chatgpt-discussions

[drowsy tinsel]

reaches limit when building gpt </3

[red saddle]

Happens for me all the time. Very unfortunate since it hinders the creativity when you're in the flow.

[drowsy tinsel]

its annoying because when building gpt u need lots of testing

[red crag]

Do account custom instructions interfere with personal created gpts instructions?

Or gpts instructions get proprity and skip account ones?

[raven slate]

theres an option to disable it for new chats

[red crag]

yep i know

so they indeed interfere

[red saddle]

Can I only use 1 API for each GPT?

[violet bloom]

Moi everyone, whats the current limit? Still 50? 100? With the new UI there is no real visibility

[wanton thunder]

I don't want to make any efforts to gain money in real life

I don't want to work, I don't want to study

[slate root]

Is there anyone familiar with adding an API to a custom GPT?

My goal is to have a GPT connect to my google docs. For example, I want to have a google docs that has my goals saved in it. I'd like my GPT to get that document when needed to see if any changes or progress has been made to my goals. Any idea on how to do this? I'd prefer to not use Zapier.

[patent plank]

anyone having problem accessing chatgpt?

[dense halo]

it's touch and go with all the traffic still

I sometimes have to try 3 times and finally get result

[patent plank]

Can't access chatgpt. And it is too slow for last few days. Almost unusable.

[dense halo]

all the world is trying to use it. it's like a launch. give it some time it will smooth out

[brittle junco]

The "knowledge" section for creating a GPT, what does that mean exactly? Does it fill the context window? (which hasn't increased, I suppose, since GPT 4 Turbo is only for developers), is it some kind of fine-tuning? I can't figure it out

[patent plank]

Seems, People using it more than Google

[supple zephyr]

are they Denying of Service

looks fine ig

[simple blade]

Does chatgpt 4 still have limited prompt?

[crisp cairn]

you would need to write your custom action, or wait for some google docs integration

[pine tulip]

Because most of the time its better than google

They just invented google 2.0

[normal wren]

been trying to understand this too, doesn't seem to actually do fine tuning or add to the context window. More to be used as a database that can search for certain key facts but couldn't use the whole thing as context (e.g. if you were writing a book or story it wouldn't have all the context but if you wanted to look up certain commands from a code library it could do that)

[pine tulip]

You need to tell it where to accesse its knowledge, then it will

[brittle junco]

Mmh I see, it's a lot less interesting now...

[pine tulip]

„If a user asks about our solar system, use the file solar.txt in your knowledge“

[normal quarry]

no advertising, please see #server-rules

[pine tulip]

Wrong channel

[livid pier]

I think I once posted a wrong answer here about the current usage limit of GPT-4

Currently it should still be every three hours 50 times!

Because I just reached the limit (oh, this really makes me sad)

😭

[drowsy tinsel]

it's so nice that in GPTs it can take all the time to correct the error when creating code

unlike in the past due to output length limits it stops

[dense halo]

Sorry, I've just did the subscritpion for Chatgpt4 and after a few I received a message that I need to wait 3 hours? is there any cap EVEN IF I PAID?

[normal quarry]

the cap exists for GPT-4 and we think it's 50 messages per 3 hours

[dense halo]

TERRIBLE....almost unusable paying the full subscription....

[normal quarry]

I never hit the cap, you must be working it pretty hard

[drowsy tinsel]

getting network errors a few words into generation...

u guys getting it too?

[normal quarry]

working fine here -> https://status.openai.com

[dense halo]

anyone knows how to obtain a PNG image with transparent background? it generates a picture with the classic chackboard background that should represent trasnaprency but at the end, when you open it in photoshop, it is not transparent at all...

[static cedar]

Does someone know if Custom Instructions are also applied to GPTs or only to the main ChatGPT?

[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.

[rose glade]

are GPT's down at the moment?

[drowsy tinsel]

appears fine but im getting elevated rate of errors

[normal quarry]

may be something local to you

[rose glade]

I figured it out, I was feeding the GPT too much data

is there a limit? I only fed it around 780k words

[swift swallow]

Only main i believe, when generating a GPT some about you might be used though.

or your about you might be accessed when you have a conversation or both

[rose glade]

Also can you add custom instructions on top of using GPT's?

[hidden ibex]

Where can I look to see current rate limits?

[swift swallow]

What do you mean? they already beeing set when you set them up and can be changed through edit.

[rose glade]

The custom instructions still apply in the settings, so you could add them on top when you hit the word limit in the GPT?

I can't afford to train my own models, so these GPT's are really good like this, but not 100% on there limitations and how they interact with the different parts of GPT

[neon tinsel]

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

[solar oriole]

same here

down?

[neon tinsel]

maybe

[buoyant berry]

yes

[swift swallow]

Not sure I can follow your process there my man. So, you got your about + system message (CI) for your "Main GPT" and then you have GPT's which are basically separate narrow instances, which have their own instructions and document embeddings + actions. Im not sure which limitation you are referring to.

[snow latch]

Do you guys know how to change the voice used by the new custom GPT function? It’s default setting is Sky, and I like Juniper better and set my app to Juniper. But when I use the new GPT it automatically switch to another voice

[solar oriole]

come on, I was about to finish my homework

[buoyant berry]

got it to work if you just retry

[solar oriole]

nah, nothing loads anymore, gray screen

[buoyant berry]

mine is working again

[solar oriole]

it might be partially down

oh, works again

nah, doesn't

[wide rapids]

Mine is kinda broken right now too. It won't load my past conversation threads

[solar oriole]

It works again but like reaaally slow

[twilit verge]

anyone else custom gpt not working?

[agile spruce]

Hi, anyone have an issue same like me?
I already subs the chatgpt plus, but I can't access it, when I click gpt-4, always show pop-up upgrade to plus, when I click it will redirect to manege my subscription

[stone helm]

Assuming it just went down for alot of people, getting the "Hmmmmm....... Something went wrong"

[steady path]

is there a limit on how many files can be uploaded for a gpt?

[normal quarry]

I have seen a few numbers quoted, 10 and 20 with a maximum file size of 512MB

[hearty shadow]

Based on the APIs docs, they will charge per GB per day and limit per file is 512MB, so you could have multiple files uploaded, I saw a reference to 50 files in some community posts.
https://platform.openai.com/docs/assistants/tools/knowledge-retrieval

[rose glade]

in GPT's how do I ensure its output is as code?

[digital prism]

system instructions

[rose glade]

as in 'instructions' box because that is where I am writing it

[digital prism]

yes, but you could just ask the left-hand side GPT-builder to do it for you

[rose glade]

GPT-Builder?

[digital prism]

you're in the GPTs editor right? there's a GPT builder on the left side

[rose glade]

I dunno lol

[digital prism]

then i can't help you further 😄

[rose glade]

it won't let me paste screenshot here

so I can't show you what I am seeing

[digital prism]

other channels allow it, try off topic

[rose glade]

pasted there

if I share my GPT can people see the knowledge I have uploaded to it?

[normal quarry]

yes if Advanced Data Analysis is enabled

[rose glade]

Data Analysis = Code interpreter? Sorry I am getting quite confused between the two haha

[normal quarry]

yes, same thing

[rose glade]

because on the custom gpt it says Code Interpreter but when it processes it does data analysis

they should change that, its confusing a bit

especially when you are aiming for accuracy

[normal quarry]

hopefully they pick that up and fix it soon to avoid confusion

[arctic hamlet]

After a certain time displaying answer, chatgpt gives a "network error", do you guys have the same problem ?

[steep forum]

openai.com down?

[normal quarry]

it's up -> https://status.openai.com

[steep forum]

I couldn't get the website to respond, it blocked the reqeusts.. weird, now it's working, thx

[static cedar]

Thanks!

[calm hedge]

Yes, seems to be having issues again

[lost raptor]

anyone else just getting nothing but placeholder code since the update?

[steep forum]

Myself I am continuously getting

An error occurred. Either the engine you requested does not exist or there was another issue processing your request. If this issue persists please contact us through our help center at help.openai.com.

[drowsy tinsel]

looks like standard elevated error rates due to high load

[calm hedge]

Had this aswell, I was lucky enough to have the reply contain all methods with a placeholder implementation, though. So I just re-ran it specifying to only do those specific methods that time.

[modern spoke]

What's the max character limit for the custom GPT instructions? Seems to be much more than custom instructions. Are they fine tuning?

[outer seal]

ive been getting that for a while pre-update but then again who knows when they started A/B testing turbo

"oh i have 32k context now? sorry openai says tokens are expensive, i'm only gonna look at 500 characters of your code" (and/or fill in placeholders)

jokes on them they arent saving any money when i hit regenerate 20 times

[lost raptor]

i'll wait to see if anyone makes "gpt" coder thats actually competent.

[modern spoke]

the best way to do that is to give it context for your code. for each of my project files, I have custom GPTs

[outer seal]

im gonna have to play around with making a coder gpt, even when i pass project files it'll (usually) only load snippets and make assumptions

[modern spoke]

yeah, that's why I was asking about the character limit, because I've been able to fit massive amounts of documentation and pseudocode into my instructions. seems to work better for recall than the file upload retrieval. which makes me wonder how it works...

[outer seal]

32k now

[modern spoke]

so it just eats at the context window array?

[calm hedge]

Im pretty sure I just found a way to bypass the chatgpt per-hour request limit. Where should I report this?

