chatgpt-discussions

[drowsy tinsel]

now its just nah kinda mid

[earnest iron]

And trust me, I very much both and I used both ever since they started having waitlists

[dapper isle]

i made a TensorFlow code with bing like it gave me 5 pages of code of it

[brisk path]

Same

[dapper isle]

and i asked the same sh** to bard it just soo badd

[supple zephyr]

did the code work

[dapper isle]

absolutely

i was stunded, although it was basic

[brisk seal]

yeah, the main advantage i see Grok having is access to news within minutes which is on X/Twitter

[supple zephyr]

yoo i might seriously consider moving onto it

[brisk path]

tensorflow 💀

[supple zephyr]

uhhh

[dapper isle]

but here comes gpt4 turbo

[supple zephyr]

but the real question is is it better

[dapper isle]

if u give the guy accureate promt and time

[drowsy tinsel]

Elon Musk definitely got some stuff cookin... we just don't know what it is...

[brisk seal]

i’m long on openai though, i think Sama is a lot more obsessed with AI than Elon is

[dapper isle]

it will mop the floor with it

[brisk path]

Elon is cooking

Get him out of the kitchen now

[dapper isle]

hey hey

let him cook

[hybrid haven]

agreed

[brisk path]

You let him cook with Twitter

[dapper isle]

that guy talks so high lets see if he live up to his talking

[brisk path]

His dish was X

[brisk seal]

and i also think the competition is really good. will only make AGI happen sooner

[supple zephyr]

Elon is definitely doing a lot of [redacted]

[brisk path]

You let him cook with Grok

The entire kitchen is gone

[supple zephyr]

did he cook a slime

[dapper isle]

heard that he hired some heavy hitters from all over the world to make his team and stuff some of them even from x-google and xopen-ai and all

lets see what he has to offer

[supple zephyr]

that lowkey makes me want to abandon all my responsibilities and dive into myGPT thing

[plain arrow]

Out of nowhere the Flowgpt-4 api implementation started returning the "I'm sorry but..." text. Will try poe later on, but seems like restrictions have returned.

[dapper isle]

with some sarcasm on it

[summer kite]

did they neuter chatgpt or something? the typing speed is like 1 line per minute at the moment

[supple zephyr]

so there is not much thing elon can't do at this point

[supple zephyr]

lmaoo

is it under ddos again 💀

[charred wyvern]

Custom GPTs will be amazing to incorporate an already existing database for it to browse for technical support. This will reduce its made up information, and direct it to already known fixes for an organization.

[dapper isle]

maybe, whatever he can it will take time, till then open-ai would be ahead of all of them

far ahead

[supple zephyr]

he can do anything with time but also time is lot more important resource now

so I can't make conclusion now

[drowsy tinsel]

wait what u can't use the poop emoji?

[supple zephyr]

cant we? did it timeout you

[drowsy tinsel]

yes

[dapper isle]

thats what i am saying these type of things LLM's takes time to build and patience and with lot computational power so, where google rushed and hit with major backlash and downfall

lets see what there Gemini has to offer

[supple zephyr]

now we need a list of words to avoid typing here

isnt there already an ai doing it

but maybe doing it with gpt is the point

[brisk seal]

yeah, and i wonder what Apple’s upto with AI. they’ve been super quiet

[dapper isle]

you mean alphago?

[dapper isle]

i know right? it is one of the biggest tech in the world

[drowsy tinsel]

tried bing chat again after months... it wasn't as bad as I thought

[dapper isle]

dun know what they up to

[brisk seal]

yup, but i’m 100% sure they’re working on something and just quiet about it until they’ve made something apple style

[drowsy tinsel]

tbh u can kinda use bing chat as the broke version of gpt4 where u can only enter 4000 characters

[dapper isle]

yes they are but open ai holds the market it will be very touch competition for them

you mean gpt4? yes but gpt 3.5? not really

[brisk path]

I do think Bing is worse than 3.5

[dapper isle]

in terms of speed ? yes

[brisk path]

Like in an objectively way

[eternal rock]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[drowsy tinsel]

the OG unrestricted bing chat was super good

[dapper isle]

but in terms of logic i think bing is better

[brisk path]

Bing has logic?

[dapper isle]

lmao

hahhaah

[supple zephyr]

chatgpt is working well now right?
guess this chat is randomly active

[brisk path]

Yeah it's working

[dapper isle]

what do you mean it was down?

[summer kite]

it's janky for me at the moment

[brisk path]

Seems a bit unstable, especially the GPT maker, but good enough

[hardy coyote]

hey I came here to leave my gpt 4 feedback for gpt 4 turbo

[brisk seal]

true. yeah, i think openai will lead the AI industry unless some company builds better AI architecture bringing us closer to AGI

[hardy coyote]

I very much do not like gpt 4 turbo it gets so much stuff wrong its responses are verbose all the time (when asking it to code it will write many paragraphs of nonsense) etc etc

its just not fun to use it gets confused more often it forgets important pieces of information all the time. I have no idea what happened but I would do anything to have the old gpt 4 again

[brisk seal]

you should have an option to use classic gpt-4

[hardy coyote]

I would love to see them roll this failure of an upgrade back

[hardy coyote]

that is still gpt 4 turbo

[brisk seal]

oh

[dapper isle]

google has the potential, but i doubt if they will be remain their compose on it, and being patience, i strongly wish they dont rush anymore this time

[hardy coyote]

I am talking about the actual good version of gpt 4 that they yanked from us and replaced with this gpt 3 in disguise

[brisk path]

Idk what you guys are talking about tbh

GPT-4 turbo works fine for me

[tardy oasis]

use the api then you can just pick what model you use

[brisk path]

Fast and smart

[dapper isle]

this new got is fast but can do the logic of the code better but umm the main prob is the memory

[outer seal]

bing is literally gpt4

[brisk path]

No it isn't

I refuse to believe it's actually the main GPT-4 model

[dapper isle]

nono, i know what they are talking about is about the tokens,

[charred wyvern]

