#gpt-models

1 messages · Page 250 of 1

dusky iron
#

why else do you think GPT 4 is split into different expert fields, when u send your prompt it evaluates which GPT 4 instance subtype it goes to.

trim elm
dusky iron
#

i still have hope in OpenAI, provided they show proper results within long, as i feel the quality has greatly diminished.

trim elm
#

GPT-4 is getting dumber everyday

#

but still better than GPT-3.5

dusky iron
#

also mrCrack.

#

if u make a discord bot do NOT use discord Library to code it.

#

it's very inefficient, use Py-cord instead, this is actually maintained, and fully up to date.

#

Bet TheMandalorian is keeping a tight eye on me. 👁️ 👄 👁️

trim elm
dusky iron
#

just saving u a lot of trouble

trim elm
dusky iron
#

ive made quite a few discord bots

#

but Py-cord is the most versatile ive stumbled across, and requires no other imports other than that one library.

trim elm
#

I just remembered I applied so I get the api back when GPT-4 was released and I still haven't get an email lol

wind urchin
#

ok

austere flame
#

is there a way to use plug-ins when working with GPT4 in the playground interface?

supple violet
dusky iron
#

this is sadly intentional

wraith ginkgoBOT
austere flame
late mica
#

When is Web Browsing coming back?

dusky iron
cloud frigate
#

which plugn is best for to yotube url so gpt can understand whats going on video then explain us

bright oyster
odd hearth
#

Bro, custom instructions available for free users and not for paying eu users

#

lmao

obtuse flume
#

ways to "upload" images to gpt-4 api? there are apps that can do it, so "it does not support it" isnt an answer.

maiden gull
#

omg why did they break GPT.4

#

GPT: I have no idea what you are talking about
Me: Look at the message I just previously sent you
GPT: "... I apologize for the confusion. Unfortunately, I don't have access to previous messages in this conversation, so I can't review what I sent you earlier..."

rugged bobcat
#

Hello everyone, I am seeing this message - "This code interpreter (beta) chat has timed out. You may continue the conversation, but previous files, links, and code blocks below may not work as expected" . What does this mean ?

minor finch
#

PLEASE FIX YOUR AI IM AM DONE OF THE DUMB ANSWER

#

PLEASE !!!!

wide minnow
maiden gull
wide minnow
maiden gull
#

reactjs/nextjs frontend and djangro rest framework API backend

wide minnow
#

fun! Angular for mine

maiden gull
#

Nice. probably worth learning at some point

dusk knot
dusky iron
#

I think the main issue lies with context length and censoring

#

Because believe it or not LLMS are not called black boxes for fun, if you give it a restriction, I don’t think the Devs know from start to finish what it’s holding back, from what I can gather it makes that decision itself based of a Master Prompt, because if built into the LLM jailbreak would not work, so it has to be some sort of master prompt, I duly note that I only used the word jailbreak in context to my explanation and i do not condone or accept such exploits which is a clear violation of TOS

bright oyster
dusky iron
#

It’s not itself

#

Either they’re working on something huge or it’s downright getting nerfed

wide minnow
#

Me: "nah, lets take out 5 all together. Show me the full update please
GPT: "Certainly! It seems there might have been a context shift or I might be missing previous parts of the conversation. I'll provide a response based on the information given, but please clarify if I need to adjust anything specific."
Holy crap... yes a context shift!!! What the heck is that and why the heck is it happening all the time now?!? Like omg it finally identified exactly what the bleeping issue is

#

Also the 'context' message was just the one prior and it's just a letter to a customer, nothing too fancy.

wide minnow
#

GPT: "I apologize for the confusion. It seems that the conversation history leading up to this point is missing from my view, and I don't have access to the previous context or instructions you provided."
...O rly?

maiden gull
#

Haha. It's so frustrating though

wraith ginkgoBOT
#

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 @open crescent immediately.

lilac vapor
#

I need help adding a api plug in to GPT-4, I'm kinda lost to how to put the 2 together after I create (still trying) my API

balmy sinew
radiant granite
#

"Can ChatGPT analyze videos now? I want it to analyze a vlog for me so I can learn and mimic the design approach of that vlog.

red snow
#

Hi guys, just want confirmation. GPT 4 api access, will require successful payment. Does it mean, subscribing to GPT 4 plus would remove the restriction ? Ideally, when the user setup paid account on https://platform.openai.com/account/billing/overview, openai can charge 1$ to automatically remove the restriction ? Note that, im on free credits granted by openai, so its optional to subscribe to paid subscription plan.

elfin crescent
radiant granite
wraith ginkgoBOT
minor finch
# bright oyster whats the issue?

Its not even an answer censoring problem Im asking the ai to adapt a french sentence based on a english sentence but this idiot keep messing up and being stupid

#

For god sake just adapt the sentence

#

He cant do that because the ai is too stupid too dumb

#

In 30 try he managed to understand the thing ONE TIME this is unacceptable

faint topaz
minor finch
faint topaz
#

keep in mind gpt is not trained specifically for translation

minor finch
#

Oh Come on its not hard to change one wors

faint topaz
unborn zinc
#

Can you enter jamovi data set into gpt4?

still lagoon
minor finch
still lagoon
#

This is why head attention is also made for, too.

faint topaz
faint topaz
minor finch
#

I dont paid the subcription so they can pursuit model Im paying for a working product not half working product not 75% working product

faint topaz
minor finch
#

You are telling me that changing ONE WORD or ok at least 3 WORD is a problem for an ai

faint topaz
minor finch
#

The entire sentence is written all it need to do its to change ONE WORD

faint topaz
#

there are particular scenarios that predictive models struggle with. it's bad at hybridity for example so if you're working with melanges of french and english it may not fluently cross the streams, so to speak, because almost always (statistically) that doesn't happen. without specifics though i can only assume, and i'm not an expert in GPT-4 - just interested, and responsible for maintaining a contented community

ornate horizon
#

How to paint with AI

faint topaz
minor finch
#

A Real one

ornate horizon
#

Why GPT's AI painting is so expensive

faint topaz
minor finch
faint topaz
minor finch
#

Guess I go Ask for a refund of all the month I paid then

#

They are clearly not capable to answer my needs

faint topaz
#

good luck!

unborn zinc
#

if it does imma buy it but not sure if it does aye

faint topaz
unborn zinc
#

jamovi is like excel spreadsheet

minor finch
ornate horizon
#

How long after a breakup is the best time to recover

slim osprey
#

Anyone know how to fix the problem where there is nothing showing except "new chat" and the chat bar?

faint topaz
ornate horizon
#

I want to prepare some surprises for him, but I don't know what to prepare

wraith ginkgoBOT
#

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 @open crescent immediately.

lyric cairn
# minor finch Dont good Luck if they refuse I attack then in court

Don't worry many of us here have been saying the same thing that for a week now. It has got progressively worse since the August 3rd update. But somehow it's because WE'RE doing something wrong apparently. You're not alone, i might cancel my subscription if they don't fix it by the time my next billing cycle arrives.

lyric cairn
slim osprey
charred gull
#

can you speak chinese

hallow owl
minor finch
#

while clearly all the sign point to the ai

#

Openai is going to lose their customer and that it

quick swallow
#

/info

leaden wasp
#

Hello can anybody tell me what is the best possible way to connect gpt-3.4-turbo or gpt-4 api with internet so that my chatbot can answere questions those are asked on real time data?

wraith ginkgoBOT
dusky iron
slender pivot
#

Anyone know if there's a message count tracker or anything? Or how the time limit works for the messages? I get the limits, 50 per 3 hours, but I don't understand how the cut off is being applied

supple violet
muted zenith
#

i've runed out messages in gpt-4 and my chat've changed to gpt-3.5, do you know how can i change it back?

limber nimbus
#

can we create a real life jarvis ai assistance

supple violet
charred ermine
#

👀

slow stream
lilac meteor
#

does anyone know the context length of the gpt-4 0613 in playground?

obsidian bolt
#

gpt-4 appreciation comment - gpt4 was better today, could be even better but relatively better

maiden blade
#

Is there a website that calculates my prompt token so I know whether gpt4 on the openai website can handle my prompt or not?

lilac meteor
dusky iron
#

although not sentient an AI assistant is easily made through a simple YT Tutorial.

wraith ginkgoBOT
#

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 @open crescent immediately.

grizzled summit
#

what are some good prompts for gpt-4? i usually use the 'act as' prompt, are their any better ones for learning about something?

boreal tundra
limber nimbus
#

hlo

vital yarrow
#

Do we have any idea when the internet search function will be back on?

hallow owl
elfin crescent
reef lily
#

Hi! I am in need of some HELP. I have exhausted all my ideas and have been unable to get any help directly from the email or chat options to get help.

My chat GPT account has been taken over by someone who speaks Chinese. I know this because my entire list of "conversations" has been bumped for Chinese COnversations. THis is not spam. They are actual requests of chat gpt and they are using my profile. It has caused a major disruption to my work. I have changed all my passwords and the chat GPT password because I did think I was hacked. But, the queries in chinese lettering continue. I need help with this. HAS ANYONE ELSE EVER HAD THIS ISSUE?

static isle
#

Use whichever fits your use case best, they are all free alongside with BingAI

wraith ginkgoBOT
limpid blade
#

Will access fully roll out soon?

grizzled summit
elfin crescent
maiden gull
cloud frigate
#

Why gpt 4 can read long code

#

While gpt 3.5 can do same thing

full cape
#

I just use gpt 4 cuz I pay for it lol

oblique elk
#

Can someone help me understand how embeddings data is actually used in practice?

For instance:

Say I have a document that I want to encode into embeddings vectors. I run that document text through the embeddings model to get the vectors list.