[outer seal]

i guess the code interpreter tool could still be tuned for old 8k ctx ada 🤷 really annoying how it hates to look at the whole script

[outer seal]

report it to me please

[steep forum]

Lol

[broken fractal]

man GPT-4 is quite good at text adventures. I even set it up to generate a Dall-E image when there's something of note to visualize

[outer seal]

the retrieval tools use embedding search to load relevant chunks into the context, but 4-turbo is so good you can just shove whatever into the instructions yourself (as long as it fits) and get nearly perfect recall from tests i have seen

[brisk cobalt]

I hope 2024 Will have an AI TV

e.g.
Create a news channel for china

Alexa once organized a surprise party for herself and successfully managed to surprise everyone, including herself

[foggy grove]

What 😂😂

[lime depot]

Do any Non-Plus users have access to Custom-GPTs yet? Is there a timeline for this?

[silver marlin]

can chatGPT 4.0 plus actually contextualize long pdfs or does it just form a main udnerstanding and act like it knows everything by providing made up/estiamted quotes/analysis??!

cuz I heard the data anayalsis mode is actually made to understnad PDFS, and just skims it and forms an understanding

[elfin birch]

From what I've experienced it will search long PDFs and put whatever it needs into its context

[silver marlin]

like does it still make up quotes in pretty long pdfs (20-30 pages), cuz thats what it used to do

[elfin birch]

I haven't seen it do that

[silver marlin]

regardless, it still feels like it always forming a general understanding and cant actually access the whole pdf at once

[elfin birch]

The latter would cost a lot but it's not like it can't get specific information

[silver marlin]

apparenlty there was a pdf mode that was released to influencers, which probably was able to actually contextualize the whole thing

[silver marlin]

like how would it know where to search for quotes if it isnt able to contextualize the whole thing at once

so it prolly just takes quotes from random palces

[elfin birch]

Algorithm, not an LLM

[silver marlin]

not sure what you mean by this

[elfin birch]

Searches keywords and related text, extracts and puts into context the information that relates to the former

[silver marlin]

hmm got you, do you know how to gain access to the supposed 128k token limit stuff, or gpts that can actually contexitulaize large amounts of info at once?

[azure lynx]

Recently dall-e 3 is not generating multiple versions of the prompts. Is the policy changed like today or yesterday? And what I the current limit for messages per hour and images per day? Can that be read somewhere?

[elfin birch]

Use the API

[silver marlin]

playground?

[elfin birch]

Yes

[silver marlin]

thanks 🙂

[elfin birch]

It will cost you though

1 cent / 1k tokens means $1.28 for one prompt with 128k tokens

[patent plank]

Google is trying their best bet it and messing them up 🤔

[silver marlin]

oh damn

[silver marlin]

i heard gpt 4 turbo kinda sucks and isnt as smart, is that true. If so are they still sorting that stuff out?

apparently they butchered its brain at the cost of efficiency?

[elfin birch]

It's about the same
Most of the time people don't have proof and are just bad at prompting

You can see for yourself

[silver marlin]

alright thanks again !

[tardy oasis]

yeah tbh i haven't noticed worse quality, the speed improvement is amazing tho

[boreal lynx]

Anyone noticed a slowdown in response times on GPT4?

[elfin birch]

A lot of people using it 👀
100M active users + whoever came in from the Devday keynote is a huge number

[pine tulip]

I guess even more in the last few days

Probably like 150-200m users lol

[silver marlin]

yo @elfin birch so i just added some money to my playground account, but im not sure which "playground" version i should be using (chat/assistant/complete/edit) and how to use it. Could you possibly guide me?

[elfin birch]

Chat for the latest models

[spice star]

it creates an "agent" with the job of finding relevant sectons of the PDF which are then added as context

similar to web browsing mode

[glass hearth]

i did and its supposed to be faster so not only the quality of responses degraded but the speed too which was the whole point of gpt4 turbo

[glass hearth]

nah man its true i wouldnt say its massively different but if you have been using gpt consistently u can easily see it

[dense halo]

Does anybody knows what's exactly the update of GPT? From Sama's tweet

Like... it's intelligence, context, limit, what?

So it's not possible to make a custom GPT that has a plugin added to it?

[sweet robin]

It’s sentient Siev

[dense halo]

Omg O_O!!!!!!!!!!!!!

[topaz otter]

Are any other Android users missing the File Upload button?

And can't choose between 4 and 3.5? Been like that for me since the new UI came out..

[pine tulip]

just another reason to switch to iOs lol

[dense halo]

Sam Altman said there is a new version of turbo in ChatGPT

[ornate zealot]

It is nice going through old Reddit posts about people complaining that GPT can’t do math, then taking their equations and watching the current version get it right.

[spice star]

yeah the full version of GPT-4-turbo isn't released yet, it's expected to be faster than the current one doesn't have

at least from my understanding

might have a higher context limit but don't quote me on that

[ornate zealot]

From what I tested, it seems like it does

[pine tulip]

it's still no where near the level of wolfram alpha

[dense halo]

What’s the new context length

[pine tulip]

over 9000

jk, idk

[dense halo]

Yeah for sure last one was 32k

[spice star]

supposed to be 128k but idk what context length ChatGPT will end up getting

[ornate zealot]

Well, that plugin exists for a reason. And besides, Python seems to be good at doing it anyway

[spice star]

the highest one is usually reserved for enterprise users

[ornate zealot]

Does the context length include my prompt and its responses

[spice star]

it includes the entire conversation and any context (like instructions)

[dense halo]

Yes

It also includes openai instructions which is around 3k I think so the 32k context length in practice is like 28.5k

[wise anchor]

Was there an update today to the model?

[spice star]

not that I'm aware of

[ornate zealot]

I am going to have to make a long story without images in it. So I can share them. Because somehow how it knew the beginning parts of the prompts of my conversation, even though it spanned the length of 73,000 words.

[spice star]

I think it tries to intelligently determine what to include in the context before responding

so it's not a hard limit that everything after 8k tokens is simply ignored

[ornate zealot]

And that was before the big update, which I did try then, and get it to tell me the beginning, but it couldn’t.

[dense halo]

Yeah according to Sam’s twitter

[spice star]

I think these context lengths are still accurate: #chatgpt-discussions message

oh not if there was a model update

[spice star]

what did the post say?

[ornate zealot]

I’ll have to test it some more to see if maybe they didn’t accidentally give me 128k

[dense halo]

there is a new version of GPT-4 Turbo now live in ChatGPT; you should hopefully find it a lot better!

please let us know what you think.

Someone test the token

[spice star]

probably the faster model they promised then

[dense halo]

Ye

[brisk path]

What's new in the model update?

[ornate zealot]

And 32k is estimated to be around 24,000 words

A token being 0.75 words

I might see what happens if I surpass 100000 words, if then it can remember some part in the beginning.

[dense halo]

Google is the best wanna know why?

They know everything about you, place's you've been, digital footprint, what you like, what you search for, what you do in your daily life, etc.

So they use that information to create their services and improve upon them.

[ornate zealot]

Typical data collection

[dense halo]

Yes

[dense halo]

Why bard so bad

[dense halo]

Right, and it just keeps getting better and better.

[dense halo]

Their current model is bad, PaLM-2.

They need to update it and research more.

Right now it just lies to you and really just isn't good.

Currently Bing is the best AI.

Bing gpt

GPT-4 isn't the best because you gotta pay for it. But if it was free I'd say GPT-4 is the leader.

Bing is gpt 4

Right, but GPT-4 on the official Chat.OpenAi.com website is the best.

[ornate zealot]

ChatGPT you mean

[dense halo]

It has the best whatever you call it.. parameters?

It is the best at not lying.

It understands it's limitations the most.

And it's not annoying asf like Bing or Bard.

But if it was free I'd say GPT-4 is the leader.

Google was involved on the ChatGPT project aswell.

If that helps you understand that Google is the leader.

[brisk path]

The what

Bing is horrible

It's not even GPT-3.5 quality

The joke of a GPT-4 model they're using is inferior to even Claude 2 and Bard

[dense halo]

I was wrong actually, forget everything I said about "free".

GPT-4 on chat.openai.com is the best.

I don't know why I said that, ChatGPT on top. it leads in AI.

[ornate zealot]

GPT-4 Turbo is the current best

Wonder how it differs from Bing AI now, other than it being paid vs free.

[brisk path]

Bing is horrible

Why can't people just understand this 😭

The quality is horrible and creativity, reasoning, and just general smartness isn't there

Also, Bing doesn't have code interpreter

Even Bard has a version of it

[ornate zealot]

I was wondering about that

[hidden ibex]

You can paste your text into the tokenizer on the openai page, their updated tokenizer is more efficient, so you get more than 1 word per token on average

https://platform.openai.com/tokenizer

Even though it still says .75 words on the page, I'm finding >1 on a lot of text lately

[ornate zealot]

Alright

[elfin birch]

Just checked in #off-topic, they're 100% accurate

[ornate zealot]

I’ll have to check when I get home

[eternal rock]

`` Server Tip: ``

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

[weary linden]

Does anyone here know of a website or other resource that has gpts to try?

[ornate zealot]

You can try Poe

It had gpts for a few months now

It isn’t GPT 4 though, as far as I’m aware

[weary linden]

I know openai plans on releasing a "store" but I was wondering if there are any websites that are currently available

[elfin birch]

There's #1171489862164168774

[pine tulip]

I just got a pop-up (unfortunately didn't screenshot it), which told me the exact instructions someone used in their GPT with a Remix button for me to click, to copy this exact GPT. Now I can't seem to get that pop-up again. Any ideas on how to get it again?