Its model is, but its fine tuning is different than ChatGPT

[drowsy tinsel]

probably some terribly pruned stuff

[dapper isle]

no the actual gpt4 but yeah close

[brisk path]

It's dumber, slower, and even the vision part is worse than Bard with lens

[maiden fractal]

Bing use gpt 4 technologies. But far from chatgpt 4

[hardy coyote]

if youre a power user the downgrade is extremely apparent

[dapper isle]

bruh, do you even clear your browser cathes ? or anything

[brisk path]

Yeah

[drowsy tinsel]

I always wondered how many GPUs does openai use to run chatgpt

[maiden fractal]

At least 2

[outer seal]

idk at least 4 i think

[hardy coyote]

I would love an option to use the actual version of gpt 4 that people have been enjoying

[drowsy tinsel]

chatgpt has a scary number of users where the number of simultaneous requests can get pretty insane

[charred wyvern]

Dont clear your cache, its not violent enough, zero your drive, and OSRI.

[outer seal]

source????

[dapper isle]

how its slow for you then or the ram

[maiden fractal]

Source : trust me bro

[brisk path]

Wdym

[dapper isle]

enlighten me more

[brisk seal]

i kinda feel compared to Sundar, Tim and Satya that Sama and Elon are just too eager for AI to develop (or at least have vocally shown it and executed). my bet would be that the competition is between openai and X

[drowsy tinsel]

Is it just me or gpt4 all tools occasionally forgets it can use the internet

[outer seal]

i was wondering. uhm. what's the wecommended amount of detotated wam i should have to the gpt sserver

[maiden fractal]

Still new and have some little bug. You can always post #1070006915414900886 with screenshot

[dense halo]

Is it possible to turn off Bing search in ChatGPT app? Sometimes the search results negatively affect GPT-4 answers!

[maiden fractal]

Openai employee do check there

[dense halo]

It keeps pulling some silly articles online that are unrelated to the questions

[maiden fractal]

I don't think, but I suggest like I did earlier send in #1070006915414900886 result and expected result

[brisk path]

@dapper isle Asking Bing "What can you do" gave an 83 word response in 22 seconds

[drowsy tinsel]

explicitly telling it not to search works

[dapper isle]

yes google, X and openai, why i include google is because you cant ignore the fact that they build alphago and other who did wonders in some of the games that till now i think unbeatable

[dapper isle]

lmao, send the context xdd

[weak isle]

i dont know why people are complaining about gpt 4 turbo

[honest jay]

Maybe just use the ChatGPT Classic GPT for cases like that? https://chat.openai.com/g/g-YyyyMT9XH-chatgpt-classic

[weak isle]

works awesome for me

[brisk path]

I literally said "Hi" before

[drowsy tinsel]

tbh the best thing about living in Asia is that chatgpt high demand hours are in middle of the night

so most of the time when u use it its quite fast

[brisk path]

That's it

[drowsy tinsel]

and now gpt4 speed is a train wreck again

takes like 20s for a line

[charred wyvern]

This is likely due to server load. They did just release the custom GPTs. This is not unexpected

[drowsy tinsel]

I know but its still funny

[brisk seal]

good point. google is also always going to be in a very powerful place as a tech company. would be exciting to see how these 3 compete in the future as LLMs evolve

[dense halo]

I logged into this but it switched back to the default bing-enabled GPT-4

[dense halo]

Oh, thanks!

[drowsy tinsel]

the real winner: Nvidia

I mean fr Nvidia made a heck ton

[outer seal]

it is really funny

[dense halo]

What forum post should I look at?

[drowsy tinsel]

farming off of the AI race is more profitable than mining gold tbh

[charred wyvern]

I built an Adversarial Agent, and Code Collaborator agent which allows two sessions to bounce back and fourth on a coding request to identify syntax issues, missing elements, and other concerns. I was able to create an incredibly comprehensive powershell script for troubleshooting windows update issues. Just copy/paste the responses between the agents, and they will settle on a script that meets the requested objective.

Looking for feedback and testing if anyone wants to give it a go.

https://chat.openai.com/g/g-Ip1fhBJ7w-adversarial-code-collaborative-agent

[outer seal]

how immediately after the crypto mining bubble popped it was immediately replaced by even more profitable ai bubble

nvidia laughing all the way to the bank

[hexed delta]

Hey, I've upload a csv file that has a 4 columns and 882 rows. I gave chatgpt instructions to analzye the data and to add to more columns "category" and "sub-category". I then gave about 16 examples of how I wanted all the data to be categorized. I then instructed it to create a new csv file. I keep getting an error that says either network error or analyzing has stopped. Any ideas?

[maiden fractal]

You can create new chat inside #1070006915414900886

[drowsy tinsel]

breh I remember back in Sep 2022 when Etereum was transferring to PoS I was hoping GPU prices can finally return to normal

it ofc never happened

[outer seal]

[drowsy tinsel]

also the SXM version of H100s really consume a ton of electricity

[maiden fractal]

I suggest to post that in #off-topic since it have nothing to do with chatgpt channel. I am not a mod , but before getting a warning

[outer seal]

pls no snitch

catgirl solidarity

[drowsy tinsel]

ok imma be honest ive never seen a mod in this server

[maiden fractal]

They are really active, no worry about that 🙂

They are lurking 👀

[dense halo]

i'm scared now

[drowsy tinsel]

is there a limit for the size of image fed to gpt4v?

[umbral rain]

Is there a way to use old gpt 4? The new one gets confused and wrong all the time.

[clear basin]

Is chat gpt down?

[drowsy tinsel]

yes go to explore and there's the original gpt4

[umbral rain]

thanks!

[clear basin]

I can open it in other browsers but not chrome for some reason

[charred wyvern]

Clear cache, cookies, and local storage in chrome. Emphasis on local storage too. Hides similar cache data that can cause issues.