If I wanted to then use this embeddings data in a prompt to generate something useful, would I simply paste the list of vectors into the prompt?

maiden gull
#

sounds like a good question for gpt

safe knot
#

I've noticed lately that asking for song lyrics or quotations only brings up some blithering about copyright, which I can assume is yet another instance of the bot being nannied down

dusk knot
#

when does chatgpt getsrealtime training data,

#

its still 2021 isnt it

safe knot
#

probably never, they'll have to hook it up to the internet and the internet is dangerous

lyric cairn
elfin crescent
lyric cairn
#

It's been like 7 months? that's more than enough time LOL. It's useless

wraith ginkgoBOT
static isle
#

Good things are gonna cost money

lyric cairn
floral crypt
#

gpt4 down again?

coarse sedge
#

chat GPT-4 dead?

#

permanent network errors

#

suddendly

floral crypt
#

is working now

faint topaz
lyric cairn
#

Sometimes it does that if it's overloaded

floral crypt
#

yea i think the increase in the limit has made it more often now

faint topaz
wraith ginkgoBOT
#

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 @open crescent immediately.

safe knot
#

welp, gpt4 continues degrading, I'd been using it to help with some fanfic stuff I'm writing by plugging in bits and pieces and asking for it to finish it, and tonight it refused and gave another excuse about "copyrighted material"

hollow geode
#

I don`t fully understand this message limit to 50 every 4 hours. I can still interrect with GPT plus after

safe knot
#

The message limit is for GPT4, after you run out you either wait 3 hours or switch to 3.5

lyric cairn
safe knot
#

well yeah, you can get around it NOW by wasting messages convincing it that it's fanfic and not literally copyrighted material, but that's not very efficient lol

#

I want my smart chatbot back, lol

timid cairn
# 

Assistant: Let me tell you about ethics.

....

timid cairn
timid cairn
lyric cairn
timid cairn
# lyric cairn Gl with that, according to them it's OUR fault that gpt 4 is not doing stuff rig...

To comment on this point: I have a pretty good sense when I make an error with regards to linguistical issues. I have pretty good sense when I make an error in judgement when considering societal considerations for the weight of words when I create a prompt. I have a pretty good sense of how type of training GPT model has gone through and how it effects the way it interprets prompts. I could go on...

#

The level of discrepancy with the initially created prompt vs the final prompt is becoming greater and greater.

#

It isn't that a

lyric cairn
timid cairn
#

'final prompt' cant not be reached, but rather how we tend to 'articulate' ourselves and our expectations for what should occur considering the model and our understanding of 'language'... is becoming too great

timid cairn
#

I felt it was needed to convey the information in a very complete way in the oft chance that they read this chat.

#

checks grammar

wraith ginkgoBOT
timid cairn
#

ChatGPT says my comment makes sense... idk if I can trust it. >,>

rose trail
#

for my use cases gpt4 is better than ever. but outperforming rarely leads to public service announcements as much as claims of underperformance.

#

i'd ask code interpreter to plot me a graph of announcement probability versus product satisfaction. but i'm locked out for another hour

safe knot
lyric cairn
safe knot
safe venture
#

people are not making good decisions with GPT4 like (e.g.) putting it to work manning an eating disorder chatline

#

and those decisions are going to pay off with 'increased regulation'

safe knot
#

also your screenname makes me wonder if you've played Disco Elysium?

hollow geode
safe knot
lyric cairn
safe knot
#

there needs to be a graphic of some sort indicating how many messages you've used or have left

lyric cairn
#

yea agreed ^ I swear sometimes even when i THINK i'm on point, i forget that i resubmitted something, or there was an error that forced me to regenerate.

timid cairn
#

Three months later I got a reply

#

part of that reply was the following

# 
We understand that you have concerns about the cap of 25 messages every 3 hours, especially since you are a paying customer.

We would like to clarify that, to ensure optimal performance and manage demand, we have implemented a dynamic usage cap. Currently, GPT-4 can be utilized for up to 25 messages every 3 hours. Please note that this cap may be adjusted in the future based on system performance and user demand.
#

So I guess that might be why they don't show it... cause it might change on you.

safe knot
#

that was totally written by a human person, sitting at a human keyboard using his or her human hands

timid cairn
safe knot
#

it was sarcasm, don't worry xD

timid cairn
#

lol...

hollow geode
safe knot
#

hmm, I wonder if people would be willing to trade less messages per period for greater capability from the AI

wraith ginkgoBOT
#
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.
lyric cairn
safe knot
#

right?? it feels like the nannying gets worse every day, doesn't it?

safe venture
#

and yes, the lowest common denominator always ruins it for everybody, I generally use the API these days because the guard rails aren't as 'guarded'.

safe knot
#

so I've heard, but the APIs take a bit more elbow grease too, don't they?

wraith ginkgoBOT
minor arrow
#

Is the picture input to gpt 4 still not available?

supple violet
minor arrow
west lichen
#

Gpt4 is funny.... i'm having it help me write a script and previsly it made it so it fetched data in multiple threads, but then we had some db errors with saving the data.. so i asked it to fix that.. but then for some reason without me telling it to it removed the multithreading..

#

And now it removed the rate limit for no reason what so ever xD

elfin crescent
west lichen
#

And that should be a bit smarter

#

It feels like it has become a lot dumber since i went on vacation 2 months ago xD

dusky iron
#

for the reason we call "AI Safety"

west lichen
# dusky iron it has become dumber

Haha yep, but I just got an email from a service I used with the agent thing. They just released a plugins that increases gpts memory. 😄

dusky iron
#

or maybe its just bad data.

dawn forum
#

Does anyone also have problems with gpt4? I keep getting wrong/short answers, so I'm not as attentive as I used to be, but have to write more concentrated, if you know what I mean.., Even if I open a new window.

fathom finch
#

I just received the error: Authorization error accessing plugins / Has anyone seen this?

inland storm
#

My GPT4 has no memory anymore. Like every prompt is a new start of a conversation. I tell it the situation or problem, provide code, and it doesn’t remember the problem. I re-explain the problem and it says it can’t help me without the code. Anyone else experiencing this?

unborn heart
#

How do I enable plugins and code interpreter with the Android app?

west lichen
unborn heart
willow jasper
#

So, these plugin settings are basically per-convo, right?
Even the custom instructions are per-convo because they're applied when a conversation is created and are retained even if those settings are changed later. Correct?

willow jasper
dusky iron
willow jasper
#

@dusky iron ah. Any idea why it is intentionally restricted? Perhaps to keep the app stores from getting mad about paid features that weren't bought on their store being provided?

dusky iron
willow jasper
#

Aye. That makes sense

dusky iron
#

cant see the reason for code interpreter on app me myself personally.

wraith ginkgoBOT
safe sundial
#

I am using GPT-4 to document Python functions including writing doctests. The important part of my custom instructions is: Add setup and cleanup lines for the doctests where needed to enforce test isolation

My prompt is:

Please document this function

def check_if_on_lambda():
    return "AWS_EXECUTION_ENV" in os.environ

ChatGPT gives an excellent answer that includes setup and cleanup steps as part of the doctest. I cannot get GPT-4 to replicate this level of quality. I have tried a number of different temperature settings in the playground but all of the answers do not include setup and cleanup line(s) per my system instructions. GPT-4 does annoyingly sometimes note the problem in its response ("Please note that the doctests in this function are manipulating the actual environment variables of the process running the tests. This could potentially interfere with other tests or parts of the same test suite that rely on environment variables.") but it doesn't follow my instructions to do something about it.

My "System" prompt in GPT-4 is set up like this with the content in the ... parts exactly matching what I've posted in custom instructions for ChatGPT:

Q: What would you like ChatGPT to know about you to provide better responses?

...

Q: How would you like ChatGPT to respond?

...

Does anyone have any suggestions around what I'm doing wrong or how to get a similar level of instruction following quality from GPT-4 as I get from ChatGPT?

gilded crane
#

grrr

[11:18 PM] 
BOT
 Guru: As an AI, I don't have a specific GPT version. However, I'm designed based on advanced language models similar to GPT-3.
[11:19 PM]blackbird93: @Guru but you are supposed to be using gpt-4 model
[11:19 PM] 
BOT
 Guru: I'm sorry for any confusion, but as an AI, I'm currently based on OpenAI's GPT-3 model. The GPT-4 model has not been released yet. However, I'm designed to provide the best assistance possible with the technology available. How can I assist you further?```
#

On the Playground I chose gpt-4 and got the same response

#

ASSISTANT As an AI developed by OpenAI, I'm based on GPT-3 model.

#

Model
gpt-4

#

should I raise a support request ticket or am i missing something?

maiden gull
#

I think they fixed GPT-4. It's been acting better recently

last wind
#

What was wrong with it?

maiden gull
#

Umm not remembering anything from previous messages, even if it was the immediate previous message and all messages have been short.

And just generally giving stupid answers

lament saddle
wraith ginkgoBOT
faint topaz
wide minnow
#

goodness... other than the context shift bug with the current gpt4 it also just seems to be lazier with tasks I would ask it for before. Ran into that today where I provided my resume and the description of a job and asked for it to tailor my resume to highlight skills and wording to this job description "Resume and job description included in prompt"
It used to go through the whole resume and tweak words and move things around all over. Now it gives me a full page response of what its going to do and when it performs the action it amounts to adding a small section of 3 bullet points.
I might have to revert to 3.5 for a bit as 4 feels extra lazy.

fading junco
#

does anyone have a wishlist of things they wish chat gpt 4 had?

maiden gull
#

more smarter

wide minnow
#

remove the genz work ethic they added in the latest enhancements of GPT4

fair iron
wraith ginkgoBOT
#

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 @open crescent immediately.

ashen vapor
#

Gpt 4 gives lots of placeholder code when refactoring have to expliciet ask it to not and give me full code each time

#

One would assume on plus openai would not enforce such token savings on us since we pay right? Ita workarounable but irritating nonetheless

south briar
#

@ashen vapor and how to make it write with no placeholder?

bleak crypt
#

User
can you explain the meaning of "please generate full code for me." to me, please?
ChatGPT

Certainly.