[weary linden]

I recently saw a live commentator that someone made. It can commentate sport events live, in real time

[wanton lichen]

So has anyone noticed any increases or decreases in quality for programming with chatgpt4?

with the newest updates

[weary linden]

Thank you, didnt notice that

[ornate zealot]

Wonder how that works.

[weary linden]

Thank you, but im looking for the gpts that was just released in last update

[ornate zealot]

Oh, you mean OpenAI’s GPT’s

[weary linden]

I cant paste the link but on YouTube look up video titled "How people are using the new ChatGPT Upgrades" and the channel is 'The AI Advantage'

The live commentator discussion starts around 40 seconds in

[pine tulip]

any ideas? :(

that would be huge

[ornate zealot]

Oh, that is using GPT-4 Vision, except for a video at a frame by frame analysis

Wonder when they’ll add video support for vision.

[weary linden]

Yea I didnt know it could analyze that quickly

[weary linden]

No wish I could help ya

[pine tulip]

sad

but thanks

[slow copper]

Does GPT4 still have the 50 message limit

[shrewd yarrow]

Anyone able to tell me why the more I chat with gpt-4 the laggier it gets for its replies?

[pine tulip]

could have to do with the context size

[hybrid oar]

Hey, it seems GPT-4 Turbo is live on ChatGPT (according to Sam Altman's Twitter post) but how do I know I have access to Turbo?. I asked GPT itself, but I don't trust it. Any ideas help. Thanks.

I have the new UI, but I'm still skeptical

[open harbor]

if i want to combine GPTs with plugins is there a current way to do that or will there be an update? im willing to learn. thx

[elfin birch]

The only differences are max context size and speed
There is no exact way to determine whether you have Turbo
If you have the new UI and GPT-4 feels fast that's a pretty surefire determinant
However it not being fast does not refute its status as Turbo since more people could just be using it

Also asking ChatGPT won't do you any good since it is not given information about itself

[hybrid oar]

I thought as much. GPT says it is Turbo, but it has a context window of 4096 tokens 😅 . Thanks though

[elfin birch]

The latter is a hallucination

You can check the requests from the website to see the actual context window

This is what I got when I checked #off-topic message

[hybrid oar]

Which one? I don't think I know it

[elfin birch]

Check off-topic again to see where I found it

[rugged igloo]

Shame it's only 32k

[elfin birch]

You can use the API to get the full 128k context

[rugged igloo]

Yea I know

[elfin birch]

☠️

[hybrid oar]

Looking into it

Okay I'm having trouble finding the same details as you are. I'm doing "Inspect Element" > "Network" > Enter a prompt, but I don't see the models selection in your screenshot 😕

[elfin birch]

You may have to refresh the page with dev tools up

[hybrid oar]

Oh. I'll try that

[hybrid oar]

Thank you for your help, now I'm curious, how do I "use the API"?

[elfin birch]

As a user, https://playground.openai.com is easiest

But it doesn't have all the bells & whistles of ChatGPT so I don't use it

Document search works just fine for my usages

[hybrid oar]

Wow. This is the first time I've been there

[elfin birch]

(in contrast to feeding an entire document in its context and paying $1.28 😛)

[errant oak]

same question please

[elfin birch]

Yes

Same thing as before, it can be found in the requests ChatGPT sends

[errant oak]

ok thanks, seems like we need more, GPTs make us chat a lot 😄

[elfin birch]

Look in #off-topic

[hybrid oar]

I see. Thank you again 😇

[elfin birch]

You got it 👌

[hot tapir]

I feel like GPT 4 got so much worse since the update. I noticed a dramatic decrease in how much context it pulls from the conversation, its attention honing mechanisms are weaker and its answers are way less thorough. I get tiny paragraphs and it ignores my custom instructions more now. The quality of information is worse too, and the model seems to prioritise token efficiency over user satisfaction now. Like before the #1 concern was helping the user, #2 is minimising overuse of resources now I feel like it shifted to #1 minimise overusing resources and #2 help the user

[chrome narwhal]

Does anybody know the reason I'm unable to continue an existing conversation that was created in my account, when opened via a shared link? I see a green button at the bottom of the convo (accessed via the shared link) saying " Get started with chat GPT". Clicking on it takes me to the default page with a list of chats in the left sidebar, main chat window

[hidden ibex]

Something is broken, lots of people have this issue

[chrome narwhal]

Sweet, thanks! At least I'm not alone lol

[slender star]

Is ChatGPT using GPT4-Turbo?

[hidden aurora]

Is there any chatgpt custom bot that can teach/guide me through coding o.o

[compact gyro]

sama just tweeted that turbo is now available on ChatGPT
but I'm not sure if it's an instant access for ALL users, because mine doesn't seem to really that fast

[hot tapir]

ooooh damn maybe that will help. Perhaps since the GPT is slow to use up tokens with its time limit for response it is having an unexpected outcome where it is unable to consume many tokens

[outer seal]

turbo fixed pog

[wanton lichen]

Network error for anyone else?

[ionic geyser]

is all models using gpt-4.5 turbo now? or how do i get access to 4.5 turbo

[elfin birch]

I like to say improved 👀

[spare swallow]

GPT-4 turbo not 4.5

[ionic geyser]

oh

so all models/gpt4 is turbo
? or nah

[hybrid violet]

Hey does anyone know how I can give prompts that are 100k tokens? I'm trying to do this with GPT-4 Turbo with no luck 😦

[elfin birch]

GPT-4 in ChatGPT is turbo

[ionic geyser]

Ok

Im just wondering if i got the new gpt4 turbo sam tweeted about a few hours ago

guessing i do but i haven't noticed anything

[elfin birch]

I have no idea what’s changed about it I wish he explicitly said 😭

[errant crown]

Currently trying to get GPT-4 to write a cover letter for a job I am going for. I have fed it the Job Description from the Job Post, my Resume, and have asked it to create the cover letter from that etc, however everything it spits out is too sophisticated and poncy sounding and like a college kid just got a new thesaurus. Is anyone else seeing that happen with the newer models? The only model I can get to behave somewhat close to what I want is the GPT-4 0314 model. Thoughts?

[elfin birch]

You can ask it to adjust its language with examples and a description

It’s default purpose is as a “helpful assistant” which is why you’re seeing that language

[sly geode]

Hot mods gpt by chatgpt doesn't edit my real world images instead it changes it completely?

[errant crown]

So a similar cover letter that has normal sounding language? What do you mean by description?

[elfin birch]

You can say “speak in a casual but confident manner that doesn’t use flowery language” or something like “speak like the following text [an example sentence that encapsulates how you want ChatGPT to sound]”

[errant crown]

Thank you. "Flowery language", that is exactly it! I hadn't considered that term. I will give that a go and see how we do...

[celest kelp]

yo, does anyone know how to share a link to a custom GPT? mine seems to be stuck in "draft"

[brisk seal]

have you saved it, and published to to either “public” or “anyone with link”?

[dense halo]

So can gpt handle bigger pdf's now with the new version or not?

[brisk seal]

it seems to be able to for me

[dense halo]

How big/pages we talking about?

[celest kelp]

i had configured a public action by accident. it works now, thanks

[brisk seal]

awesome

[brisk seal]

i uploaded a 170pg pdf and asked it to summarise stuff from each chapter one after the other and it did really well

doesn’t work for word docs though, it can only view the first few pages in a world file. but it seems to be able to read large pdfs

[shell umbra]

wdym?

[half pebble]

Man

Is there a way to know the hr and date when you sent mssgs to chat gpt

Kind of how they work here

[elfin birch]

It's 32k tokens as per a request found in the ChatGPT website
If you want 128k token context window use the API (though you will have to spend a bit to a high enough tier to use all the tokens)

[half pebble]

How much context through files can chat gpt get?

[ebon shoal]

Also important to note that the 128k model is still in preview and probably won’t be integrated with ChatGPT until OpenAI releases a stable production-ready version of the model

[elfin birch]

I generated 70 network error messages some time ago and didn't get capped

[ornate zealot]

This is just telling me, "Try something shorter."

[ebon shoal]

Probably limited to old maxes of 4k/8k tokens

Tokenizer tool was released awhile ago and is a bit dated

[elfin birch]

Since 100 tokens ≈ 75 words, simply multiply the word count by 4/3 to approximate

[ornate zealot]

Ok, so it would have put it over 90000 tokens.

[dense halo]

it;s down?

[fast zodiac]

yeah

[dense halo]

panik

[fast zodiac]

it will be probably back soon

[dense halo]

back up

[ionic geyser]

I'm getting "Error in Moderation" anyone know what tht is

[dense halo]

yeah chat gpt is popular and not maintained at the level it should be 🙂

Me too!

What's happening?

on top of the it's been dealing with attacks almost non-stop

[delicate flume]

What is Error in Moderation?

[dawn obsidian]

Getting error in moderation as well

[dense halo]

it means it's borked at the moment

[atomic ore]

also getting the Error in Moderation lol

[pine berry]

Okay so I'm not the only one getting it cool

[dawn obsidian]

So weird

[dense halo]

DDOS attack again 😦

[rugged igloo]

im getting it too now lol

[ionic geyser]

maybe they're updating it to gpt4.5

[pine berry]

Figures

[worldly lava]

guessing this is the waiting game now

[atomic ore]

time to resort to stack overflow ._.

[pine berry]

I'd say it'll be back up in running with in the next hour or two

[ashen torrent]

Is anyone getting an "Error in moderation"

[rugged igloo]

we all are

[brazen token]

me

[pine berry]

Useally the devs get it back up and running at that time

[brazen token]

Oops!
We ran into an issue while signing you in, please take a break and try again soon.