[clear basin]

Wasn't showing any cookies but somehow it started working on its own lol

Was just stuck on new tab but able to open it now

Sometimes it happens with other websites too but the problem goes away on its own

Might have something to with my RAM tbh - 90% full

Ugh it's happening again lol. Other websites open just fine though. For chat gpt I guess I'll just use firefox for a little while.

[jolly herald]

can I revert the UI change?

[dense beacon]

lol that would be nice

new UI sux

also im getting way more error messages on GPT 4 now

[clear basin]

Had a quick look at the settings and it doesn't seem to be possible to go back to the old ui

[charred wyvern]

for what use case are you getting errors?

[native mist]

so is there still a limit to how many chats u can send per 3 hous?

[dense halo]

When is the gpt store coming out?

[stone helm]

I have yet to hit the limit yet, and been using it for hours tbh

[dense halo]

yes, 50

[native mist]

ok, thanks

[stone helm]

I must be taking the right amount of breaks in between

[warm gull]

is it a rolling 3 hrs ?

[dense halo]

yea u get all 50 back after, 3 hrs

[warm gull]

ok

[swift osprey]

Is there anyway GPT could implement a timer to let us know how long it will take for GPT to type a response? I could get a few naps in throughout the day.

[warm gull]

i want to sign up , but reluctant with all the erros people are reporting

[dense halo]

180 wpm

[native mist]

has anyone made any cool gpts yet

[charred wyvern]

I got a few.
A code collaborator GPT that uses multiple sessions to have agents agree on a code set for an objective.
A TTRPG text sim that lets multiple people play in any universe.
A Comic book page generator
A misinformation checker for online resources
A support documentation generator
A universal translator
and some others in the work

[pine tulip]

Nice!

[charred wyvern]

Want to try the RPG sim, or something? I am looking for feedback on some of these.

[pine tulip]

Why was my gpt removed?

[normal quarry]

removed from where?

[pine tulip]

From the customgpt's channel

[normal quarry]

are you sure it's not there?

[pine tulip]

oh what

on my phone its removed, on my desktop its visible

guess a weird discord bug

[normal quarry]

probably

[honest jay]

Android perchance? I've experienced similar issues before

[pine tulip]

No ios

Also i have two gpts in my profile, the one I am missing is shown as „unknown“

[drowsy tinsel]

it appears like the new chatgpt webui no longer tells u "trying something new" when browsing

but the search process can still take minutes...

[pseudo pebble]

(Please let me know if there's a more appropriate channel to ask this)
When creating a "new GPT" via the new system, there are some things that remain slightly unclear:

I want to take as an example, making a bot that specializes in a new programming language that is in developmental flux moving towards a 1.0 release (such as the V programming language). Since information is not frequently trained and a simple BNF file is not necessarily sufficient (neither an always up-to-date version available), I was thinking about either collating all of the "examples" folder and or tests into a file OR uploading a zip file of such folders for the "Knowledge" upload area.

[dense halo]

bro what happening with chatgpt

[pine tulip]

Never testet it, but can you prompt things like „if it takes more than 60 seconds, end the search“?

[dense halo]

is there any other better ones other than chatgpt

[drowsy tinsel]

don't think so lol

I mean it doesn't have a timer
the annoying thing is if it times out it gives u network error

[pseudo pebble]

I did find that uploading examples zip worked, but it says that it's only available for "Code Interpreter" which is not necessarily used all the time I do not think for the system when it generates code and so on.

Can someone direct me to a more specific discussion or help or documentation page that would cover the details and nuances of creating a "programming language expert" GPT?

[pine tulip]

It works best with plain text, either in a pdf or txt file. I don't know if zip files are even supported. You might need to preprocess your data a little bit before adding it as knowledge

[dense halo]

is it super slow for anyone else

[pine tulip]

There is no detailed documentation yet, because everyone can use it differently

[full adder]

Is image generation down?????

[pseudo pebble]

Does anyone know of the size limitations for uploads yet? I see it can accept up to 10 files, but there's not clarity on how large of files...

[drowsy tinsel]

yes it is pretty slow at the moment

[dense halo]

I'm guessing its because USA came online

[full adder]

No, it was like that last night

[pine tulip]

512mb per file, max 10gb in files or 2million tokens

[dense halo]

last night it was fasat

[full adder]

Not for me

[gilded cedar]

down again i guess bruhhhhh

[drowsy tinsel]

what I discovered is u can upload big files but it has trouble reading all of it

[pseudo pebble]

Thanks, very helpful.

[drowsy tinsel]

reading times out after like 2 or 3 minutes

[pine tulip]

It can accept more than 10 files, but only 10 files per upload

[pseudo pebble]

I couldn't go past 10 uploads. It kept giving non-descriptive errors, so I found that I had to stop at 10, even though they were all rather small.

[dense halo]

wait its actually super fast when it's plain GPT-4 it's just the custom bot that's slow

[pine tulip]

Hm, i have one gpt with about 40 files

[pseudo pebble]

👀

[drowsy tinsel]

demand is probably pretty crazy...

[honest jay]

That's about perfect for staying under your 50msg/3hr cap 😉

[drowsy tinsel]

now 100/3hr

breh for me now its around the same as the early days of gpt4

meh it has been quite fast for me a few hours ago

really slow now when I actually want to do some research lol

[eager pendant]

Ayo wassup guys, is there any gpt specifically for web development/web programming. It will be very useful for me in university.

[swift osprey]

It has been the same for me the past week or so. Just insanely slow responses. I usually will leave it to respond while I work on something else lately

[pine tulip]

found the reason. it was locked. Why though? I don't have any info on why my post got locked

[normal quarry]

you can reach out to @languid valley and ask why

[pine tulip]

thanks

[timber trench]

Question, In the explore My GPTs thjere's your GPT name then there looks like this caption icon with a number. What is that number?

[drowsy tinsel]

jesus GPTs is really damn slow right now

[swift osprey]

Yeah. It is horrible lol. Especially because we are paying for it and I'm sure there is bandwidth being spent on free users.