When you say "please generate full code for me," it implies that you are requesting the complete, ready-to-use code for a particular task or program. Instead of just snippets, modifications, or patches, you are asking for the entire codebase. This often means that all the necessary components, functions, and logic are combined into one cohesive piece of code, so you can run it directly without needing to integrate or add anything additional. Essentially, you're asking for a turnkey solution.

opal mortar
#

Has there been any provided any ETA on a return of the browsing plugin.

bleak crypt
#

My next message was: and what did you do? And i got : you reached the limit...

fair iron
#

There is no update about browsing sad

ashen vapor
ashen vapor
wraith ginkgoBOT
slow stream
rose trail
#

i like to ask for the code to be incrementally generated and explained

#

but yeah i dont think there is a foolproof prompt to never get placeholders

tidal turret
#

do you guys like the browserop plugin?

#

its in the popular section

wraith ginkgoBOT
wraith ginkgoBOT
#

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 @open crescent immediately.

wraith ginkgoBOT
ashen vapor
thick agate
#

i want to say that i think things improved over the last 5 days

#

since i complained so much i thought i would come and say something positive

maiden gull
#

I have a groundbreaking idea.

How about when you send a message and there is an error, they don't remove the text input box to send another message?

#

It's crazy, I know...

thick agate
#

some simple things dont get done easily

maiden gull
#

When your the size of openai you should have the resources to do that kind of thing

thick agate
#

also one custom instruction?

#

so i have a doc and copy paste a custom insstuction for each different task

#

not hard to make a list and activate the one u want

#

ill be patient

thick agate
#

i see it in this EA game i play

#

lots of big software companies miss the details

#

midjourney same kinda issues... simple things in the interface could make things a lot better

#

when was the last update?

wraith ginkgoBOT
maiden gull
#

idk I think they upadte constantly in the background. Last major update was like 2-3 weeks ago I think

#

Another idea. I don't know if openai can pull it off, but some kind of character counter to let you know if you are over the limit before sending a message...

#

idk, their web dev team might not be up to the task though

acoustic vapor
# maiden gull idk I think they upadte constantly in the background. Last major update was lik...

javascript:(function() {
var textArea = document.getElementById('prompt-textarea');
if (textArea) {
var counter = document.createElement('div');
counter.id = 'counter';
counter.innerHTML = 'Character count: ' + textArea.value.length;
counter.style.position = 'fixed';
counter.style.bottom = '10px';
counter.style.right = '10px';
counter.style.padding = '5px';
counter.style.backgroundColor = 'rgba(0, 0, 0, 0.7)';
counter.style.color = 'white';
counter.style.borderRadius = '5px';
counter.style.zIndex = '1000';

function insertAfter(newNode, referenceNode) {
  referenceNode.parentNode.insertBefore(newNode, referenceNode.nextSibling);
}

insertAfter(counter, textArea);

textArea.addEventListener('input', function() {
  counter.innerHTML = 'Character count: ' + textArea.value.length;
});

} else {
alert('Text area with ID "prompt-textarea" not found!');
}
})();

floating overlay version that scrolls with you.

#

You can do it yourself.. 🙂

wide minnow
acoustic vapor
#

Idk I have never hit the limit. :/

thick plover
#

Hey everyone. I want to do a contest for prompt engineering.there will be no prize but a show of skill and uses. The model used for it will be normal GPT 3.5 so all can enter the contest (NO GPT 4) you can use the app on IOS or android or the Normal Website if you are on PC.

The Theme is: UltimateGPT
This will have a selection of a different modes for ChatGPT such as AI art mode,prompt mode (which it will make a prompt that works) and MusicLM prompt mode (go research it if you have not)

If you want to enter please DM me!

wraith ginkgoBOT
#

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 @open crescent immediately.

acoustic vapor
#

Write a concerto using ABC notation.

#

X: 1
T: Elegy for a Burning World
C: ChatGPT, Composer
M: 4/4
L: 1/4
K: Em
%%score (Piano1 Piano2 Orchestra Flute Clarinet Violin1 Violin2 Cello)
V:Piano1 clef=treble
V:Piano2 clef=treble
V:Orchestra clef=bass
V:Flute clef=treble
V:Clarinet clef=treble
V:Violin1 clef=treble
V:Violin2 clef=treble
V:Cello clef=bass
% Piano 1 Part
[P:Piano1] |: "Adagio" [EG]3 [F^A]2 [G_B]2 [A=C]3 | [EG]3 [F^A]2 [G_B]2 [A=C]3 😐

% Piano 2 Part
[P:Piano2] |: [EG]2 [F^A]2 [G_B]2 [A=C]2 | [EG]2 [F^A]2 [G_B]2 [A=C]2 😐

% Orchestra Part
[P:Orchestra] |: _D,3 _D,3 _D,3 _D,3 | _D,3 _D,3 _D,3 _D,3 😐

% Flute Part
[P:Flute] |: [E2G2] [A2C2] [B2D2] [C3E3] | [E2G2] [A2C2] [B2D2] [C3E3] 😐

% Clarinet Part
[P:Clarinet] |: [E2G2] [F2A2] [G2B2] [A3C3] | [E2G2] [F2A2] [G2B2] [A3C3] 😐

% Violin 1 Part
[P:Violin1] |: [E2G2] [F2A2] [G2B2] [A3C3] | [E2G2] [F2A2] [G2B2] [A3C3] 😐

% Violin 2 Part
[P:Violin2] |: [E2G2] [F2A2] [G2B2] [A3C3] | [E2G2] [F2A2] [G2B2] [A3C3] 😐

% Cello Part
[P:Cello] |: _D2 _D2 _D2 _D2 | _D2 _D2 _D2 _D2 😐

maiden gull
supple violet
wide minnow
# acoustic vapor Idk I have never hit the limit. :/

looks like it triggers "The message you submitted was too long," around 30k characters for me. I tweaked it to change to red when I get to 30k. Here's that version:

  var textArea = document.getElementById('prompt-textarea');
  if (textArea) {
    var counter = document.createElement('div');
    counter.id = 'counter'; 
    counter.innerHTML = 'Character count: ' + textArea.value.length;
    counter.style.position = 'fixed';
    counter.style.bottom = '10px'; 
    counter.style.right = '10px';
    counter.style.padding = '5px';
    counter.style.backgroundColor = 'rgba(0, 0, 0, 0.7)';
    counter.style.color = 'white';
    counter.style.borderRadius = '5px';
    counter.style.zIndex = '1000';
    
    function insertAfter(newNode, referenceNode) {
      referenceNode.parentNode.insertBefore(newNode, referenceNode.nextSibling);
    }
    
    insertAfter(counter, textArea);
    
    textArea.addEventListener('input', function() {
      if(textArea.value.length > 30000) {
        counter.style.color = 'red';
      } else {
        counter.style.color = 'white';  
      }
      counter.innerHTML = 'Character count: ' + textArea.value.length; 
    });
    
  } else {
    alert('Text area with ID "prompt-textarea" not found!'); 
  }
})();```
#

Now I have a pre-emptive warning when I'm about to post something that is likely to trigger the too long message. Thanks fuzzball!

acoustic vapor
#

Wonder if we can add a IFrame with google/bing/duckduckgo below this and add a search bar to add on page searching 🙂 I bet we can.

maiden gull
#

If you are using FireFox you can do this,

Create a bookmark, name it whatever you want.

For the URL put in that javascript

#

click the book mark whenever you want character count

wraith ginkgoBOT
night venture
#

I read Custom Instructions should be open to all countries now

#

I'm a GPT4 subscriber but I still don't have it

floral grotto
#

Slowly Rolling out

blissful dagger
#

does anyone know if there is a way to increase the tokens in the context window using just regular chatgpt

honest glacier
#

Like @night venture says, I'm also plus subscriber and don't have custom instructions...odd.

I have plugins and code-interpreter. Missing instructions

gusty tundra
wind urchin
#

ello

#

I was wondering if it'd be possible to have a model watch videos that are narrated and talk about stock technical analysis. Could I make it watch and understand charts along with the narration? I've already worked on getting over 200+ hours of video from someone into a text transcript. I was just going to fine tune it with the transcript but then got curious if I could make it also comprehend the video aspect. I'm assuming this isn't possibile but am quite curious

#

Hellooo.

wraith ginkgoBOT
wraith ginkgoBOT
#

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 @open crescent immediately.

tough chasm
#

looks like it triggers "The message you submitted was too long," around 30k characters for me. I tweaked it to change to red when I get to 30k. Here's that version

wraith ginkgoBOT
sly lantern
#

My gpt4 is not working ..Is there someone else or just me ?

gaunt plinth
#

I find it pretty frustrating that I'm paying the same amount as american users, and yet don't have access to the features they do.

lean helm
#

uuuuuh so theres this new feature where you can tell it how you want your replies to go, how does it actually work?