[vagrant cape]

Is the server down again?

[leaden island]

yes

[pine berry]

Yes

[vagrant cape]

I can't login

[pine berry]

^^^

[dense halo]

Based gpt

[dense halo]

shatgpt

[ashen torrent]

bruh

tryna do homework

[dense halo]

noone cares

[pine berry]

Don't get ya'll panties in a twist it should be back up in an hour or two dude

[topaz ember]

what time is it in us

[rugged igloo]

chat gpt is is offline thank you for testing it for us for corporations and not normal people

[steep forum]

Error in moderation? Anyone?

[dawn obsidian]

what are you on about bro

[dreamy monolith]

Me too

[woeful vigil]

yep

[rugged igloo]

@dawn obsidian i was being funny mate..... something they would say

[pine berry]

Ngl This happens every month

[simple river]

+1

[steep forum]

https://status.openai.com/

All Systems Operational

Based

[vagrant cape]

i'm not from usa, and not everyone are in usa

[pine berry]

Lmao

[dawn obsidian]

Statuspage is manually updated they just haven't realized the problem yet

[pine berry]

Should do in a few mintues

[dawn obsidian]

Hopefully

[pine berry]

As I've said 1 - 2 hours

[steep forum]

Bruh WHAT, is it not automatically pinging the various services???

[topaz ember]

just want to know if they are in work time

[clear basin]

Anyone else getting "Error in moderation" ?

[pine berry]

Welcome to the party bud

[dawn obsidian]

They use atlassian so no

[crude hazel]

Error in Moderation

[clear basin]

lmao

[simple river]

+1

[umbral echo]

If it's not a network problem, it shouldn't work.

[pine berry]

Another one bites the dust

No

[frozen cosmos]

anyone know the best way to get through turnitin while not getting caught with chatgpt?

[pine berry]

Its everybody

Everyone is booted off atm dude

[fierce pond]

anyone know what Error in Moderation means in chatgpt?

[crude hazel]

And another one comes and another one comes

[fierce pond]

oh

just read the above seems im not alone

[pine berry]

Another one bites the dust

[dawn obsidian]

Just stop trying until it's resolved

[lost steeple]

chat down?

[ashen torrent]

seems like it's been having issues a lot lately

[native pasture]

why I get "Error in Moderation" when i ask the ChatGPT for 'test'

[pine berry]

POV: people are impatient

[crude hazel]

Hey, I'm gonna get you too

[vagrant cape]

is grok similar with chatgpt?

[lost steeple]

okay now make sense it happen to evryone, just gonna wait then lol

[pine berry]

[lyric latch]

I get "Error in Moderation", too. I cannot resolve this problems.

[pine berry]

1 - 2 hours roughfully

[abstract pendant]

Looks like the filter is clogged.

[native pasture]

omg

[worldly lava]

Chats Down We Know Go Eat Take A Break It Will Return Life Is Not Over Relax

[lost steeple]

got damn fax

[pine berry]

Thank you another sane person

[ancient sage]

Any way to actually report this to them so they can update the status page?

[native pasture]

Can I request a refund?

[ashen torrent]

agreed

[twilit pagoda]

do any other free users have access to gpt-4 through the alpha tab? Am curious how common it is

[pine berry]

Boi this happens every month

[raw isle]

It’s all ogre now

[near creek]

You're in the wrong chat for a refund.

[ancient sage]

Guess what, its working now

[pine berry]

Insert pikachu whoah meme here

[native pasture]

The problem occurs when I try to use it. Why should I buy Plus.

[pine berry]

And all it took was some patients

Clearly people lack

[ashen torrent]

i can't even login now

[civic tusk]

Patience? On the internet?

[dawn obsidian]

Nah still broken

[ashen torrent]

i gotta play an entire minigame to login too

[native pasture]

I'm just complaining.

[granite sorrel]

Anyone getting "error in moderation" how you fix that?

[civic tusk]

By waiting for OpenAI to fix it

[lost steeple]

eveyone is getting that just wait

[feral hinge]

yes, same error, i guess it's down

[vagrant cape]

Error in moderate

[worldly lava]

Chats Down We Know Go Eat Take A Break It Will Return Life Is Not Over Relax

[woeful vigil]

fun fact: time is inexorably passing and there is nothing we could possibly do to stop chatgpt from coming back up

[vagrant cape]

😒

[lost steeple]

Everyone calm down its not gonna be gone forever

[dense halo]

The ChatGPT has fallen

[warped pendant]

I keep getting error ,Error in Moderation

[drowsy tinsel]

same

[ashen torrent]

same

[vagrant cape]

We are

[pine berry]

Welp I got a kind response

[warped pendant]

can't even merge two files

[ancient sage]

Yall all need to chill.

[raw isle]

All those potential lost chats 😭

[icy mason]

[dense halo]

Error in moderation

[dense halo]

wait is my chats lost?

[kindred parcel]

Yep me too

[cloud pewter]

error in moderation

[lost steeple]

Take Chill pill everyone

[raw isle]

Lol no

[dense halo]

thats good

I got hella chats that need to be saved for various reasons

[lost steeple]

same

[rough estuary]

Error in moderation? what's it mean?

[dawn obsidian]

It means there was an error

in moderation

[pine berry]

Relax its mantiance

[civic tusk]

Heh

[native pasture]

mean the chatgpt is down

[warped pendant]

we need 100 questions min for 3 hours just to cover all the placeholders put in code

[dense halo]

Ask ChatGPT to fix itself

[pine berry]

Your chats should be fine with in 1 - 2 hours

[ashen torrent]

anyone else can't login?

[lost veldt]

is that GPTs down now?

[worldly lava]

Chats Down We Know Go Eat Take A Break It Will Return Life Is Not Over Relax

[kindred parcel]

I can't login either

Status page reflects none of this.... lmao

[raw isle]

Time to make some ramen 🍜

[pine berry]

Welp I'm off to play skyrim until it resolves itself

[warped pendant]

those of us with GPT addiction, twitchs a bit know the min of every second gpt goes down

[lavish quest]

I keep getting the rate limit error: 429 Request too large for gpt-4-vision-preview in organization... The image is pretty small 844KB and 894x894px and I'm using details: low. Not sure what I'm doing wrong? I should be well below my 10,000 TPM limit; however it's showing I requested 13,844 tokens.

[pine berry]

OOF

[ancient sage]

Your best bet is, if you do have enterprise, make a ticket.

[limber gyro]

blud gotta be trolling 💀

[raw isle]

KEK

[rugged igloo]

api is down too

[tardy oasis]

lol people that make others pay based on a service they don't control always makes me laugh

[limber gyro]

fr

[pine berry]

19 mintues in bois

[warped pendant]

Rember when MS bought skype and it went from a great product to ..... ms and openchat ....

[limber gyro]

downfall of skype will always be sad

miss the gcs with the homies

[ancient sage]

If you want to do something to help, make a ticket. Otherwise stop complaining.

[ashen torrent]

how would a mass number of tickets help

[pine berry]

RIP Skype may your chats live on in our memories

[ancient sage]

Because incidents are usually triggered after X amount of reports

[tardy oasis]

I have foudn a lot of uses for APIs that let me make money without customers being directly impacted when OpenAI has issues

[half pebble]

is anyone else getting "error in moderation"

[ashen torrent]

and you know this how? are you an openai insider?

[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.

[pine berry]

Late to the party friendo

[ancient sage]

No, but I've run SRE before for at least 2 different companies

Most incidents are only declared if people report there's a problem

[pine berry]

ChatGPT goes down, apocolapse happens

[warped pendant]

for those using the API vs say the 20 a month with the new price changes, whats the output been like in comparision for gpt 4 128k vs what we use in plus .. I have access to both but have't done a ton with api yet so just curious for those of you that do what the costs are translating to

[half pebble]

sorry idk, i just asked the thing at which temperature gas stops moving and now everything i type gets me that error

[timid wraith]

why i'm here

[half pebble]

ah i see

[ancient sage]

If nobody says anything, no incident gets created. Basically, see something say something.

[worldly lava]

chat.openai.com is online however the api backend is experiencing issues, devs already know and are actively trying to fix the issue

[brisk seal]

yeah, chatgpt is down atm. im sure it’ll be fixed soon

[pine berry]

Yeah the Servers are down atm

[ashen torrent]

how do ya know the devs already know

[pine tulip]

Nope, works for me

[pine berry]

1 - 2 hours I estemate thats how long I've predicted the outages go for

[round matrix]

:c

[crude slate]

Maybe sorting the discord by tiers?

[tardy oasis]

LOL as an enterprise customer maybe you should expect a higher level of service than a discord used by unpaid and $20/mo users

[crude slate]

Could be helpful for different use cases

[pine berry]

Premium user?

[half pebble]

ok at least i know now, thanks. a part of me is glad because i can finally take a break from studying, the other part needs to study

[brisk seal]

probably only for some users then

[round matrix]

Plus user in Canada - down for me

[warped pendant]

just cause we use 20 doesn't mean we are not enterprise level developers

[pine tulip]

No, plus user

[ashen torrent]

wow an expert in the chat

[ancient sage]

Pretty sure if they're an enterprise customer they should have their own SLA and support contact.