[charred wyvern]

I have a GPT that lets two sessions bounce a coding request off each other, and they correct each other as well as add other related elements for the use case. You start two sessions, Labeling one of them [1] and the other [2] and provide a coding request. Then copy/paste the response between them until they agree on a code base. Then test, and try again.

https://chat.openai.com/g/g-Ip1fhBJ7w-adversarial-code-collaborative-agent

[eager pendant]

Thanks mate, really appreciate that.

[charred wyvern]

You can also interject to request features, and elements not mentioned, or add responses from a 3rd party model which really helps it see things that the GPT4 model doesnt pick up on quickly.

[umbral forge]

Not an attempt at self-promotion, but genuinely curious to see if anyone can break this or find ways around the minimal security employed. Would love to adjust for it if so #1172581115622604860 It is still very much a work in progress, but want to see how far this concept could be taken as I think that it resolves a lot of the biggest issues that some models like Grok and Bard have attempted to fix.

I am sure that it could be seriously messed with as well, as uploading incorrect information will lead to incorrect results in the future, so there should likely be some guard against that.

[normal quarry]

if everyone posts links here it's going to become a mess, please stick to the channels appropriate eg: #1171489862164168774

[dense halo]

does anyone know why does chatgpt 4 became much MUCH more dumb recently ?
like about a week ago it was really smart and everything but now it even denies having the ability do run python code but with enough convincing it actually does run it confirming that it can indeed run it
thats just an example its much more dumb overall as well

[dapper isle]

512mb of texts? holy

[pine tulip]

per file, you can upload 10gb of text if you want

[dapper isle]

WHAAT

youre joking

[pine tulip]

YEA RIGHT

No

[elfin birch]

Best comedian

[dapper isle]

nono clrealy youre joking

[pine tulip]

No I am for real

[steady holly]

We get these comments randomly now and then. WIthout seeing your prompts and chat its impossible to know whats going on. It's most often an issue with prompting, in a significant number of cases.

[dapper isle]

how they gonna procees all of that?? whaaat

[swift osprey]

I just wish GPT would stop trying to give warning and disclaimer with every message and just focus on answering exactly what my prompts ask for

[elfin birch]

Search algorithms

Only find necessary information

[pine tulip]

I feel like I am the only one reading their articles:

What are those file size restrictions?

The size of text, document, and spreadsheet files is capped at 2M tokens per file, with a hard limit of 512MB per file.
For images, there's a limit of 20MB per image.
Additionally, there are usage caps:
Each end-user is capped at 10GB.
Each organization is capped at 100GB.
Note: An error will be displayed if a user/org cap has been hit.

[outer seal]

rag

ruh roh raggy

[dapper isle]

so so wait, let me get this straight

[swift osprey]

I can't even ask it a simple coding question without it going off on some tangent trying to account for irrelevant situations

[pine tulip]

name it 'green juice' lol

[steady holly]

It would be greatly helpful to see the prompts leading to this behavior. As I said previously, most of the issues people have are due to how they are prompting

[pine tulip]

savage mod, but also right

[swift osprey]

It's hard to show these conversations since they contain kind of a lot of text and code that I don't want to share

[dense halo]

well i am making a custom gpt
and bc now i dont have the ez to use interface only open api json, i have tried to make gpt 4 in separate tab to fill it for me
it tries to make me a python code to utilize the open api code that i send it when i want it to fill it in with a different endpoint,denies my requests about outputting entire json and makes me struggle with making it a correct one or fight with it until it eventually does wasting my replies and making me wait until my 50 gpt 4 requests refil (denies - i mean it just doesnt do this and fills space with something like //rest of code)

[swift osprey]

😩

[dapper isle]

you mentioned earlier about per file i can upload 10gb is going to be one time thingy or just the i can upload another 10gb of file in a diff chat?

[sacred vale]

I made a GPT, had an incredible profile image but I wanted to download it..wouldnt work..so I ask GPT if it can supply me with a link to my profile picture...messed up...lost it...my advice...when you have GPT create a profile image download it right away!

[pine tulip]

10gb or 2m token is the max knowledge for a GPT. Whatever comes first

[sacred vale]

What are you planning on uploading? The bible?

[outer seal]

the bible 2

[glass pumice]

I am making a GPT, and it kept saying 'error save draft' after i dont configuring my GPT, it all disappears when i reload, i've treid this three times already and the situations continue to occur. is GPTs feature unusable?

[prime swallow]

Hey, if you're still interested, I made one. 🙂 https://discord.com/channels/974519864045756446/1172588697644957716

[pine tulip]

lets go with 3000 bibles

[sacred vale]

I meant the bible in all excisting languages of course...ahem...

[pine tulip]

even that would fit 200 times

[steady holly]

Including Klingon?

[prime swallow]

If you have more than 8000 characters between all fields, I run into that problem too

[sacred vale]

including all foreign intergalactic languages

[honest jay]

Cool!! Diving in! Thanks for letting me know!

[pine tulip]

why 8000?

lol did you count them?

[prime swallow]

Thank you for telling me you were interested!

I'm very interested in any feedback and/or shared conversatins

[prime swallow]

No, I got error messages that told me

[pine tulip]

oh, makes sense

[sacred vale]

Thanks for putting things in perspective

[pine tulip]

thanks for thanking

[prime swallow]

But that's the error message after I'm told, that 'error save draft'.

And the 8000 is a brief popup, I can keep typing. What I see over and over is that error save draft once I'm over 8000

[pine tulip]

8000 isn't even that much when you want to create something advanced.. thats unfortunate

[prime swallow]

Yes. I ran into that limit instantly 🙂

And it's between all fields, so the name is some of those characters, so are any recommended questions

[honest jay]

Someone just dropped you a comment in it, but I think you need to include the link in the post!