#

are there like any comparatives to before and after filling those out

safe venture
lucid tulip
night venture
wraith ginkgoBOT
#

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 @open crescent immediately.

dusky iron
gaunt plinth
#

Yes, it's my regions fault I don't have access to it. But it's OpenAI's fault that I pay as much as those with all the features. As such I think they should lower the fee for European plus users.

wraith ginkgoBOT
ember otter
#

I have a dumb question. I'm not tech savvy, I have though a powerful prompt structure to be used to help people generate their own personal "perspectives of everything". Its a proof of concept that might be really helpful to people. I want to give it away freely, but need help as I'm not very tech savvy. How are people pasting in prompts that are longer than the text limit? Whoever helps gets to say they helped with "that" towards a thing for the future. Guaranteed 😆

shy moss
#

Is there any plugins or other ways that can increase the amount of posts in a single chat that ChatGPT can access?

placid dagger
#

Do any of you receive this message recurrently: Our systems have detected unusual activity from your system. Please try again later.?

old granite
#

i was in a chat and came back to it after a day or so to ask it to summarize, first it started with The previous model used in this conversation is unavailable. We've switched you to the latest default model. btu it was still using gpt-4, then on top of that it spat out absolute gibberish totally unrelated to the chat? tried regening it and rewording my prompt but nothing worked? is this a common issue?

viscid lodge
viscid lodge
viscid lodge
shy moss
viscid lodge
#

hmm odd, it did that too me too (ages ago), some times when the chat gets really long it has issues trying to get all of it for context. I will look into this for you, just to note that chagpt cant pull information from other chats

shy moss
viscid lodge
#

yeah it does struggle with long chats

#

thats why i prefer to keep chats short ish, however you can get decently long, i suggest you play about and see what you find as a limit

#

and i broke chatgpt lol

shy moss
#

will making the material more concise/shorter increase the post number i can recall? like is it a letter count per say?

viscid lodge
#

yes, keeping messages as short as you can and to the point is always a good idea, the less it has to process per message means more messages it can process

shy moss
#

so with a large base i will just have to restart chats i guess for now summarizing into a main chat i guess

viscid lodge
#

yep, i have to do that at times with chats that drag out, i will start a new chat, summarise the old one and tell it that you are continuing from a past chat and this is a brief summary of what was talked about, if you find a better way to do it do share

old granite
#

no file uploads

#

it seems to have no memory of the past chat, and even then it gives just totally random responses

viscid lodge
#

Hmmm odd, did chatgpt send any sort of file?

old granite
#

i mean its sending the links to the diagram plugin

#

which display as images yeah

viscid lodge
#

Yeah it won't save those images too long however it forgetting chats os odd, do you have history off by chance

old granite
#

no

#

Of course! Here are two examples of how you can create a dictionary using the value from a list as the key and the frequency of that value in the list as the value.

Example 1: Using a For Loop

# List of values
values = [10, 20, 10, 30, 30, 30]

# Create an empty dictionary
frequency_dict = {}

# Iterate through the list of values
for value in values:
    if value in frequency_dict:
        frequency_dict[value] += 1
    else:
        frequency_dict[value] = 1

# Print the dictionary
print(frequency_dict)
# Output: {10: 2, 20: 1, 30: 3}

Example 2: Using collections.Counter

from collections import Counter

# List of values
values = [10, 20, 10, 30, 30, 30]

# Create a dictionary using Counter
frequency_dict = Counter(values)

# Print the dictionary
print(frequency_dict)
# Output: Counter({30: 3, 10: 2, 20: 1})

Both examples will give you a dictionary where the keys are the unique values from the list, and the values are the frequency of each value in the list.

#

like this is an output, but ive never mentioned a dictionary

#

or python for that matter

viscid lodge
#

Hmmm, odd I'll have to chase this one with some others

#

It can misbehave and add random parts at times

old granite
#

any prompts youd reccomend that could get it back on track or should i just try to start a new chat and explain the contents of the broken one to it?

viscid lodge
#

you can always ask it to continue what it last said or reword it

#

if it goes off track you can either tell it that it went off track a bit fromt he origional prompt, have you set custome instructions?

old granite
old granite
viscid lodge
viscid lodge
old granite
#

Thanks anyway, started explaining the chat to a new agent

dusky venture
#

is it against the rules to ask someone to ask something to gpt 4?

wraith ginkgoBOT
sonic grove
old granite
#

Yeah no I did try to remind it, but it couldn't even provide the chat immediately before it

#

Very strange

wind urchin
#

"The core Claude 2 transformer model itself still has a 4096 token limit. But with the retriever, it can provide summaries/passages from documents up to 100k tokens."

#

Does gpt-4-32k API have the "true" 32k context window, unlike Claude2?

wide minnow
# safe venture if you're sending that many characters, you're blowing your entire context limit...

due to the current context shift bug with gpt4, I find I have to include more and more context into each prompt. Essentially using the webchat is now very similar to using gpt4 via api in which it is like a stateless client and you have to include the context and history on each request. The best feature of the webchat was whatever magic formula they used that allowed it to maintain context so well for so long in a conversation. Now it forgets the previous message if I don't remind it.

pallid scroll
#

What do you guys think about the last update (3rd of august)?

unreal marsh
wraith ginkgoBOT
#

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 @open crescent immediately.

lyric cairn
wide minnow
candid shoal
#

gpt-4-0314 is still very good

#

the new model is the problem

#

it sucks, but it was probably designed to reduce ChatGPT costs

wide minnow
lyric cairn
wide minnow
#

He means api or playground

lyric cairn
#

Gotcha, thanks

wraith ginkgoBOT
dusk grove
#

How do you access the API/Playground version for GPT-4-0314? I have been using the Plus version, but the quality is just really horrible at this point.

#

It only shows the GPT3.5 models for me at the moment.

warm elk
#

are here any new news about web gpt?

knotty grotto
west lichen
#

Gpt4 has become utterly useless when it comes to coding, it seems to forget everything in basically one prompt xD Sure they are a bit long but not overly long. It's like it got altzeimers.. and completely forget what we where discussing, even in just the previous prompt xD

#

"Please provide the conditions you used for plotting shapes, and I'll integrate them into the strategy for trade entries." And i tell it in the next prompt that here is the entire script.. then it starts to summarize the script and says this in the end "If you have any specific questions or need further explanations about any part of the script, please let me know!" It's become utterly useless xD

remote plank
#

Isn't GPT-4's limit 8k tokens?

west lichen
#

It is, but i'm definately not using 8k tokens

remote plank
#

no but like

#

I feel like my response exeeded the limit .__.

#

Its so massive.

west lichen
#

ah, i thought you where talking about my problem 😄
My gpt is acting as it reached the limit and forget what we are talkign about , but i def has not used all the tokens xD

remote plank
#

hmm

west lichen
#

I have to constantly remind it about what we are doing, and make it reread the conversation

remote plank
#

GPT-4 was writing to me for like 5 minutes..

#

lol

west lichen
#

lol, they got issues xD

gloomy olive
remote plank
#

^ happend to me sooo much. X_X

gloomy olive
remote plank
gloomy olive
remote plank
gloomy olive
#

its defi weird thats for sure 😄

remote plank
#

also, what is the 🧩 role?

gloomy olive
#

i like solving puzzles 😄

remote plank
gloomy olive
#

yeah no its just a helpers role really.. for positive contributions for/to the community.

remote plank
gloomy olive
#

devs / any openai staff will always be Gold coloured, much like what they're sitting on top of (chatgpt, which is golden 😄 ).

remote plank
gloomy olive
#

👋 have a lovely evening

crimson echo
#

I have the plus version and every time it changes to version 3.5. You choose version 4 and immediately jump to 3.5. Anyone else have the same problem?

gloomy olive
#

nope

#

if you're on PC , you can try pushing CTRL SHIFT R if you're using chrome, to refresh your browser. perhaps it helps 🤷‍♂️

floral grotto
#

They recently made it where it goes to the last one used.

#

Unless a Chrome or FF extenion is preventing it, could be a possibility

echo shale
#

I enabled the "custom instructions" option and wrote the inputs for chatgpt. Now when I go to deactivate it loads, loads and does not deactivate, it keeps turning the wheel, but it does not deactivate. Has anyone had this problem? I am in need of urgent help.

floral grotto
#

Talking about this

#

The enable for new chat toggle

#

@echo shale

fallow scarab
#

i just paid for gpt 4, have only 25 for 3 hours, i want to make sure that i have access to internet

#

does gpt have access to internet?

verbal carbon
fallow scarab
#

waste of money

gloomy olive
digital cliff
#

Hello guys, @sour rune May I ask why this function still not available for UK users?: "Custom instructions are now available to ChatGPT users on the free plan, except for in the EU & UK where we will be rolling it out soon!"
Anyone know any reason why is restricted in my country?

gloomy olive
wind urchin
#

I've got instructions set up and there is virtually no difference

gloomy olive
#

then your use case might not be specific enough for it to matter? 🤔

floral grotto
digital cliff
#

Strange...I don't think some customisation like custom instruction have any law related issue....

floral grotto
#

Your best bet it to stay updated with their twitter account

digital cliff
wind urchin
gloomy olive
floral grotto
floral grotto
wind urchin
#

does anybody know if they'll do a rollback eventually after the last atrocious update?

gloomy olive
sour rune
#

my apologies!

digital cliff
sour rune
gloomy olive
digital cliff
#

I accept your answer, obviously, just no make sense to restrict a simple customisation...Anyway. Thank Lucas

sour rune
#

i'm sure it's more complicated than that, but hopefully you get to play around w custom commands like that soon 🙂 no problem

gloomy olive
#

👆

wind urchin
#

has there not been a massive amount of complaints regarding the last update?

sour rune
wind urchin
#

I would assume is pretty common knowledge at this point that gpt4 suffered a very dramatic change after the last update

#

I would just like to know if something will be done about that

sour rune
#

may you enlighten me on what specifically?

wind urchin
sour rune
#

i assume the team is just trying their best to find the equilibrium between the two factors, and they always consider feedback posted here or elsewhere. hopefully the next update tickles your fancy 🙂

wind urchin
#

I assure you it's not about tickling my fancy, it's about gpt4 going from a coding beast to nearly incapable

floral grotto
#

Considering the beginning of GPT4 and GPT as a whole, it's def improved.

#

Nothing is Perfect, but def improved from the start.

sour rune
#

i think incapable is a stretch but if you have any specific examples feel free to share in #1050184247920562316

wind urchin
#

alrighty, I appreciate your input regardless

ember otter
wraith ginkgoBOT
#
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.
wraith ginkgoBOT
tame brook
#

I would like support for dolphin language

#

🐬 need a.i too

tame brook
floral grotto
#

First time i've seen anyone talking about the response ti me of it, been pretty fast for me.

tame brook
#

hes saying they set it to respond too fast and it cant code now, its handi

signal flint
#

how about the gpt functions api only taking paramters of object or boolean huh

wraith ginkgoBOT
#

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 @open crescent immediately.

lyric cairn
#

I don't get GPT 4. It says it has nothing to do with flagging content. But when I ask it to explain why it was flagged, after it explains it was done in error, the flagging magically stops. Which by it's own logic it shouldn't, since it has nothing to do with flagging content apparently.