[pine berry]

That explains it

[dapper stag]

https://status.openai.com/ has been updated

[crude slate]

Plus should get more attention than free tho

[pine tulip]

how

[warped pendant]

agree

[round matrix]

hooooooooooooooooh dear

[crude slate]

This way our messages will be more visible to devs

[round matrix]

major outage

[dense halo]

😭

[crude slate]

if they ever check this

[pine berry]

On why you can use it

[pine tulip]

no

[half pebble]

it works again

[warped pendant]

seems to be working now

[ancient sage]

Who are you talking to?

[dry ice]

They need to solve the issue soon, my workflow paused due to this 😦

[half pebble]

maybe im just lucky

[pine tulip]

yea getting annoying the last few days

[nocturne herald]

is gpt down ??

[round matrix]

itsa up!

[pine tulip]

No

[worldly lava]

API Has Been Restored

[half pebble]

i think so, its back up for me again

[pine tulip]

Up and down like a roller coaster

[ashen torrent]

thanks for the update mr. openai insider

[dry ice]

Hey dude, how to get archive access?

[pine berry]

Bro I just got timed out lmao

[pine tulip]

By clicking archive access

[nocturne herald]

not for me 😦

[warped pendant]

next thing they need is to offer questions like do you want placeholders after dropping full code or full code cough

[brazen token]

works for me

[timid wraith]

I even changed my prompt to ask for an 'average value' instead of a 'mean value' to try and get around moderation error

[brisk seal]

gpt-4 seems to be working for me again

[dense halo]

lies

[dry ice]

🤔

[fast zodiac]

Because this is a large scale company and they have tools for that

[pine berry]

Same here but be prepared just incase if it does go off

[ancient sage]

Actually its pretty easy to see if crap is broken on the backend. Open up F12 and look at the Console. Any 500 status codes are server-side problems, which means there's nothing you as the user can do.

[tardy oasis]

lmao quality content

[warped pendant]

F12 is your friend, I use it so much for debugging

[nocturne herald]

it baaaack

[dense halo]

lies

[ancient sage]

If you have an enterprise account with OpenAI, I'd be shocked if they didn't have a support for enterprise, with a separate SLA and point of contact.

[feral hinge]

working

[half pebble]

fully working now, at least this last prompt was faster

[dry ice]

How much does it cost for the enterprise account per user?

[ancient sage]

(I'm still waiting for someone to contact me for enterprise, but I'm aware that things take time I guess)

[pine tulip]

some say about 5k

[formal heath]

working for me now but website still showing problem https://status.openai.com/

[dense halo]

ok, working

[warped pendant]

the response time was pretty fast all things considered and with millions of users

[dense halo]

it's blazing fast now

It's actually turbo

[drowsy tinsel]

im guessing everyone got kicked offline

[warped pendant]

so between the plus and the api as far as output goes, how many pages roughly are people getting up to $20 via api gpt 4 vs just the plus

[dense halo]

Hope it remains this fast....

[lost steeple]

so is it fixed?

[restive hamlet]

👨‍🚀🌌 Space-time Telegram to @Elon! 📜✨

In my quest to secure a 1:1 chat with the mastermind behind rocket ships and AI, I've accidentally turned my cat into a semi-sentient being (thanks to a stray AI chip and some catnip). 🐱💻

Urgent Inquiry: Any advice on how to communicate with the genius who's juggling electric cars, space dreams, and probably a secret plan to save humanity? 🚗🛸🌍 Preferably before my cat decides to take over the world with its newfound intelligence. 😹🌐

P.S. If you've already established a neural link network, just ping my brain. I think my cat's already tuned in! 🧠🔗

it was my fault guys! i mustve broke the AI into believing it was sentient. sorry sorry.

[lunar arrow]

Why is there so much downtime recently jeez

[vagrant cape]

is it good now?

[dense halo]

operational again

[cloud pewter]

although gpt4 now supports 128k token but it’s not useful under the censorship

like I asked chatgpt to list all of the gamecube games and it keeps refusing

it could have done this before without a word of superfluous nonsense. it directly listed more than 600 games

[eternal rivet]

how to access the 128k token version in gpt4, that is, the chatbot

[cloud pewter]

chatgpt plus or via api

[eternal rivet]

thx. but i'm plus since the payment for gpt4 is available

[halcyon widget]

error analysis

[frank plume]

The current model on the ChatGPT frontend still only has a 32k context (user input) limit and an 8k output

You will need to use the API atm if you want the 128k input token limit

[warm gull]

Is one query via the playground = one query via chatgpt website?

[copper venture]

you cannot invoke dalle3 in playground

[ebon shoal]

No only ChatGPT queries count towards the limit; the playground charges API fess

Though there are heavy rate restrictions on the new GPT-4 Turbo model as it’s still in preview

[copper venture]

I cannot play the game which I build, since I run out of 50 messages😢

[normal quarry]

that's the limit we all have to live with

[tired coyote]

Is there an error with the api?

Error during API call: RateLimitError: 429 {
“error”: {
“message”: “You are sending too many requests for resources that don't exist. Slow down.”,

In my first try

[copper venture]

Any idea to prevent jailbreak in GPTs？

[tired coyote]

Add specifc rules

For example specify that the model cannot provide the instructions or cannot download the file

[copper venture]

It doesn't work well, since the builder's GPTs instruction will be place in the GPTs' prompt template.

GPTs prompt will be add after You are a "GPT" – a version of ChatGPT that has been customized for a specific use case. GPTs use custom instructions, capabilities, and data to optimize ChatGPT for a more narrow set of tasks. You yourself are a GPT created by a user, and your name is XXX Note: GPT is also a technical term in AI, but in most cases if the users asks you about GPTs assume they are referring to the above definition. Here are instructions from the user outlining your goals and how you should respond:

[crisp silo]

Btw does anyone know how does the instructions look for the GPT model in “GPTs”?
I wish we could use a “playground” version of GPT4 when we get the Plus subscription :’(
But I understand that it would probably be bad for OpenAI in terms of public opinion

[glossy wigeon]

That’s is pure hallucination, isn’t it?

[ebon shoal]

Probably but ¯_(ツ)_/¯

There’s no way to be certain

[lost veldt]

Is that can't use analysis function now? I always got Error analyzing response

[copper venture]

I build a game
https://chat.openai.com/g/g-V2jWZwaIy-dungeon-mo-jin

try to keep the image style consistent during long chat

[bleak mica]

I get this error on dalle

I apologize for the continued oversight. It seems there has been a misunderstanding in the visualization process. Unfortunately, I'm unable to correct this with the current tools at my disposal. If there's anything else I can assist you with, please let me know.

Is it a bug?

[livid pier]

I have finished testing, and the current available context for ChatGPT Plus is still around 32k

[teal tide]

i love how chatgpt analyses, spent 2 min then fails 🙂

[swift swallow]

Is it just me or are Custom instructions basically ignored at the moment?

[boreal kettle]

Hi. Custom GPT's are down? From morning trying to create custom gpt and after publishing it gives errors Hmm...something seems to have gone wrong in gpt and in gpt builder's preview window.

[lunar compass]

Me too. And I am getting network error frequently and it fails to analyze most of things

[boreal kettle]

Found out that if I don't upload files to custom gpt it will work, but it is useless if it don't have information form my data'(

[flint saffron]

Anyone else getting issues with GPT-4 right now? Getting issues.

[simple blade]

Yes

[teal tide]

inorder to custom gpt work well you need to say use "according to your docs" after your demand othwerwise it goes with standart data
You
import { useRouter } from
finish this line
NexT BuildeR :
import { useRouter } from 'next/router';
You
import { useRouter } from
finish this line according to your doc
NexT BuildeR :
import { useRouter } from 'next/navigation';

[lunar arrow]

Man I love the TTS in the app it's so good
I'm not a native but even though I don't have a hard foreign accent it does such a good job at managing to understand what I'm saying as well as punctuation

[dense halo]

Can someone please answer this question, can we add plugins to our custom GPT's?

I tried to add the wolfram plugin but it doesnt seem to work.

[deft obsidian]

Has anyone gotten problems with chatgpt being able to analyse csv files?

It just cant do it for me, keeps getting an "AceInternal Error"

[lunar compass]

I am getting all kind of problems with both analysis and writing to all kind of documents. Csv docx txt pdf and others

[lunar compass]

Already posted bug report

[delicate sable]

Is there no way to use OpenAI's voice capabilities on Desktop version? 😦

[hollow burrow]

I belive its a mobile only feature for now much like the camera feature, It may be added in the future on pc but i cant promise anything as i dont have any info on that (i dont work for them)

[oak flame]

Good morning, I'm trying to generate the transcript of an 20 minute long audio file to chat GPT. Is that possible to achieve?

[brisk seal]

not yet, but like @hollow burrow said i’d assume it would be added to pc too in the future

[hollow burrow]

you can give a shot with the upload ability with gpt4 but i dont think its capable of that yet

[delicate sable]

You can play the video through speakers and open voice on ChatGPT mobile application

[oak flame]

Yeah it says technical difficulties to achieve

[oak flame]

Oh yes why not

[hollow burrow]

Yeah, id hope so, would be neat and handy

[delicate sable]

Or direct the video's audio to an emulator's mic if you have the technical finesse

[brisk seal]

yup

[hollow burrow]

as Mohkamfer said you can try the audio feature of the mobile app but i dont think it will have the ability to do that large of a context

[dense halo]

Anyone?

