#

Read the fine print

plucky lava
zinc valley
tawny crater
#

You're either going to pay 500 dollars right now for 2 people or 60 dollars a month for 2 people.

bronze mulch
glad zinc
#

how do you make an ai for sketch to image

foggy magnet
tawny crater
#

Sorry 600 dollars not 500

plucky timber
#

60 dollars for what higher limits though, any clarification on the supposed higher limit

#

am i getting 2 extra messages ? 20 , 200?

tawny crater
plucky lava
foggy magnet
#

Okay, so instead of making assumptions you could just say that. The fine print doesn't clarify until that point.

wise yacht
#

Store not loading for me, but it does offer the teams upgrade. Anything I need to do to get access to the GPT store? Right now its still recently used list

foggy magnet
azure moat
#

What happened to those messages from last night saying "ChatGPT will remember more about you etc etc" Did I dream that?

plucky timber
tawny crater
foggy magnet
#

What is the cap?

zinc valley
#

So did the gpt store launch

foggy magnet
vernal ingot
#

Here it is: Builders can earn based on GPT usage
In Q1 we will launch a GPT builder revenue program. As a first step, US builders will be paid based on user engagement with their GPTs. We'll provide details on the criteria for payments as we get closer.

rose glade
foggy magnet
rose glade
#

you need to translate the dynamics of the outlines of the image into the photograph

reef cloak
#

Did the GPT store release or did they just announce it? If so, please give the URL.

glad zinc
wise yacht
#

I only see the recently used list on mine

glad zinc
#

basics

foggy magnet
opal yoke
#

it's the same openai/gpts, but it's still showing the old "My GPTs" landing page.

foggy magnet
wise yacht
foggy magnet
#

OpenAI does not do product launches properly, ever

#

just treat the announcement as just that, access comes later

wise yacht
sterile glacier
#

Restart your brain, you are hallucinating.

foggy magnet
glad zinc
#

give

vagrant ginkgo
#

I’m about to use teams as a family plan lol

foggy magnet
#

Literally me

#

Someone has to take 1 for the team and find out the cap, that's the make or break.

reef cloak
#

If you're in a computer, just keep the GPT's page open and keep updating it. If it changes, tell everybody.

vagrant ginkgo
foggy magnet
#

Teams gets 32k context out the gate smh

#

Damn 500 error when I try to look at the pricing page.

opal yoke
rose glade
reef cloak
#

This is one of the most messed up product releases I've seen. They said that it released, but it didn't.

plucky timber
maiden fractal
plucky timber
#

Can anyone find a fine print on the teams purchase or am I buying a mystery box?

zinc valley
#

so what is the update that just happened? THe gpt store and the new payment method is all i see

opal yoke
#

the gpus are burning again

maiden fractal
#

Some people said 100 prompt /3h can't confirm if real or not

zinc valley
plucky timber
#

We need that info upfront 😛

foggy magnet
plucky timber
#

You cant just sell "increased" :kek:

maiden fractal
warm gull
foggy magnet
#

100 per person is massive

warm gull
#

no that would be per TEAM

zinc valley
#

so what else is new with the update

foggy magnet