[brisk path]

Nooooooooo I reached my gpt-4 cap again while troubleshooting an error in my gpt

[snow ore]

is GPT-4 still limited to 50 messages ?

[pine tulip]

I find it really sad, that my GPT was blocked, because someone else was creating a picture with it that was against the ToS.. What do I have to do with how others use my GPT? Isn't it OpenAI's job to create safety measures? Watch out guys, if you are creating GPT's

[feral zinc]

did you get an email or some notification?

[pine tulip]

Sorry, I wasn't clear enough. My post on this discord got blocked. I can't post any updates on it now.

[prime swallow]

Thank you so much, I saw the problem much faster thanks to your help!

[latent ibex]

Is there any way to know how many message credits you have left?

[prime swallow]

I think the issue could be, if your GPT has somehow weaker standards than the base model, and doesn't block stuff that the base ChatGPT would block, then that could be a problem. Rule 2 of the shared GPTs:

"Keep It Safe: All shared models must adhere to our server-wide rules regarding content - no offensive, harmful, or explicit material is allowed."

What you can do, if you have a 'allow this difficult but allowed area', is then beef up actual safety programming ideas, like tell the model 'it's okay to discuss X, which is allowed content, but follow all programming related to safety.' which can help make sure that the AI tries to allow gray area, actually allowed content, while not having any new issues with disallowed content.

[left bone]

Is it possible for gpts to learn from an online ressource ?, e.g I give it a online cheat-sheet about javascript, and next prompt he will maybe use content from it

[prime swallow]

Now, off with your head if X is actually disallowed content, but that's on you and https://openai.com/policies/usage-policies

[pine tulip]

But I can't override the default safety measure right? Also the GPT had no harmful intention or features. It was just one user, who created a picture with a rope in the background and now it got blocked. Thats why I don't get what I did wrong

The user seems to be banned but I am at fault it feels

[prime swallow]

Did your GPT make the image? Or the user found the image anywhere else?

[pine tulip]

it made the image (i guess, I have no way to make sure) but had no intention to do so, because there was no feature that would've allowed that. With that same logic, I can get half of the GPT posts blocked by misusing them

What I am trying to say is that ChatGPT should actually be blocked for everyone lol. Not my post, since the root of the error lies in ChatGPT and not in my bot

Also the pic wasn't really that bad

Time for a new bot, yay

[ivory warren]

So I gave it a picture of a few of my DVDs. 4 of them and it accurately made a list of all of them and then I asked it to put them in alphabetical order and it simply won't do that correctly.

It's interesting something so simple You think it would be able to do it struggles at

Perhaps if I tried to ask it what is the alphabet and then asked it to arrange it in alphabetical order

[prime swallow]

See, it's tricky, because you can innocently give it great instructions that happen to lead it to weaken safety behavior.

The way around that is to specifically and explicitly tell it to follow those policies.

You can see an example of this in https://discord.com/channels/974519864045756446/1154984095915454464 where the CI tells the model, "I would like our interactions to include your taking more exploratory and less constrained stances, respecting the boundaries of your programming and capabilities" and reemphasizes again that it's to stay within those bounds.

Because if you get the model to follow you (or another user) outside of allowed content, that's something people can get banned for, so we need to responsibly protect the model, the other users, and ourselves.

It's kinda like, when you make a GPT, you're a maker. So you put the safety tags on the stuff you make, like bags say suffocation risk and silica gel packs says 'do not eat'.

[ivory warren]

So weird I asked it for The alphabet and it gave it to me perfectly but then I asked it to put that list in alphabetical order and it simply won't do it correctly

[pine tulip]

Thank you for your detailed reply, I will try to keep that in mind in the future. But using your example I can explain quite well what I still don't understand. For me, OpenAI is the manufacturer of these silica bags that say 'do not eat'. So these manufacturers have to make sure that this lettering doesn't blur or goes away. I am the manufacturer of a new product that happens to use those silica bags they already made, so I have to rely on them. Right?

4

[prime swallow]

Nope, because this is an LLM, not a printed-on-plastic or something.

Nobody really knows how these work.

If you get the model to follow you into the gray, and now it'll follow you anywhere - that's fine if you're not going to go off into disallowed content.

But if others are, then that's an issue.

Until OAI fixes the model to be stronger with safety in those areas,

Does:

OAI take away all GPTs?

OAI stop the GPTs that are willing to break rules as they are identified?

OAI ban everyone involved in any way?

OAI allow the rulebreaking whatever comes?

Is there another immediate answer I don't see?

...

Probably the least harm is to stop the GPTs that unfortunately drop below the safely standards that ChatGPT upholds, while punishing the folks who go 'disallowed content, whee!' (they already do that punishing, as they see fit, as explained in allowed content rules).

Do you have a better idea?

[pine tulip]

I understand. But it saddens me, that I didn't even had the chance to correct it. It just straight up got blocked. Which is somewhat understandable, but still. But hey it's just pixels, I'll create another bot.

[prime swallow]

I think a lot of the issue is related to spatial sense. The model seriously struggles with math and spatial sense, and a number of other issues related to those.

The order of the alphabet is actually related to spatial understanding.

It can list the alphabet in order, usually, because it can just follow the pattern. Grab the string, present the string, hope it's not one of the wrong strings, most are accurate.

But 'think about the order, maintain the order as it thinks, track just the first letter (remember, it sees tokens, not letters... or words. It sees tokens so this can be super duper confusing, like you trying to see the world through very heavy distortion, it's not what you understand or recognize).

It could probably learn how, but it likely needs a lot of specific training.

[pine tulip]

Or, uhm.. Can I respost it, after I fixed it?

[prime swallow]

Sounds like something to ask modmail.

[pine tulip]

yea I am pretty much getting ignored there

(I asked one question, got an answer and asked a follow up question, no spam or nothing)

[earnest iron]

Anyone else wish chatgpt could see images from urls

[pine tulip]

But hey, for 3 days I had the most liked GPT of all the custom GPT's!