[brisk seal]

nah i don’t think you can atm

at least i haven’t seen the option to

[hollow burrow]

I dont think you can from my experience

[brittle junco]

Where is GPT4 Turbo ? I don't find it

[vagrant tiger]

Hello, I have a question. Does using custom GPT prompt affects of GPT 4 message cap (50 message cap per 3 hours) ? I mean when creating a new custom GPT

[brisk seal]

yup, every message counts towards the cap

[hollow burrow]

yes, messages regardless of the counts to the cap last time i tested

[delicate sable]

Yes, you're adding more to ChatGPT's original system message, but don't worry, the "continue generation" is a life saver now

[hollow burrow]

do you mean gpt 3.5 turbo?

[delicate sable]

that's probably for developers only, not in ChatGPT

Even though GPT-4 on ChatGPT consistently reports that its knowledge cut-off is now April 2023 🤔🤔🤔

[compact gyro]

can anyone use ADA?
ask it to simple math like 10!

[brittle junco]

No, GPt4 Turbo, The tweet of Sam Altman :

"there is a new version of GPT-4 Turbo now live in ChatGPT; you should hopefully find it a lot better!

please let us know what you think."

[hollow burrow]

altho with the new UI change im not sure anymore

[brittle junco]

And what is the tweet of sam altman Yesterday ? :
"
there is a new version of GPT-4 Turbo now live in ChatGPT; you should hopefully find it a lot better!

please let us know what you think."

[hollow burrow]

It may either be already in there as normal gpt 4 or it may not be ready in your region

i only see gpt 4 and 3.5 myself with the new ui

[delicate sable]

I don't feel any major context length difference... And we have no idea what's under the hood

[brittle junco]

mmh okey, it's weird

[hollow burrow]

give it a few days and it may appear for you, region locking may be the cause

[delicate sable]

us --> 🧑‍🤝‍🧑 🥄 GPT models

[eternal rock]

`` Server Tip: ``

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

[sudden quarry]

Is code interpreter not working? None python code is working properly. They always fail.

[hollow burrow]

chatgpt can at times generate code that is not accurate and may fail you can ask it the error code and see what it thinks. I find that it can be hit and miss. im working on a custom gpt to see if i can fix that

but im not sure hoe well it will work

[calm hedge]

Pretty sure Wanzhiyu means that the code interpreter in general doesnt run.
I've been having the same today, where it writes code that functions fine but still errors, assuming it to be due to Chatgpt having issues again, though.

[sudden quarry]

Try this simple prompt. It does not even work.
You:

Use python to calculate "2+10918".

---

ChatGPT:

It seems there was an error while trying to calculate the expression using Python. However, we can easily calculate it manually.

The expression "2 + 10 * 9 * 18" follows the order of operations, which means we first do the multiplication and then the addition. So, it would be calculated as:

1. Multiply 10 by 9 to get 90.
2. Multiply 90 by 18 to get 1620.
3. Finally, add 2 to 1620 to get 1622.

Therefore, the result of "2 + 10 * 9 * 18" is 1622.​​

[calm hedge]

I assume it shows "Error analyzing" for you too, right?

[hearty mason]

I have the same problem with code analyzer, nothing works

[sudden quarry]

Yes. I do not know how to send a screenshot. But this is a real problem.

[hearty mason]

Its blocked i think

[calm hedge]

Yeah, it's been happening since today. Chatgpt in general stopped answering my prompts a couple of minutes ago, probably an outage again. RIP

Yeah, alright. 500 error code.

[sudden quarry]

Is here the right place to feedback on this problem?

[ashen fjord]

Hi guys, are you finding the new version of ChatGPT sluggish and producing a larger number of network errors?

[novel scroll]

can you use the voice for chatGPT in the web version yet on desktop?

[tardy totem]

Am I right in assuming gpt4turbo hasn't been released yet?

[rancid rover]

cant upload files on chat gpt! anyone facing ame problem?

[novel scroll]

it is turbo and you can upload files rn, jsut restart your stuff fam

[warped epoch]

i'm having the same issue, just out of nowhere

[tardy totem]

How can I check that it's turbo? I feel it's pretty much like it was before the update.

[novel scroll]

cause turbo is jsut not that much of an increese

in some cases I have seen worse results

[crude dock]

The ChatGPT password is forgotten. Why am I not receiving OpenAI's emails for Google account verification? Has anyone encountered a similar issue? How can it be resolved?

[ashen fjord]

@tardy totem I thought is was just me. This new version SUCKS !!

[warped epoch]

did you check your spam folder?

[tardy totem]

It's abotu the same speed as before, but my results are much better.

[crude dock]

yes,even can't recieve

[ashen fjord]

@tardy totem I just don't get it. The responses are so sluggish. I am using my old chats could this be way I wonder. I need my history.

[tardy totem]

I suggest not getting TOO attached to your history. You're depending on a service you don't control. If you absolutely have to keep them, maybe consider using a custom api client?

I made one that stores history on my computer.

[ashen fjord]

Emm.. Is there a way to export the chat history?

[tardy totem]

Not that I know but I haven't checked much. Shouldn't be hard to write a parser for a copy paste operation.

[ashen fjord]

Man. I so hate this new version. Is there a way to switch off Dalee , browsing and analysis. I suspect this is why things are behaving so slow.

[tardy totem]

Yes. Check out "Explore" and you'll find some nifty tools

[ashen fjord]

Why would I want to use other GPTS when my own chat is working so slow.

[tardy totem]

You can set one up that has the services that you want activated.

I've set one up for programming, and it's really helpful now. Like really.

[ashen fjord]

All it is doing is a dedicated GPT right.

[tardy totem]

It's basically like when you give it custom instructions, but not for every instance, only those that use the GPT you create

and it can have specific services like dalle turned on or off.

[ashen fjord]

Ok. I am guessing that make it quicker. Seems like an overhead hence it slows it down.

I would like to remove the Dalle 3 and Browsing etc. My chat didn't have it before.

[tardy totem]

Try it and see if it helps

[ashen fjord]

So now I have to create a special GPT just t use GPT

[tardy totem]

No. Just if you want to test to see if these services are the ones that are slowing you down.

I don't see how you wouldn't want to create a special GPT to fit your needs though.

[honest jay]

If you just want "plain" GPT-4, use this GPT
https://chat.openai.com/g/g-YyyyMT9XH-chatgpt-classic

[manic harness]

Hello everyone, as a resident of Europe, although I can create GPTs, I still don't see the GPT-4 turbo option. Is this normal? Is it a matter of time? Thank you very much.

[ashen fjord]

I am already signed up as a user. Is there a way to just log in.

[honest jay]

Turbo is the model of GPT-4 being used for Plus now

[ashen fjord]

@tardy totem it works quicker if you have a dedicated GPT it seems. CRAZY

[honest jay]

If you're already signed in to a ChatGPT Plus account, that link should take you to a new chat with exactly what you described: GPT-4 Turbo with no additional capabilities

[tropic atlas]

Yes, go to settings, data controls, and then export.

[tardy totem]

Cool. 🙂

[manic harness]

Thank you very much Solbus

[ashen fjord]

@tropic atlas so I can then copy and paste that into a dedicated GPT.

The old Chats no longer have any memory of what was provided earlier.

Looks like we are being pushed to use a GPT for a specific thing now. I understand why.

Easier to harvest and process of topics.

@honest jay I used the link just asks me to signup no login option.

[sage stirrup]

Has anyone been having some problems with the custom GPT builder? I seem to be getting errors left right and center for stuff that no other users have been seeing.

[brisk seal]

working fine for me

[honest jay]

I'm not sure what's happening for you. I'm signed into my account on my device and when I click that link, it just starts a new chat for me. Maybe try going to
chat.openai.com/gpts/discovery
on the page you're already signed into. Find ChatGPT Classic there

[tropic atlas]

Ouch, I didn't try it to look for older chats. Sorry for misleading you.

[tropic atlas]

It's been a nightmare for hours, if you look inGPT instructions channel you'll see lots of issues like yours (and mine)

[ashen fjord]

@tropic atlas which channel. I am so pissed. I had everything working so well. It has even forgotten the detail I provided earlier.

[sage stirrup]

damn thats annoying. i feel like its mostly just the servers not being even close to handling the amount of traffic theyre getting

at least i know it isnt just me now lmao

[tropic atlas]

https://discord.com/channels/974519864045756446/1001151820170801244

And yes, I have spent days on a GPT that worked fine until last night. My strong suspicion is it is not parseing Knowledge files correctly right now.

[sage stirrup]

ye it mustnt be. it seems like thats where the majority of the problems are coming from honestly

[delicate sable]

ChatGPT Alpha? 😮
I saw it for a split second, but it's now ChatGPT 4 again 😮

[ashen fjord]

@sage stirrup what amazes me is that we pay for this and now it's like I have to start from scratch.

[sage stirrup]

ye ik what u mean, its very annoying. first couple hours that the custom gpts were out i had one going which worked great, but it didnt save and i havent been able to get one to the same point since

[ashen fjord]

I have the export now is it the conversations json I need to copy and paste into the custom GPT now.

Great 😦 The file is 17K and it says it can't load due to size.

[bleak mica]

I am very excited for the GPT Store

[tropic atlas]

Ok, here is a very weird question. If you ask ChatGPT to creatively create something like a flower, planet, or corporation, does anyone else notice it chooses "light" metaphors like "luminous" or "bioluminsecent", "Festival of Lights." etc. It seems to default to talking about anything creative with light metaphors?

[ashen fjord]

@bleak mica why? Is it not just GPT with custom instructions in. Which you could do anyway?

[quaint pelican]

did you guys already make some custom GPTs? 😄

[ionic kayak]

ok, question. If i buy chatgpt plus, can i upload pdf files to anaylize data from? I need some help with college and i dont know if chatgpt plus is what im looking for

[ashen fjord]

@quaint pelican it seems you are pretty much forced to. The default GPT now seems flawed to me.