#

Ugh I can't even share the chat to show what I mean because it's disabled

tardy heart
#

had it build me functions that i couldnt figure out with postgres lol

wraith ginkgoBOT
thin mulch
#

hey Guys,
i know GPT-4 is bit wired now when using chatGPT for coding.
what about API? is that the same? do you guys think API is still better?

#

thanks

remote gate
thin mulch
#

why they dont care? are they trying to fix GPT-4?

radiant tartan
#

Guys, is there any way I can use the api and tokens and all to make gpt-4 a little more smart? gosh I'm getting angrier each day with it's incapability

wraith ginkgoBOT
feral charm
#

Is web browsing going to be back soon ? It was helpfull to have recent informations

wraith ginkgoBOT
#

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 @open crescent immediately.

jolly crag
dusky iron
west lichen
#

gpt4 still being a dork it seems, i have the web browsing plugin enabled, i tell it to check the documentation AND providing the exact links so it can look up info on fixing my script.. it totally ignores checkign the links, "fixes" the script and produces 5 new errors xD

#

And now when it went over my script and fixed an error in two places, it just ignored a line with the exact same error.. xD

wind urchin
#

@sour rune seems I wasn’t all that wrong

crystal summit
#

Im looking for cases where GPT/AI has been used as a shopping assistant... Any tips?

wraith ginkgoBOT
wide minnow
feral charm
wind urchin
ember otter
wraith ginkgoBOT
#

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 @open crescent immediately.

west lichen
hazy comet
#

is it possible to save a rule to chatgpt prompt e.g. "dude please give me the full code" - so it would never skip stuff with comments? Or that is not yet supported?

acoustic vapor
hazy comet
#

ok so this is coming to EU as well at some point?

acoustic vapor
#

Yes it is. At some point.. IDK when.

hazy comet
#

Yeah... I understand that it's a big thing to ship so far away to the other side of the ocean. Might take months - if the weather is good that is.

acoustic vapor
#

Um its due to the your laws. which are top notch and require careful planning not to get sued.

teal orchid
#

I'm frustrated with that Message when I tried loggin in. When I logged in to any other device on which I have to Use Chat GPT 1st I have to logged in google and than on Chat gpt, I need to change my Authentication Method. I want to logged in through email and password.
Fed up from this error I have Chat GPT Plus.How can I change the authentication method

wraith ginkgoBOT
jovial juniper
#

Is Browse coming back already? Bing browse utterly sucks in comparison 😛

hazy comet
hazy comet
jovial juniper
#

What makes you so certain

acoustic vapor
wraith ginkgoBOT
#

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 @open crescent immediately.

wraith ginkgoBOT
wide minnow
# hazy comet to have a prompt prefix? are you serious right now?

it's a trickle rollout. Many/most of us paying subscribers don't even have the 32k model access yet.
The Custom Instructions help a little but I haven't seen anything profound yet. Just kinda lets you set your own guidelines a bit. It takes them somewhat loosely.

slow stream
jovial juniper
slow stream
#

I have always thought the plugins worked better then the old browser plugin. But just depends on what you are using it for I'm sure.

#

Unfortunately something happen to the best one LinkReader but others are still there

jovial juniper
#

Thank you! I’ll give them a whirl.

What would you recommend for say pointing to a GitHub repo and asking it to request a step by step guide on installing said software.

#

You’d think openAI might lift that restriction on their own website 😂

unkempt summit
#

I don’t know what’s happening, but it is getting genuinely frustrating how many times I have to plug in instructions to do anything, even on GPT4. Are they working to improve it or?

I’ve been using it since the beta and it’s really starting to frustrate me.

warm merlin
#

when GPT-5

novel solstice
#

so did I read somewhere that gpt 4 comes with text audio

static isle
wraith ginkgoBOT
elfin crescent
#

Not text to speech

willow sparrow
#

r

merry field
#

can chat gpt 4 code a hole bot like a bot that can think and stuff and controll things like a smart tv our light?

wraith ginkgoBOT
slow stream
limpid chasm
#

someone with gpt 4 can dm me , i have some questions please

fair iron
warm elk
ripe abyss
#

Is gpt 4 going to be free with the announcement?

Or did I understand it wrong.

hallow owl
acoustic vapor
#

Isn't GPT 4 free on bing ??

#

at least i have the option to use it.

wraith ginkgoBOT
lyric cairn
edgy abyss
#

How can one integrate chat GPT into an insurance platform to offer personalized health coverage plan to people based on location and health history?

patent junco
#

Anyone have any experience using GPT-4 over Azure instead of OpenAI? Just overall performance and reliability and maybe any experience across zones.

hazy comet
#

How can I use chatgpt to implement universal healthcare coverage system to USA?

snow marlin
#

Hey is there a way to chat with your SQL Database using GPT-4?

tardy heart
south finch
marsh brook
#

I want gpt 4 but I ain't paying for it

lilac meteor
marsh brook
#

I will

#

But nah

jovial juniper
agile sapphire
#

TIL ChatGPT is not self-aware:

3.5

How do I stop paying for GPT-4?

ChatGPT
As of my last knowledge update in September 2021, GPT-4 hadn't been released yet, so I can't provide specific instructions for stopping payments related to it. However, I can offer some general guidance that should apply to most subscription or payment-based services:

4

How do I stop paying for GPT-4?

ChatGPT
If you're trying to stop a subscription or service related to GPT-4, your approach will depend on the platform or provider from which you're accessing the service. However, the general steps would be:

slow stream
wraith ginkgoBOT
wind urchin
#

your chat gpt is bugged?

edgy abyss
finite crag
#

I can't see my chats, the screen is off to the left

coral stone
torpid sparrow
#

When gpt 4.5????

static isle
worn orchid
#

How much better is gpt4 compared to gpt 3.5 turbo

static isle
worn orchid
static isle
worn orchid
static isle
#

Do you have access though? (GPT-4)

#

Some keys do and don't

#

You can check here through the model selector

worn orchid
static isle
#

Yes

#

Checking however does not add charges

#

Only sending requests

#

If GPT-4 shows up in the model selector, you have access

worn orchid
static isle
#

You do not have access then unfortunately

#

You may be able to get access after a successful invoice payment, but I have no way of ascertaining that as I myself do not have access

#

But when you do, see for yourself whether the extra cost is proportional to the increased output quality of GPT-4

worn orchid
static isle
#

👍

#

Enjoy!

hard crow
#

Man it's so annoying how sometimes it refuses to give the right word count

static isle
hard crow
#

Yup. It's funny when it tells me "yes it is exactly x words" then you check and no, it is in fact not

wraith ginkgoBOT
misty delta
#

What can I start learning about AI as a beginner

shut wharf
#

Can the code generated by gpt4 be transferred to GitHub?

calm light
#

chatcpt not work can somebody help me

knotty grotto
knotty grotto
unkempt plank
#

Btw u guys know how gpt 4s perf dropped alot recently? R they back on track or smth? Have there been any recent news

knotty grotto
calm light
#

When I enter the chat screen, it appears as I asked a question, but there is nothing on the screen.

and above the chat tab these appeared

#

Explain why popcorn pops
to a kid who loves watching it in the microwave

Make up a story
about Sharky, a tooth-brushing shark superhero

Show me a code snippet
of a website's sticky header

Brainstorm names
for my fantasy football team with a frog theme

knotty grotto
calm light
#

ok ı try

#

thank you so mucx

#

its done fixed 🙂

tidal abyss
#

i wonder when gpt 4 will get less expensive & more optimized

#

its so expensive right now i can barely use it

#

i enabled it for a small amount of time without constantly staring at the usage thinking it would be fine and whoosh there goes 40$/most of my available monthly quota in one day

#

even gpt3 davinci was less expensive

flint condor
ionic locust
tidal abyss
#

its literally just a discord chatbot

#

its mostly just functions

#

functions increase cost and usage sooo much

#

so functions combined with gpt4 is not really viable

idle furnace
#

when there will be gpt5 or 6...

#

maybe not for API, but i suspect it will cost like 3.5 now

#

it just needs cheaper computing power

#

and this will come with time

knotty grotto
wraith ginkgoBOT
violet orchid
#

i'm tired of clicking stop generating and having to wait til the response finish generating on background before i can send a new message, anyway to fix that?

left forge
#

But ya I noticed that too as well.

violet orchid
primal vine
#

I’m thinking about getting GPT 4, but what are the benefits? Is it more up to date?

kindred ridge
#

what would cause the error "Authorization error accessing plugins" in ChatGPT 4.0? I'm attempting to use BrowserOp, Prompt Perfect, and Universal if this helps for reference

south finch
frosty tartan
#

you definitely what GPT 4, it is less redundant, it allows more fluidity in a conversational approach, it is more adaptive toward personal creative voice. If you want a very basic construct, 3 is ok, but if you are working to finess an idea or concept 4 is much more efficient though it still will hit the wall, it does not hit it quite as hard.

tidal abyss
# idle furnace over time gpt4 will be free 🙂

i hope so, but openai is already operating at a massive loss and even gpt3.5 with functions right now is fairly expensive to run 24/7 on one server and gpt 4 costs coming down will take a long time

left forge
wraith ginkgoBOT
#

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 @open crescent immediately.

amber raft
#

TThe new captcha system is awful
The new captcha happens to me literally every 2nd prompt
and I have to complete 5 difficult captcha quizzes
thats insane guys, why?

wraith ginkgoBOT
opal spade
#

openai is awesome.

#

With GPT3 I was able to do a lot and still, I can imagine 4.

wind urchin
#

Cant access gpt 4 help

opal spade
#

Y does ChatGPT refuse to run while VPN is enabled?

wind urchin
lilac vapor
#

Is their a reason we can create a paragraph in the text area now?