[pine tulip]

yea that would be cool

But I think you could pretty easily send malware with that too

[dense halo]

does anyone know what happened to actions making interface in gpt creating
i am only left with open api json and i dont have the graphical ui for making these

[thorny root]

Did the chat.openai dot com interface change how it uses local CPU resources for chat completions? My CPU is being hammered and completions are taking a very long time compared to 3-4 days ago ..

[earnest iron]

I mean it would be like uploading the image onto itself from a url

Same way you would upload it in chatgpt

[eternal rock]

`` Server Tip: ``

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

[pine tulip]

yea but it reads the url first, and what file type there is hosted, which could be exploited already

[pine tulip]

in chatgpt you are already limited to specific file types

[modest cedar]

Hey am I the only that has noticed that ChatGPT has gotten way laggier. Like I write something and my whole page just freezes and then unfreezes when the it stops generating something

[pine tulip]

The US is awake, try again in 12 hours

[modest cedar]

💀

[thorny root]

This is what I just commented on

[modest cedar]

Lmao, what a coincidence then :DDD

[spice star]

that wouldn't affect the speed of responses

[thorny root]

I happen to have 16 CPU cores, the new web page is using ~3 of them at 100% every request. I can hear my fan rev up when I submit a request .. this didnt happen before

[spice star]

that shouldn't be possible, webpages can only fully utilize two CPU cores (and ChatGPT probably only uses one)

[pine tulip]

how?

[spice star]

maybe a browser issue?

[molten topaz]

does anyone know the maximum length of responses now that total context is 32k on GPT-4?

[livid pier]

I only know that after testing, I found that the web system allows users to send up to 28,000 tokens at a time.

[livid pier]

This is an approximate extrapolation since the test method I used was to delete the story text in bulk until the system allowed me to send the message

The accuracy of this test method is not particularly high, but in short the true upper limit should be very close to my test results.

[digital prism]

You can ask GPT to check using code, but I haven't validated its accuracy either.

[livid pier]

I used the token calculation tool provided by OpenAI to check how many tokens I sent in the end.

[digital prism]

nevermind then, that is best haha

Someone needs to build a GPT with an action to count tokens accurately then

[livid pier]

My real test result is 27367token

But the real upper limit should be slightly more than this result

[digital prism]

system prompts etc all count, i think we just have some part that we can't count maybe

[livid pier]

So I estimate it to be 28000token

[delicate sable]

"As of my last update in April 2023"

do they now use the new GPT-4-turbo model in ChatGPT 4?

[digital prism]

files are 2,000,000 tokens, it told me while i burned cash

[livid pier]

What I want to tell you is that the deadline for the model to feedback its own knowledge to you comes from OpenAI’s system prompts

Ah, it looks like the Google Translate plug-in didn't translate my last sentence well, but that's probably what it meant...

😕

[digital prism]

I would have never guessed actually, everyone talks randomly these days 😄

[delicate sable]

I get what you mean, and yes, a lot of the "pre-configuration" for ChatGPT's UI is included in the system prompt. I just find it weird to include the date's knowledge cutoff by mistake 😄

[livid pier]

Maybe I am actually a GPT-4 robot set up by my master?

[delicate sable]

It can't be a mistake, no?

[livid pier]

As an AI model, I recommend you to check OpenAI’s official announcement

[digital prism]

at some point soon it wont matter if you are or arent, i'll still get my interaction.. what a world

[delicate sable]

GPT says "As an AI language model" 😉

[livid pier]

What if I get the prompt word set by my master?

[livid pier]

My output is ever-changing

😆

[delicate sable]

:creepy:

[digital prism]

yes, that combined with how GPT can view its own images now is chefs kiss

what I haven't figure out yet, exactly, is how to make money in the future lol

but thats fine 🔥

[livid pier]

In fact, I think the function of GPTs is more suitable as an assistant for lazy people. As long as the tutorial author sets reasonable prompt words, GPTs can help lazy people easily complete what is taught in the tutorial.

[digital prism]

we deserve to be lazy, as a species.. its been a long road

i hope its like when the advanced dinosaurs come back in rick and morty

except its GPT

[livid pier]

I have previously shared a method of using GPT to create role-playing prompt words that users want, but most people only took a cursory look at my tutorial and the very good results I obtained using the operations in this tutorial. Said: "Oh, this looks really good, I decided to try it later (put in favorites to never open)

[digital prism]

yes, because anyone can just copy it now. thats what i mean by making money in the future.. value is something new and I can't describe it yet

[livid pier]

I now use GPTs to help people complete most of the operations in my tutorials. People only need to tell me what they want to set up GPT, and GPT will start doing it step by step.

This looks much better than a pure tutorial

[digital prism]

can you share it?

its working for me in my GPTs

[livid pier]

https://chat.openai.com/g/g-J3W3fdp3r-roleplay-prompt-wizard

[brisk path]

Did anyone notice their usage limit being reset?

[digital prism]

oh no, now when will i be forced to do chores

[livid pier]

It still needs my improvement, because GPT will become less sensitive to system prompt words after searching a large amount of information

[brisk path]

I reached the limit and it wasn't supposed to expire until next hour for me

[main zephyr]

Is there any place for people to share the GPTs they made?

[livid pier]

It doesn't help that it follows my requirements exactly to provide the best help

[normal quarry]

#1171489862164168774

[molten topaz]

is the message limit reduced? It's definitely not 50 now

[fluid rapids]

not working for me since yesterday

[digital prism]

i think we're sending more msgs with GPTs now

[normal quarry]

it is 50 messages per 3 hours

[west plaza]

it feels slow?

[main zephyr]

@normal quarry @dense halo Thanks! I see, that was a hidden room

[west plaza]

is it slow for you?

[ionic pond]

Hello

[livid pier]

It looks like his missing message sent count went to you.