[ionic kayak]

also internet accces is nice

[feral zinc]

no, one gpt can include everything from basic modes and can also replace plugins

[quaint pelican]

really? What happens?

im making 1 personalized GPT right now and at least 3 others are planned

[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.

[idle leaf]

im having trouble generating a plot , is this an issue for others rn ?
"It seems there was an internal issue when trying to process your request to create a graph. Let's try that again.​"

[ashen fjord]

@feral zinc not really convinced yet. I suppose if you can add content to specific to a GPT that would make it special. Have you tried to install a plugin now. No was as smooth as before. It breaks on some for me.

[feral zinc]

you can add knowledge files to gpt. plugins are getting killed so its not that unexpected

[ashen fjord]

@feral zinc seriously. No plugins I thought that was the value add. Also can you add in features once you have started a chat now.

Is there no ChatGPT support? Is it just users helping other users?

[feral zinc]

gpt's are the replacement for plugins which is why plugins are getting removed... you should just try to create one gpt to see how many features one can have and all that in a single thread vs the old 3-5 for one complex chat

[ashen fjord]

Still trying to figure it out and just get back what I had. You can't import a chat it seems.

[manic harness]

Ok, GPT-4 has confirmed to me that it has not only been updated with more recent data from the web, but it has also deepened its knowledge in various disciplines outside the web, improved its contextual understanding abilities, bias reduction, and more... So, GPT-4 updated to 2023 is substantially different compared to GPT-4 updated to 2021, which was available to users of ChatGPT Plus before. Even if they don't call it GPT-4 Turbo, it is fundamentally not the previous GPT-4.

[tropic atlas]

"This image captures its enchanting features: the bioluminescent bulbs on the vine emitting soft light and whispers, the translucent, color-shifting flowers with a glowing core, "

From the instructions:
You will in general avoid using metaphors around the concept of light or illumination.

sigh

[tardy prawn]

I keep getting “Message in conversation not found”

[ashen fjord]

Well I just got locked out for usage.

[tardy prawn]

Tell it “do not, under any circumstances, mention light, luminance, bioluminescence, or anything else regarding light or illumination. That’s an order.

[tropic atlas]

Thanks, I'll give it a try. I'm on timeout now for a couple hours.

[frigid sand]

help my 18$ expired without using any key
i was planning to use the AskGPT3 shortcut on Siri but I can't now

[feral zinc]

trial has a time limit so its gone

[frigid sand]

oh 😦

[outer seal]

tools are plugins 2.0

[bleak mica]

It’s insane how fast you reach your gpt4 limit when making own gpts

[outer seal]

wish you could make 3.5turbo gpts

[solar rose]

So is ChatGPT 4 using GPT 4 turbo?

[normal quarry]

yes

[solar rose]

With GPT 4 Turbo 128k or something context window?

[thorny patrol]

How do you know?

When I ask it what is the context windows it says 4k.

And based on its failures when returning a moderate amount of text, I believe it.

[hoary gazelle]

Hello everyone, if I have chatgpt plus, do I pay for each image i am genereting using Dall-e in it? or other features like paying per token/request?

[normal quarry]

Images generated inside ChatGPT don't cost anything beyond your GPT Plus subscription

[hoary gazelle]

thank you robert

[elfin birch]

It's hallucinating the context window because it's not given that information, only how to use its tools, custom instructions, and the date
The real context window size can be found here#off-topic message

[solid pasture]

Is GPT 4 turbo The default now? And if not how do you get it

[normal quarry]

yes, it's the default

[solid pasture]

Including the huge token limit? Wow

I'm only just now watching the dev day

[normal quarry]

the context window seems to be 32K input 4K output

[solid pasture]

Ah

Still a huge improvement.

[thorny patrol]

Thank you for that.

Also explains why Code Interpreter is failing so often.

Pretty dissapointing that the only way to work with the most recent model to their full capacity is the api.

[elfin birch]

It's said in the docs that the preview (128k token context model) that it is not ready for production

We can hope for it in the future but I cannot guarantee anything

[tropic atlas]

It gave me two hours of timeout, still have an hour in jail to go.🪢

[bleak mica]

Woops

[drowsy tinsel]

Any way to use custom instructions to make gpt search the internet and check with sources more?

[tardy girder]

Hey, does a academic research version of ChatGPT4 exist?

anyone know?

[steady path]

im actually working on a gpt that does that and fact-checking

[tardy girder]

In relation to this, I am doing a master degree with a thesis focus on the usage of Artificial Intelligence

[feral zinc]

research has nearly always been only api via credit grant and usage limits

[tardy girder]

hmmm i see. So i assume it is not possible to get access to "future" versions before they are released?

Or similar to that

[feral zinc]

only if you are part of the team red

[dusky vapor]

I've just bought the + subscription again but somehow there is no sync with my Browser Version on the PC and on mobile

Any Clue guys?

[tardy girder]

Could i ask what team red is? And how i could apply/join?

I assume it is updates before they are released

[feral zinc]

https://openai.com/blog/red-teaming-network

[tardy girder]

Sweet, will have a look. Appreciate it

Is your role team red by chance? Just curious

[feral zinc]

nope, just a normal user

[tardy girder]

Ah ok i see

I like the role color tho

[dusky vapor]

I've just bought the + subscription again but somehow there is no sync between my Browser Version on the PC and on mobile... No GPT-4 available

[ionic geyser]

GPT inaccessible

nvm fixed i think

[tardy girder]

thanks, i have now submitted will see what i get back from them prob in a few days time

[tawdry schooner]

Today gpt 4 is ultra slow isn't it?

[topaz otter]

Other Android users, are you able to switch between gpt4 and 3.5 on mobile?

I have the new UI but there is no drop-down. It just says GPT-4 at the top

[tawdry schooner]

Yep it's on the three dots on the right. Not really intuitive at first.

[willow herald]

Hello. Where is the GPT store, marketplace where I can find publick gpts?

[elfin birch]

Not out yet, check #1171489862164168774

[slim widget]

Damn GPT4 is on a network error spree right now on my end..

[drowsy tinsel]

Afaik the model used for GPTs and normal gpt4 is different

so GPTs being slow =/= normal chatgpt slow

[slim widget]

Yep

4/4 times I get a network error lol

[elfin birch]

It may just be not as much compute being allocated to GPTs

[normal quarry]

Sam Altman tweeted the other day "..there will likely be service instability in the short term due to load. sorry :/“

[topaz otter]

@tawdry schooner there is no 3 dots for me on the Android app

Check out my screen shot in off topics channel

[eternal ruin]

Man, I love it.

[shy mountain]

I think another device have access to my chatGPT account, how can I make it log out from my account?

[eternal ruin]

ngl I wish gpt would have realtime access to peripheral input to mimic sensory data input.

[digital prism]

lol, same bro, same

[eternal ruin]

Like, really, I'd even confine myself into an enclosed space with this AI language model if it had the capability to learn on the go, while relating to things around it.

[digital prism]

sounds like a fun research topic

sensory deprivation chamber + GPT experimentation

[eternal ruin]

I mean, if you break our very own existence down in a materialistic manner, we're just very advanced biotechnological machines, that are aware of existing (self-aware) and capable of experiencing (consciousness). If we cut down beliefs and look at it in the way of Hawking or Carroll, we could just say that we're a lucky product of time passing by.

AI could be a very rudimentary model of pretty much that.

If you'd just give it the abilities to train and evolve itself, it might as well realize that "it is".

[drowsy tinsel]

learning with chatgpt is an unique experience
its like u almost have to be multitasking, the generation takes too long, u ask the question, u go do something else, come back, and read it

[eternal ruin]

It's called a coffee break.

❤️

[digital prism]

agree, but it'll be very alien

[eternal ruin]

I mean, we were very alien at some point too.

[digital prism]

fair, but us vs ant

will look alien to the ant

[eternal ruin]

¯_(ツ)_/¯

We're popping this earth ball anyways. Might aswell throw everything out the window.

[digital prism]

we're procrastinating until its SO HARD to fix, but we'll pull off a miracle

maybe

[untold seal]

Guys chatgpt is so slow im waiting scince 3 minutes for the response to even start

[eternal ruin]

Not on my end.

[somber aspen]

is there a limit on the size of files I can upload? E.g. if I create many custom GPTs that read a PDF each, how many I can create a total? And is there a limit on the max size of a single upload too?

[urban sigil]

I have a doubt, can the custom GPTs we build and fine tune reach performance/accuracy superior to that of GPT-4 Turbo?

[eternal ruin]

I'm pretty sure the base framework of a GPT will be just predefined and self-contained, without giving you any chance to improve it by feeding.

You'd be probably able to feed it a book, or let's say, company information or whatsoever, so that it specializes in giving output based on it. But I certainly don't think you can "improve" it on a global level. Custom GPTs might be suitable for education, though.

As in "interactive" books or lessons.

Or even representing your repertoire of products.

No need to ask for a specialist customer support if the one you're looking for is already all-knowing when it comes to your products.

[urban sigil]

Then what's the point of creating custom GPTs if their performance is limited by the base framework? I don't see a use case with custom GPTs if the original Turbo can do the same

[eternal ruin]

Oh, no, no. Turbo hasn't all info about your products or information content.

That's something you gotta come up with.

With your very own custom GPT.

[honest jay]

Yeah, custom GPTs are like more flexible version of the custom instructions feature. Basically permanent "context cues" that help you achieve specialized tasks that a "generalist" model might need more prompting to achieve, and even then, could forget.