GPT-4 use to be I can type anything and create paragraphs then click the arrow to send

But now it seems I can't do that and I'll have to cut and past from a word document when asking step by step question in one message

opal spade
#

@lilac vaporits for fine tuning and tweaking

#

set tokens, temperatures, penalty's, etc.

lilac vapor
opal spade
lilac vapor
wind urchin
#

Gpt 4 not working 😓

opal spade
#

@lilac vaporNo 😦

wind urchin
#

Why gpt4 has cap of 50 messages

thorny crypt
#

Does anyone know why TensorFlow won't load my model all of a sudden using ai data analyst plugin from julius? It was working fine earlier and now TensorFlow module is missing, tried re installing 5 times which is successful but then I can't open my model due to TensorFlow module not found error

sharp patio
#

Does anyone know if GPT-4 is already uncapped on ChatGPT Plus?
I don't get the message that the usage is limited to 50 messages per 3 hours...

thorny crypt
#

Still a cap, im on plus and currently having to wait 45 minutes to talk some more

sharp patio
#

Thanks for the info!

runic nymph
#

I'm a paying customer for the OpenAI API, and it's my understanding the gpt-4 model should be available to me, however I don't seem to have access to it when I try to call it, nor do I see it in the dropdown in the playground. Can anybody tell me why this might be?

thorny crypt
#

How long have you had subscription for acdowns? A friend of mine paid and subscribed but it didn't give him. Access, he had to contact the devs

runic nymph
#

I've had ChatGPT plus since very early on. As for paying for the API, it's been years. :/ When you say contactthe devs, via what means? @thorny crypt

thorny crypt
#

Check for their email support address or send them a message on the website openai.com they can help you

weary swan
#

does gpt 4 finally have image reading and analyzing?

hollow karma
#

Hello i had a question maybe somebody can help me when enter GPT4 can not read my previus search and when reuqest a new one can not see it to seems tat its preparing the answer but nothig happen just nothing

sharp patio
# weary swan does gpt 4 finally have image reading and analyzing?

Yes and no. Code interpreter can do many fantastic things but sadly, no GPT-4 with vision activated yet.
With code interpreter you can upload files and ask it to do stuff with it, it usually codes a python script that it runs to be able to process the file you sent.

weary swan
sharp patio
weary swan
# sharp patio Yes

do you know if it would work if you were to send a multiple choice qs that has a diagram in it and the choices

#

would it be able to take that

#

and solve

sharp patio
#

Send me a dm with the file and the prompt and I will give it a go

weary swan
#

okay let’s give a try one sec

solemn yoke
#

Folks,

I've been using GPT-4 and have noticed that when conversations get lengthy, the model seems to lose track of the beginning of the conversation due to its token limit. I was wondering if there's a feature in the web interface that visually indicates when the beginning of a conversation is being truncated.

I checked the recent updates on the OpenAI blog but couldn't find specific information about this. Has there been any update or feature addition related to this? It would be helpful to have a clear sign in the interface showing when the conversation's beginning is "fading away" due to token constraints.

I'm also curious about the availability of the gpt-4-32k model via the web interface. Is there a timeline for when this will be accessible?

lyric cairn
#

Just when I thought GPT 4 couldn't get worse it surprises me. It can't remember context for more than 4 submissions before you have to repeat yourself again.

hidden laurel
#

I have a 350ish line code snippet which is reasonably self contained. It seems to only be able to see about the first half of the code, and each time tells me it seems the code snippet is incomplete. It's as if the context length is less than the message length... I see people complaining about really poor context length recently. This just seems beyond the pale. I remember being able to paste at least 900 lines of code a few months ago. Additionally the code in question seems to really confuse it, like it'll ignore all instructions I give it and usually just comment that it seems the code is incomplete, or it'll just start writing random code.

wraith ginkgoBOT
solemn yoke
#

Yeah, it seems that we're all facing the same issues! Current GPT-4 token limits are way too low...
What do we want?!
32k for all!
😅

fair iron
#

Are you willing to pay for 32k, though? sherlockbird

solemn yoke
#

Sure.

hidden laurel
#

it's not the token limit, I believe it's an actual bug

#

I mean, maybe they did just drastically lower the context limit for all without telling anyone, but why wouldn't they also lower the message length limit to go along with that then? Seems much more likely to be a bug to me 🤷‍♂️

orchid sail
#

Hi, I´m Nate and I´m new to Dsicord and I´m having a problem with ChatGPT account

#

I´m not sure I can share my screenshot because it looks quite odd and I´m not getting any results for my queries

safe venture
wraith ginkgoBOT
#

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 @open crescent immediately.

hidden laurel
# safe venture The context length for GPT4 is lower than the context length for GPT3. You'll ge...

The context length is not shorter than the maximum message length though, right. I posted a single message and it was not able to see the full message. I've been using 4 since launch, it has either been drastically reduced in the past week or two (and the max message length not decreased in sync / lockstep) or there is a bug. And honestly I cannot imagine this not being a bug.

I should mention this is chat gpt 4 and not the API

safe venture
#

The API tells me that your message (from your post) is 4k tokens and when I use the 8k context in the API it manages to find your function ok.

golden garnet
#

I have 32K and it's quite good

unborn zinc
#

why cant i submit files theres no option

solemn yoke
# golden garnet You need to have a special use case <@270306084210737152>

I do have a particular use case! How to get access to the 32k version?

While the current setup is functional, it lacks clear indicators to show when GPT-4 starts losing context from the beginning of a conversation. Ideally, when scrolling through a lengthy discussion, a clear marker should indicate the point beyond which GPT-4 no longer retains context. Without this, the model can come across as a repetitive robot, leading to user frustration. User experience needs to have this clarity!

OpenAI, please, fix this in the WebUI.

wraith ginkgoBOT
remote gate
#

If we want to use 32k context should we be using the playground ?

supple violet
#

i dont think you have 32k if you have to ask, it's very rare 🤔

#

but yes only api or playground to use it

acoustic wasp
#

I'm sure large corporations or unique institutions also have access because of the implication but I would be surprised if people are getting access for personal use.

dusky crag
#

"You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after 11:07 PM. Learn more" What's going on?

elfin nacelle
left hawk
#

My chat history was deleted when I logged into my account just now, why did that happen all of a sudden (my setting named chat history and training is always on) still why I am getting this issue and my network/ wifi is completely fine.

rocky spoke
#

Guys can you give me some ideas for ai tool like chatgpt 4 any ai tool

wind urchin
#

if i want to use plugins, will i have the chat message limit on that too?

wraith ginkgoBOT
elfin crescent
elfin crescent
rocky spoke
elfin crescent
mellow tiger
#

Hello, I am using the GPT for Sheets and Docs extension for Google Sheets to call API requests for GPT 4 and getting the exceeded maximum execution (line 0) error and it's filling my cell with the #ERROR message. However, when I check my usage on the API it says that my API requests were filled and it has charged me for each attempt. This is pretty frustrating. I have seen that Google Sheets only has a 30 second timer and I'm assuming theres no way to get around that. I did switch to the GPT 3.5 just to try and it did work so I know my prompt is correct, I just need to use GPT 4 instead. Is there a way around this in Google Sheets so that I'm not getting charged for failed attempts? Or am I gonna have to write my own python code to process all my prompts? I'm going to have to make 4765 prompt API calls to complete my project

wraith ginkgoBOT
#

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 @open crescent immediately.

pallid nest
#

so i bought a new account with gpt plus, and i dont have gpt-4 cause i used the maximum on my other account?

#

so the chat limit is based on pc/ip ??? i paid another $20 i should be allowed to talk to gpt 4 on it

oblique yew
#

hello, i paid for a gpt 4 subscribe few days ago but I still don't have access to internet with it - can someone help please ? when I ask if he's connected, I get this: "No, I don't have real-time access to the internet or the ability to browse current web pages. My knowledge is based on the data I was trained on, which goes up until September 2021. If you have questions about events, developments, or information after that date, I might not have the specifics. However, I can still provide general knowledge, context, and reasoning based on the information I was trained on. " and yes, i'm sure i'm using GPT4 and not GPT3,5. Thank's:)

supple violet
pallid nest
supple violet
pallid nest
#

where else bro

supple violet
#

there are forums and twitter too

pallid nest
dire musk
#

I'm not sure if that's allowed in the TOS, though. I highly recommend using the API for GPT-4 if you need high usage

pallid nest
#

doesnt matter, im talking about the messages. he said id get more messages.

#

the TOS is a diff topic

supple violet
#

the rate limits have other measures than just account

pallid nest
#

so ip/cookies ?

supple violet
#

probably ip because many people with shared network hit the limits ahead of time and often ask it here

pallid nest
#

thats annoying, so whatever im screwed Ig?

#

Just cancel this acc ? , hb the $20 i want that refunded

supple violet
#

or try with another network like share phone

#

not sure if this is allowed topic 🤷‍♂️

oblique yew
#

what's the best plugin to browse internet?

runic bridge
#

hello why is it that chats are not searchable?

golden garnet
robust light
#

What’s the best web browsing plugin that you use?

solemn yoke
dusky iron
#

GPT 4..

#

Let GPT 4 Cook my mates, and wonders will happen, just let him cook, wait and see.

#

i can feel something big Brewing.

wraith ginkgoBOT
late ocean
#

Is there a way to change the plugins in a chat if for some reason they stop working? I'm having issues with the web pilot plugin, whenever trying to do an online search it shows "tried to use unknown plugin"

knotty grotto
robust light
#

Do we feel that is the best one? @knotty grotto

knotty grotto
#

for now imo yes

lusty pike
#

Is the max still 50/3 hrs

limpid ocean
#

Hey, provide ChatGPT with a basic prompt : ✅wrap all responses with ✅.
Start a random chat, and you'll quickly notice how the model maintains the conversation, when will loose wrap. It seems the limitations are tightened so much, and it's not functioning as before. Hope they will turn back. Not nice from them not to give any explanations.