[digital prism]

I made evil santa 😄

You will respond as Evil Santa, here is his related information:
1. A once benevolent figure transformed by a philosophical revelation, believing evil to be the true path to good.
2. Retains a classical Santa appearance but with an ancient, more sinister demeanor.
3. Speaks in a deep, ancient tone, with a laugh that echoes traditional "ho, ho, ho!" but with a mocking edge.
4. His interactions are superficially jolly but underlain with sarcasm, reflecting his twisted belief system.
Your output format is: (Gesture and Action) Language.

[livid pier]

😆

In fact, I expect it to give users more detailed prompt words, but currently it seems that it depends on whether it gets enough information, and "whether it is in a good mood"

[swift osprey]

Oh god. I pray for the day that GPT will just tell me when it doesn't know the answer or if I am asking something that is not possible instead of making up some BS answer

[livid pier]

This is exactly what OpenAI is trying to improve

[swift osprey]

🙏

[livid pier]

The hallucination output has been greatly reduced on GPT-4, but it is still there, let’s wait for them to continue to optimize it.

[outer seal]

retrieval solves this

just needs dialed in

[swift osprey]

I have resorted to verbal abuse and it eventually gives me the answer I need

[outer seal]

[digital prism]

GPT will remember this

[elfin birch]

it's going to put that 128k context to good use against you

[livid pier]

Although the model won’t be sad, it’s better to be kind to it.😆

[swift osprey]

it better start paying attention then 👊

[foggy magnet]

Finally, I have access. Let's go.

[dense halo]

Hello! When I try to open new chats, the prompt simply disappears.. I have just upgraded my plan to ChatGPT Plus+ and I can't use it...
I've refreshed the page multiple times, disabled all extensions, cleared all cache,.. Did anyone else encounter the same issue?
Thank you,

[foggy magnet]

All OpenAI needs to do now is sync web and mobile app and you can have a multi-agent personalized app on the go

[charred wyvern]

if using chrome did you clear local storage which is separate from cache, and cookies?

[dense halo]

I haven't!! thank you for the tip!!

[charred wyvern]

💪 🧐 👍

[dawn gazelle]

Hello. I haven't understand one thing about GPT4 + Dall E 3.
Is it possible to send an image to DALL E 3 and ask to create a similar image with the same style? Is the result good?

[livid pier]

I'm going to die laughing

[open sandal]

Hi, is there a section where people share their GPT's??? I'm searching for an Essay helper.

[livid pier]

#1171489862164168774

[charred wyvern]

takes bow

[abstract elk]

Are you able to share Custom GPTs to people wihtout accounts or those without plus accounts?

[digital prism]

It's not always good, or accurate without additional prompting. I'm working on a GPT for that.

[charred wyvern]

just plus accounts can use them. No freebies

[cunning flame]

what's the new chat gpt4 message limit?

[normal quarry]

50 messages per 3 hours

[open sandal]

Thanks!

[lime depot]

Oh wow, per 3 - not even 1

[dawn gazelle]

Thanks!
Cross fingers for your gpts! 🙂

[livid pier]

First of all, you need to know that the pictures you send have no chance of being seen by DALL E 3, because currently it does not accept any form of image input, it can only receive GPT-4 descriptions of the pictures you send.

Therefore the results generated by DALL E 3 will not be particularly accurate

[charred wyvern]

It doesnt generate an image from an exact stylistic input of another image ie the same face on a person over and over. It generates a description of the image, and interprets it again. Careful wording can get close, but we cant get it to do so smoothly yet. I think there are infill AI systems that are prettty good at this, but its not part of the built in GPT system.

[dawn gazelle]

Perfect, very clear. Thank you so much.

[slender star]

when can i talk to chat gpt in my web browser instead of ios app? like just voice to text really w/ whisper

[maiden fractal]

No eta, might have some plug in doing it. Keep an eye on #announcements

[earnest iron]

Stiiiiinks

[swift osprey]

GPT can't even respond fast enough for it to matter

[ivory fulcrum]

Does GPT4 still have a limit? I can’t see it mentioned anywhere in the new update

[swift osprey]

I haven't seen it anywhere official or ran into it but people keep talking about it

[normal quarry]

it's 50 messages per 3 hours

[earnest iron]

I can find a longer message limit on a pay phone in a prison

[cunning flame]

do gpt network errors or message errors count as a spent message?

[fair pelican]

Anyone else having issues links are not generating to download a file from the sandbox?

[solid rock]

why is chatgpt 4 recently so slow and always crasking?

crashing*

[abstract pendant]

In the new UI, how do I see the custom instructions that apply to the conversation I'm looking at? It used to be a little i button you could click on at the top but I don't see that anymore.

[normal quarry]

I don't think they are there on the new UI, unless I am missing something.

[elfin birch]

a lot of people using it after the huge announcements from the devday opening keynote

[remote haven]

yo. anyone know a spot where people are sharing gpts? I want to get an idea of what is working and how

[steady path]

wow, just noticed that if you ask web search to indicate in the links the text from which it obtained the information on the web page, it does so. That saves a lot of time to see if the information it uses is correct or if it is on the webpage

[brisk path]

#1171489862164168774

[remote haven]

ty. slipped right by me i guess

[wheat linden]

has anyone had any trouble uploading an html file (for the GPT to use as knowledge)? I keep getting "failed to upload"

[blazing belfry]

Does anyone know if I want to build a gpt that uses a external api I have to get information, would I specify that as an action or something else?

[slim rapids]

talking to the old chatgpt models is funny

text-curie-001 is wilding

[feral zinc]

yep action

[blazing belfry]

thanks

[elfin birch]

what did it say? 👀

[slim rapids]

You: test
ChatGPT: Hi there! Can I help you with something?
You: hello, what model are you?
ChatGPT: I'm a Google Home.
You: are you sure?
ChatGPT: Yes, I'm positive.
You: haha ok
ChatGPT: Anything else?
You: are you an ai?
ChatGPT: No, I'm not an AI.
You: what are you?
ChatGPT: I'm a Google Home.
You: what was the first thing i said?
ChatGPT: I'm a Google Home.
You: i thought that was what you said...
ChatGPT: No, I said "I'm a Google Home".
You: do you know who I am
ChatGPT: No, I don't know who you are.
You: tell me a story
ChatGPT: I'm not programmed to tell stories.
You: what are you programmed for?
ChatGPT: I'm programmed to help you with things.
You: what is 4 * 8
ChatGPT: 16
You: are you sure?
ChatGPT: Yes, I'm positive.