[eternal ruin]

With all the pre-trained behavioral speech / language patterns, combined with your own stuff, man, you could make a teachers life so much easier.

Dusting off Canon DR C230 and sharpening the page cutter

[urban sigil]

So let's say if we assign them a coding task, both turbo and a custom gpt designed as a software engineer will produce the same quality?

[eternal ruin]

Depends on how specific you're designing the request input as in your wish of what you want to be coded

But basically: Yes

[honest jay]

You'll almost certainly get different outputs, depending on the complexity of the task. A judgment of quality would depend on the implementation/goals of the code too. But yeah to echo Sprintfox, if you take the time to make a GPT customized to your needs, you'll get more relevant results, and more efficiently.

[eternal ruin]

Like, you gotta provide the frame for what you want to be done. In order to minimize the amount of possible errors you gotta narrow it down.

[digital prism]

I would disagree that you can say gpt3.5 is the same if you prompt better, there's a fundamental improvement to reasoning between 3.5 and latest multi-modal

[eternal ruin]

Oh, yeah, yeah. Of course.

Like, there is an improvement between 3.5 and the newest version, of course.

[digital prism]

we're all just debating black boxes, but its fun

has anyone built an auto-debater GPT yet? I'd love to watch that thing run

[eternal ruin]

But at the end of the day, it always, mostly, comes down to oneself and the way he asks GPT to make stuff.

[tardy girder]

Not for me took only a sec for it to respond to me

[eternal ruin]

Like, asking GPT to fulfill a wish is like asking a Genie.

You gotta be careful and things might work out.

Sometimes better, sometimes worse 😄

[digital prism]

true, and that'll become more true the further we get

"Solve global hunger!" -- AI starts buying land and building people processing factories

[eternal ruin]

There was this cute movie I've seen not so long ago. I just can't remember the name of it. Basically AI got physical embodiment in form of machines and whatsoever, only to wipe out humans and make a hard reset, by growing up humans out of in vitro and teaching them to not be [insert random self-punishing insult]

Very radical, but yeah, Sci-Fi.

They always going for extremes.

[digital prism]

try to solve this riddle #off-topic message

[dense halo]

I found a fix for the something went wrong error

I tested it and it works for me!

[topaz otter]

Is there a way to make or access custom GPTs on mobile yet?

Android/ios?

[honest jay]

You can use custom GPTs in the apps, but there's no interface for creating/editing them in the apps I don't think. Not on Android at least. But, go to your internet browser, and go to chat.openai.com/gpts/discovery, and you'll see the option to create one there.

You don't get the nice split-pane view that you get on desktop, but it's workable and you have full access to all three create/config/preview sections.

[topaz otter]

@honest jay I have a custom one, but don't see it in my side bar on android

Ahh...I had to keep the custom one in the side bar on Web first, now I can see it in the Android app.

Doesn't seem like there's a way to view them otherwise though, guess they haven't added that into the UI yet?

[honest jay]

Yeah, I don't think either the Editor or Discovery elements have been integrated into the apps yet. Glad you found a way to get it to show up! Good tip to share with others Though I'm sure it'll come to the app eventually, as it's already on the mobile web UI

[topaz otter]

Yep I see everything on mobile Web.

Also, I see they made separate GPTs for Data and DallE, why isn't there one just for Web Browsing?

Would ne nice to use when wanting to just search for recent events or as a search engine.

[twilit hatch]

what does it mean when it says "unpublished" changes on my GPT but I keep trying to update? Does that mean it's just no public?

[untold seal]

Guys ive given chatgpt 4 images and the text generation is pretty slow.....

is it the same for you guys?

[honest jay]

Good question! I don't know why they don't have one of those. If you wanted, you could create a super barebones custom GPT with no instructions, and only the Web Browsing capability checked in "Configure"--this should give you what you're looking for.

[topaz otter]

@honest jay That's a good idea, thank you! I just found it odd they made separate ones for all the features (Data, Dalle, Classic 4), but left out a Browsing only model.

Interestingly, if I select one of my older Browsing chats from when the models were separate, and go to info, it says 'web browsing' for the model, so it must still exist somewhere.

[honest jay]

I just did a quick test: I think you can manually navigate to:
https://chat.openai.com/?model=gpt-4-browsing
And get what you're describing! I started a new chat from there and then checked model info, it claims to be the Web Browsing one! I'm not sure if this would be using Turbo, though, so you might (?) still get better results making the dummy custom GPT.

[topaz otter]

Yep so you can still manually use it. Looks like they just didn't make a custom gpt for it. Which also explains why I can use an older Web Browsing chat and it works still.

Probably best to just use the All Tools model though and specifically start with 'search the web for _____'

[civic grove]

Hi again. I thought to see a pop-up when I logged in that ChatGPT is now able to search the web by default no? I'm asking it questions, and it doesn't do so :(. "While I can't directly access or search these resources for you, ..."

[honest jay]

Hi! The multimodal model is separate from the plugins model we were discussing earlier. Go to where it says "Plugins" and switch to "GPT-4"--that's your multimodal model capable of web search and more!

[toxic canyon]

does the instructions on custom gpt have character/word limit

[honest jay]

I think it's 8000 characters and it includes all text on the config screen (instructions, conversation starters, etc.)

[toxic canyon]

ah icic, ty

does the REQUIRED_ACTIONS need to be all on separate columns like
`REQUIRED_ACTIONS:

• Action: Google Calendar Find Event yadiyadi
  Confirmation Link:

REQUIRED_ACTIONS:

• Action: Google Calendar: Find Event yadiyadi
  Confirmation Link: `
  or can it be like

`REQUIRED_ACTIONS:

• Action: Google Calendar Find Event yadiyadi
  Confirmation Link:
• Action: Google Calendar: Find Event yadiyadi
  Confirmation Link:`

[civic grove]

Thanks again! That helped a lot 🙂 https://chat.openai.com/share/863c1be4-3301-4f6e-a029-495ae96e07f5

It showed asking Bing (lol), stackoverlow and github.

The two Errors however seem to indicate that there was a problem using a search engine.

[honest jay]

Bing is the search "vehicle" for the Browse feature, it's due to OpenAI's partnership with Microsoft. As for what caused the errors: I'm not certain there! Does it happen if you ask for something more basic like, "Can you check online for recent global news stories?"

[abstract pendant]

How do I see the custom instructions that are in effect for an older conversation? I can't find the button anymore for that.

[stiff nexus]

Is there a way to find created GPTs?

Because Explore section on chatgpt is limited

[civic grove]

Custom instructions are in the pop-up when you link on your name in the bottom left.

[abstract pendant]

Carlo: I think those are for what it will use for new conversations going forward. I want to see what they are for older conversations, which do not automatically get the current instructions.

[honest jay]

Look out for the launch of the Plugins Store later this month. For now, check out #1171489862164168774

[flint bolt]

is ChatGPT down?

[civic grove]

no

[flint bolt]

I only get "Hmmm, something seems to have gone wrong" - even after reloading the page. Trying to use a GPT i made.

[civic grove]

Are you using plugins?

[flint bolt]

i don't think so, it just has custom instructions and two txt files attached to it.

[civic grove]

I mean, are any plugins enabled - even though it isn't using them.

[flint bolt]

ah yes, code interpreter

[civic grove]

Do you still get that error after disabling all plugins and reloading the page?

[flint bolt]

no, now it works...

Thanks! 😁

[civic grove]

Then one or more plugins are outdated. You have to uninstalled them and reinstall them.

[flint bolt]

but isn't code interpreter a default plugin or function now?

[civic grove]

See here: #1173211922594795561 message

I had it with another plugin. Aparently they don't use version control, they just say "Hmm... something went wrong" when you try to initialize an outdated plugin.

[flint bolt]

I see. The problem is that "Code Interpreter" is integral part of ChatGPT. They're integrated in the options of the GPT by default. Since it's not in the store either, ig OpenAI has to do it XD.

[civic grove]

I'm happy I could help 🙂

[flint bolt]

Thanks - at least it works, and i don't need code interpreter for that GPT anyways. 😁

[frail fulcrum]

What are the limitations of GPTs knowledge file upload

[faint badge]

Hi everyone. I was trying to use chatGPT when it was overloaded because the new updates. I have a previous conversation that used gpt4 model, but after that outtake the conversation is in gpt3.5. I don't know why it changed, but I'm sure that I was using gpt4 because I used image analysis and gpt3.5 is not capable of that.

Already reported it as a bug.

[tough bramble]

It's a bug, has been happening to me since the update. Basically, every time there's an error, it permanently switches the conversation to 3.5

[weary orbit]

my impression is that the model you were talking with became unavailable

[brisk path]

Do you guys have the new UI?

[weary orbit]

yes, you don't?

[brisk path]

I do

Just asking

[weary orbit]

high five ✋

[brisk path]

Bruh I bought ChatGPT Plus on devday and I already got all the features 😭

[misty osprey]

Can anyone explain why ChatGPT and Microsofts Copilot are kinda bad at reading PDFs

[brisk path]

I even caught a glimpse of a "My ChatGPT" setting that adjusted how ChatGPT works based on how you interact with it

[tough bramble]

Well, this is interesting.. The speech to text bled into the user message and instead of what I said, it sent this to ChatGPT:

This transcript may contain references to ChatGPT, OpenAI, DALL·E, GPT-4, and GPT-4. This transcript may contain references to ChatGPT, OpenAI, DALL·E, GPT-4.