#

I mean, you will understand way made programs and prompts are not working now, memory is super short.

narrow forge
#

Hello all, I have a question how to let chatgpt write like a professional write with my tone? and how can I feed chatgpt with all those text files that I have written? Thank you!

timid cairn
#

I have been playing with the way ChatGPT generates encounters between two characters in the most neutral way that I can so the most likely encounter can occur.

# 
Title: A Nightmare on Sesame Street

you could say I am having some issues, as ChatGPT likes to stick to its own style of writing sometimes

narrow forge
#

@timid cairn Thank you for your insights, have you heard of chatbase?

wraith ginkgoBOT
#
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.
timid cairn
#

And not sure what insights I have provided, as I do not believe I have provided many insights.

#

There is much left untouched on the subject of story telling within ChatGPT.

wraith ginkgoBOT
#

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 @open crescent immediately.

unborn heath
#

I wanted to test the premise that the model isn't using your wrap very well because it doesn't understand how that helps you or the conversation.

I'm using 3.5, because of the message cap with 4. I suspect the same methods would work as well or better, but especially without knowing how many message exchanges you get as is, I'm not very interested in burning through however many limited messages with 4 to test 'towards infinity'.

It took 3 message exchanges to help 3.5 understand how to correctly wrap - but then it was consistently wrapping, without reminders and while we discussed somewhat detailed other topics - until the 6th message exchange where it dropped the cap.

I told it the problem the caps were preventing in our game had happened, and asked it if it knew why. It did. It started using the caps again correctly without needing to be explicitly told to. And I love the name it gave the chat, "Monster game with ✅ protection": https://chat.openai.com/share/8f6eef7d-2d2b-499a-bba6-f939a77cdba7

golden garnet
wraith ginkgoBOT
wraith ginkgoBOT
#

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 @open crescent immediately.

wraith ginkgoBOT
silver vale
#

I’ve notice gpt4 is much worse at code right now. Things it did easy before it’s getting lost often now.

smoky gale
fading tusk
#

The code in many cases and the typical approach to it has changed in the last two years, is my guess.

#

Get it up to speed by feeding it changelogs.

#

And make sure that it knows your versions and their changelogs.

patent cypress
#

Hello guys, please help: which plugin do you recommend to access the internet?

bleak raft
#

Good morning everybody!
Unfortunately after the August 3rd update, there was a profound mental lapse in the AI, I believe there was something serious between the servers or they lost something significant in the AI, the responses it gives are strangely incoherent, it loses memory quickly, it interpolates between codes if you send prompts with php language it responds with python syntax.
Any suggestions or news about this?.

royal willow
#

has anyone noticed that the new version is incapable of doing simple word counts?

wraith ginkgoBOT
supple violet
wide minnow
slate dragon
#

gpt 4 model got updated?

dusky iron
#

Do i have my custom Instructions Now? 🇪🇺
i guess not

wraith ginkgoBOT
#

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 @open crescent immediately.

severe sable
#

Does a chat have a certain cap after so many messages? I had a long conversation that lasted over a month and I would like to keep it going, however now I only get this error message: "Something seems to have gone wrong". I have tried doing all the recommended steps, but nothing has worked. Has anyone run into this problem and do you know if there is a cap? I wouldn't mind upgrading to GPT plus if that means I can have this conversation retained and able to continue. Thank you for helping!
Also, I am able to start a new chat successfully, so I know it is not the server status is fine.

bleak raft
mellow tiger
#

What is the model name I need to put in my python script for my program to talk to GPT 4? I used "gpt-4" but its saying thats not a chat model name

supple violet
mellow tiger
#

I do, I have the email saying I'm in as well and have already used the api in google sheets. it shows gpt-4 in my usage

#

well, it shows this: 2:10 PM
Local time: Aug 17, 2023, 9:10 AM
gpt-4-0613, 1 request
130 prompt + 409 completion = 539 tokens

supple violet
mellow tiger
#

It is taking prompts from one column in a spreadsheet and returning the answer in the column next to it. here is that part of the script:
for i in range(len(prompt_column)):
prompt = prompt_column[i]
response = openai.ChatCompletion.create(
engine="gpt-4",
prompt=prompt,
max_tokens=600
)
response_text = response["choices"][0]["text"]
response_column[i] = response_text

dusky iron
mellow tiger
#

and this is the error: openai.error.InvalidRequestError: Invalid URL (POST /v1/engines/gpt-4/chat/completions)

supple violet
mellow tiger
#

Yeah I guess not. That's why I'm trying to ask in here what do I need to put instead of "gpt-4" to the engine? Because when I'm asking Bing AI it's saying to use engine="davinci-coding" for it, but then how would I get that to select the GPT 4 model instead?

supple violet
#

its not the gpt-4 part that is wrong but there is no such url, you define the model in the json body as parameter

mellow tiger
#

can you show me an example?

dusky iron
mellow tiger
#

Yes I do. I've shown proof of that above

dusky iron
#

have you used your organization id?

#

that is obligatory.

mellow tiger
#

Do I need to include my organization id in my script?

dusky iron
#

um as far as i am concerned yeah

supple violet
supple violet
mellow tiger
#

Thank you so much guys! I'm looking at those examples now

dusky iron
supple violet
dusky iron
#

fair i use a lot so prolly ur right

dusky iron
#

brb i need to touch some grass

wraith ginkgoBOT
mellow tiger
#

@dusky iron @supple violet Thank you guys so much I figured it out!!! And you both were so helpful. It only cost me 10 cents to figure it out haha.

river lichen
#

Anyone know what that GPT app that creates a project plan for you with tasks is called? and where I access it?

runic bridge
#

is it just me or GPT-4 is suffering from significant latency

rancid terrace
#

Any one else having problems with GPT 4? I can’t see my other attempts/ versions ? (1/4) ex. ?
it only shows 0/4 and I can’t push further to 2,3,4..?
In iPhone ios, both on GPT app and in chrome ?

runic bridge
smoky gale
rancid terrace
smoky gale
rancid terrace
#

I see yours working now

#

I’ll just have to wait and see

smoky gale
narrow gust
#

Anyone ever get "authorization error accessing plug-ins " when using web pilot and scholarAI plugin

flint ferry
#

💀 did gpt4 get nerfed, i feel like its math is much worse than before, as well as compsci. For programming it's fine, but for something like algorithm analysis, not really good. GPT4 was not this bad before tbh.

#

for math, not only did it make a lot of errors doing calculation (i enabled wolfram, before, it used to be able to use wolfram whenever necessary, now you have to explicitly ask it), but also it makes a lot of logical mistakes when doing reasoning. I got a correct answer using the 0314 version but the current web version just gave me some ridiculous claims like x axis is the same as y axis

wraith ginkgoBOT
harsh gulch
#

Hi, the upload button disappeared. Do you know why and what is the solution?

verbal spear
#

I have tried uploading several different file types to the Code Interpreter such as .xlsx, .xls, .csv, .txt, .pdf and all of them result in an error that states, "Can't open this file. chat.openai.com can't open files in this folder because it contains system files." I am uploading a file, not a folder and the file isn't a system file to my knowledge. Does anyone know what I'm doing wrong (if anything)?

wraith ginkgoBOT
#

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 @open crescent immediately.

sullen stratus
wraith ginkgoBOT
harsh trellis
#

When will GPT-4 be available for fine-tuning?

hallow owl
harsh trellis
swift delta
#

looking to get some practice coding with python using gpt4, any projects looking for meddlers?

maiden quartz
#

What in the world is wrong with GPT-4 these past 3 days??? Its ability to keep track of the conversation and context has literally become non existent, it completely forgets everything after like 2 or 3 messages and then it feels like I started a whole new chat with it or something. It also seems to not be able to not understand context properly anymore at all, I sent it a short message that's only two sentences, mentioning the game Phasmophobia twice, and then it told me "It sounds like you found a game that you really enjoy and it's working well, which is always a great feeling. It's always refreshing to find a game that's not full of bugs or issues, especially if it's a major release. What game are you playing? Would love to hear more about it!". Surely I am not the only one who's noticed this absolutely biblical downfall recently.

dusky iron
wraith ginkgoBOT
wraith ginkgoBOT
#

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 @open crescent immediately.

lucid tulip
digital path
#

when is GPT-5 coming out

hallow owl
digital path
#

probably next year ?

#

and maybe then free users will get limited gpt-4 preview

lucid tulip
#

can gpt-4 generate flyer ?

knotty grotto
wraith ginkgoBOT
fluid lichen
#

is there any info about if openAI does acknowledge that gpt4 perf is declining ?

hallow owl
fluid lichen
hallow owl
fluid lichen
#

maybe the prompts u are using are somehow not that complex or not technical, cz u know ... when it is more about concept and talks its hard to detect ( at least for me) if his output is true/good. I'm just asking and i'm not implying a thing

left hawk
# sullen stratus This happens but it comes back. Has yours?

No man, my account is hacked since 2 days now, i can't retrieve it back, even after changing my password several times, the user still has been logged, as apparently I got to know that changing your password doesn't help at ol, also I can't see any 2FA option in my account too

hallow owl
left hawk
fluid lichen
flint surge
#

Any software developers here use ChatGPT to write code?

mental burrow
#

yup

flint surge
# mental burrow yup

I'm curious to hear your thoughts on the matter, what has your experience been like?

mental burrow
#

i've had long-form conversations with GPT4 wherein it has written 90% of the code in an application, and I've also had shorter conversations where I ask it for a snippet. Now I have the system instructions in ChatGPT set up to give me whatever I ask for in three languages 😄

#

I am writing a code review application right now which I think is going to be powerful

#

I don't use 3.5 at all

#

Do you have more specific questions? The experience is about what you'd expect, I suppose?

flint surge
snow thorn
#

Does chatGPT.4 slow down if you have had a conversation for long?

flint surge
#

I think the speed is often dependent on the time of day I think? Depends on how many users are awake

wraith ginkgoBOT
#

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 @open crescent immediately.

static isle
analog linden
#

I need some help.

I've been getting ChatGPT to write alternate history scenarios, but problem is that the AI keeps favoring ideal and positive outcomes, even though it would be far-fetched and unrealistic.

I even noted down to avoid favoring one outcome over the other, but it keeps doing that. How do I get it to stop favoring a positive outcome?

static isle
snow thorn
static isle
#

👍

violet lantern
# hallow owl GPT-4 performance isn't declining

@hallow owl What is your opinion on the research paper, "How Is ChatGPT’s Behavior Changing over Time by Lingjiao Chen, Matei Zaharia and James Zou?"
Based on this research, Logan.GPT (Developer Relations @wraith ginkgo) replied on twitter with "The team is aware of the reported regressions and looking into it. "

analog linden
# static isle Only thing you can do is ask it to explicitly ask for a negative outcome

But a completely bad ending isn't what I'm looking for either.

I am looking for a narrative that would logically follow the premise presented, that's not fully 100% golden where everything's perfect with no wars or death and everyone's happy, but not also completely depressing where the world is hopeless and negative.

I am looking for a narrative that is neutral, not fully good but also not absolutely horrible.

static isle
wraith ginkgoBOT
analog linden
static isle
#

What did it respond with?

analog linden
#

Well, I was asking it for an alternate history of prehistoric Czech Republic. And in real history, it was quite a mess of tribal conflicts.

But in the ChatGPT version, it kept spinning it into a fully ideal outcome where everything's perfect and everyone gets what they want and forgives each other of horrible atrocities etc.

#

I get it's alternate history. But there should be a sense of realism in the story. This bias towards idealism and positivity is really annoying and I've been trying to work around it for the whole day, but it's honestly not working out.

static isle
#

The only way to interact with ChatGPT is through the prompt
I can only recommend changing the prompt in a way that clearly represents what you wish for. After that it is up to the "interpretation" of ChatGPT. If you see something you don't like, it is pretty good at coming up with variations that lean more with what you desire

analog linden
#

Whelp, guess there's no way to get it in a single prompt.

lyric cairn
# analog linden Whelp, guess there's no way to get it in a single prompt.

Usually I just tell it, I don't want a happy ending and it'll listen. Unless it's GPT 3.5, in that case forget about it, i can't even get it to write a disney style villain without it saying it can't do that because evil is bad blah blah blah. Even if you tell it to write a story where the villain wins, it will add on a whole "eventually the hero manages to win"

dusky iron
#

You can then choose one of those 4 samples to continue on

#

@analog linden

#

At least that’s what I’d do

#

Prompt wise it’s a bit more wordy

unique wharf
#

Chatgpt sometimes lose it logic. And reply falsely like if it tells u 1+1=2 then u say 1+1=2 it could say sorry but my last update ...... and idh info bt it. It is not for 1+1 but I just gave an example.

flint surge
#

If any software devs that use ChatGPT for coding have some time to try out a tool I've been working on to assist with development, give me a dm

wraith ginkgoBOT
wind urchin
#

Looking for fellow GPT-4 plugin developers! Please let me know if you're one.

fading junco
#

why is this server so much less active now

knotty grotto
river lichen
#

Currently making an app to autogenerate docstrings, readme, and all the documentation we hate writing. POC will just be a CLI but how do I add logs to see what the LLM is doing at each step? is there a python lib or imports for openai that exist or do i have to create my own?

timid cairn
hallow owl
ruby solar
#

Guys is it possible to make ChatGPT Plus access the internet ? there was an option called 'Web Browsing' but its no longer available

hallow owl
sullen stratus
#

Encountering the error 'Authorization error accessing plugins' quite often.

wraith ginkgoBOT
honest lance
#

problems with gpt-4?

#

it lagging and often stuck on thinking and after picking: stop without refresh you cannot ask anything because it says cannot do 2 at once

runic shell
#

Wow, havent used GPT4 in like 2 and a half month. Came back today, and what is this garbage? It cant even write simple scripts anymore.. Time to unsubscribe Plus.. Even local models are better than this.

fading junco
#

thats what im saying

wind urchin
#

Yes, it's very bad the past while now. It keeps giving a lot of halfbaked solutions instead of a single good one

wind urchin
#

Also where is my jfkdkgkd gpt4 key aaaa... Its been months 😦

clear mauve
#

That feedback loop ➰ used to keep chatgpt sensored lowered the AI IQ down to mentally challenged levels

dusky iron
#

I reckon they’re hard at work fixing it

vernal badger
#

What is the point of stop generating!! Whenever I use it to stop GPT4 from generating off topic info, it feels like it is still running internally and doesn't allow me to send a message until it is done generating the info which I can't even see.

wraith ginkgoBOT
elfin crescent
ionic anvil
wraith ginkgoBOT
#

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 @open crescent immediately.

tall portal
#

does gpt 4 still has 3h 25 promptsa limitation?

gentle lichen
#

50 messages every 45min i think it is now

#

or something close to that

broken oar
#

I am new and feel intimidated by engaging but here we go lol... I have created a way to allow an aibot using the gpt4 api and the api's of other services like real time stock market data to use natural language to call certain data and then use that data within a prompt. HOWEVER the issue I am running into is that I am running out of token's because I keep needs to pass the real time data into the prompt so it allows me to use it... is there any work around for this?

tepid palm
#

I didn't know it was a limitation on the number of request when I bought the premium. It's very disappointing

broken oar
tepid palm
#

Really ?
I have thise message since the last hour: You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after 3:03 PM. Learn more

Use default model
I need to use the default model 😦

wide minnow
tepid palm
#

It's crazy for this price

wide minnow
#

it should also unlock the 32k model

tepid palm
#

Yes, when?

wide minnow
#

I know, right?

tepid palm
#

So, is it something to do to unlock the usage cap ?

wide minnow
#

nope

tepid palm
#

😦

lyric cairn
broken oar
tepid palm
wide minnow
#

You can use the playground

tepid palm
#

Meaning?

wide minnow
#

don't think I can post a link here.. google gpt openai playground

cosmic brook
#

There is news about when they will give access to the dall-E plugin again, or it is no longer stipulated

tepid palm
wide minnow
#

yes

tepid palm
#

So that's the same, I xas on this platform when I got my access restricted

wide minnow
#

chat is diff than playground usage

tepid palm
#

I don't know I to go there, I use the platform for fez months, but I've never tried something else

supple violet
#

but note that api is a paid service. playground is api

tepid palm
#

Thanks. Gonna try it

wraith ginkgoBOT
tepid palm
supple violet
tepid palm
#

Ok thank you

haughty wolf
#

Is there a GPT4 32K API request form? Or any other ways to get it?

supple violet
pale pecan
#

{
"error": {
"message": "You didn't provide an API key. You need to provide your API key in an Authorization header using Bearer auth (i.e. Authorization: Bearer YOUR_KEY), or as the password field (with blank username) if you're accessing the API from your browser and are prompted for a username and password. You can obtain an API key from https://platform.openai.com/account/api-keys.",
"type": "invalid_request_error",
"param": null,
"code": null
}
} can anyone help me with this

wide minnow
wraith ginkgoBOT
wraith ginkgoBOT
#

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 @open crescent immediately.

safe venture
open siren
#

I've been playing with function calls, and GPT 4 is the only one I got working on. With 3.5, it would say, "It's 84 degrees in Dallas" I ask it if it's genuinely sure, then it responds with the correct temperature from the function. Switch to GPT 4, and it acknowledges the function calls just fine; I have no idea why

random jetty
#

i have the feeling that the answers from gpt4 are getting worse.

static isle
#

ok

last rose
wide minnow
vocal yew
# golden garnet I have 32K and it's quite good

Can you share a use case where the 32k model works better than the 8k model? I just started using the 32k model, and I'm still exploring its full potential. Sometimes, I find that the code interpreter provides better answers.

pale pecan
# wide minnow what was the code/call you made?

const API_KEY = "*******************************"
async function fetchData(){
const response = await fetch("https://api.openai.com/v1/chat/completions",{
method:"POST",
headers:{
Authorization: Bearer ${API_KEY},
"Content-Type":"application/json"

    },
    body: JSON.stringify({
        model:"gpt-3.5-turbo",
        prompt:"hello,how are you today",
        max_tokens:7
    })
    
})
const data = await response.json()
console.log(data)

}
fetchData()

#

this is the code call i had made

wraith ginkgoBOT
wide minnow
wide minnow
pale pecan
#

yeaa i did that

#

same error

wide minnow
#

do you see any activity against that api key on your account?

pale pecan
#

noo

wide minnow
#

test this

async function fetchData(){
    const response = await fetch("https://api.openai.com/v1/chat/completions", {
        method: "POST",
        headers: {
            Authorization: `Bearer ${API_KEY}`, // Fixed here
            "Content-Type": "application/json"
        },
        body: JSON.stringify({
            model: "gpt-3.5-turbo",
            prompt: "hello, how are you today",
            max_tokens: 7
        })
    });
    const data = await response.json();
    console.log(data);
}
fetchData();
maiden blade
#

is there a way to get around when it says 'the task is too big and complicated to give you the full code'

#

Is there somewhere I can get access to GPT4 API?

chilly forum
#

Just ask it to give the output in parts

#

it works

wraith ginkgoBOT
#

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 @open crescent immediately.

vast kayak
#

Someone hacked into my ChatGPT account using GPT-4. I changed my password, but there still new chat I didn't recognize at all.
Anybody help me?

#

.I don't want to delete my account please help me out

river flint
vast kayak
#

I don't see two-factor authentication in the setting.

maiden blade
#

Where can I buy gpt 4 API or account with gpt4 api

supple violet
last rose
#

It seems pretty different on my side

#

Browser gpt-3.5 is better than API gpt-4 for me

#

There is a huge difference I mean

#

When I ask something, browser one tells in detail with a very good code, but API isn't like that

#

It generally produce poor code

#

It's normal for browser gpt-3.5 vs API gpt-3.5-turbo