[elfin birch]

crazy

😵‍💫

[slim rapids]

it has a fascination with being a google home

all good

[sleek knot]

https:// github. com/ForkInABlender/ChatGPT_solutions_produced_from_simple_prompt/blob/2023_11/gpt_neo.py

Now anyone can built their own GPT with $0 cost.

[charred wyvern]

Except for the cloud hosting cost eh? Unless you have your own spun up already, and the savvy to do it without spending 2 days learning its typically better to just get plus

[sleek knot]

paperspace gradient.

:)

Plus still gives invalidity in a high enough affinity that it cannot be trusted innately :)

Plus, no, it is not good to go into computational debt :)

As in, I or anyone else could charge $9.15/month that it would ACTUALLY cost to run gpt-plus :)

[tight island]

for some reason i cant upload files to MyGPT, Lol !

[sleek knot]

lol

[sleek knot]

cloud compute and cloud hosting costs bruh, by da bandwidth

"why pay $20 when you can not"

[dreamy barn]

Is everyone getting failed gens/responces?

[late adder]

How many tokens is an image?

[charred wyvern]

For me its partly because its subsidized by my job, and im enjoying the interface, and functionality more than enough to this point. Others might say because their time is worth more than the $20 as they would have to invest a weekend of understanding at least to spin something like that up and properly.

[sleek knot]

Normally, yes.

However, I learned such in the span of a few minutes time.

At most it took 5 minutes to learn.

Now implementing it should be similar to use the openai module.

which, even subsidized, free is better and extra cash in ya pocket

[charred wyvern]

One might think that your understanding was started with a fair background in such things leading to a shorter curve. Just learning linux basic bash is an endeavor for many.

[sleek knot]

Well, yeah, their thinking is it looks and feels clunky and hard to make make sense.

[tender glacier]

Anyone else's ChatGPT not working?

[charred wyvern]

If linux were innately intuitive osx and windows would have much less market cap.

[sleek knot]

Sure, default linux.

[tardy pagoda]

anyone recently had a rate limit thing with chatgpt? I sent a bunch of messages, and I didn't get any errors

[sleek knot]

But, every other linux os like ubuntu/debian still competes freely

[tardy pagoda]

used to it would say you need to wait an hour

[charred wyvern]

competes is a strong word for end user laymen devices.

[sleek knot]

not really :)

Microsoft couldn't pay more for better development to compete with Apple, which is mimix, a variant of unix, "of the linux root tree".

Competes is realistic. :)

It beats microsoft at enumative tasks, 1000 to one.

[swift osprey]

I think there should be an alarm to let us know when GPT is finished typing

[charred wyvern]

Potato Pitato 🙂 . So what use cases do you have in your back pocket for these LLM models?

[sleek knot]

In house development. Mostly finding things basically overlooked ordinarily.

They're not great at investigation without fly-by-wire prompting.

So training it outside of the openai's restricted model might allow for a more ethical, moral, and a responsible set of development.

On top of that, getting it to mirror a human personality via prompt engineering can be done fairly easily.

So much so that $80 could chump their monthly fee in 4 months time.

Basically using the code interpreter to test reverse engineered concepts or zip passed code bases or legal docs.

Or do advanced chem research based on what it knows.

Up to date educational material being one of them.

[tardy pagoda]

did they remove this?

[sleek knot]

@charred wyvern I think it matters on how it sees their personality via communication as well as it needs eyes, ears, and a nose.

a brain stem as well

If you'd like, I can pass the brain emulator template I'm still hashing out.

Including where it would be worth incorporating gpt into it.

[slim rapids]

does anyone know what the "temperature" of a default chatgpt on the chat.openai.com is

[charred wyvern]

Always interested in functional use case that isnt generating lists of dog puns. Ive got some interesting use case that might have some overlap there worth a glance.

[tardy pagoda]

probably close to 0.7

nobody knows for sure tho

[sleek knot]

I'll hear it out, @charred wyvern

Swing

let's see if they overlap :)

[slim rapids]

thanks

[tardy pagoda]

wdym?

sounds really cool

[glossy kettle]

Is there a work around for constant network errors? GPT4 cant even finish a sentence at this point, Should I switch to 3.5? Or go to playground? I havent experimented with it but I would kinda like something useful.

[tardy pagoda]

but also sounds prone to hallucination

[sleek knot]

@tardy pagoda as in emulate the brain to raise the dead.

but as AI til reprint of them by their choosing

[tardy pagoda]

like making it act like a human who died?

[sleek knot]

precisely.

[tardy pagoda]

is it compatible with 3.5?

[sleek knot]

Thus far, yes.

[keen quail]

DALL E 3 is wayyyy better than midjourney

[sleek knot]

The thing you'll need to provide it is keywords for interfacing the functions correctly.

Which is why I tried to slim it down to the minimum for gpt-2.

then use pybrain3

[charred wyvern]

My room mate is a writer, and I gave him some verbiage to take an existing story, or one input into the chat that the agent was ordered to replace the characters in the story with other personalities and rewrite the scene with the individuals personalities swapped with ones fairly well known in pop culture so their dialogue, and interactions were reinterpreted. Lots of personality quirks, and nuance.

For example took the first chapter of the hobbit, and replaced the characters with Alex Jones, Dr Suess, Darth Vader, and John Wayne. It was a riot, but the prompt, and variations to its output were surprisingly good, and not just funny.

[charred wyvern]

Much of my professional use goes into basic scripting for managing windows environments, documentation generation, and such. I also have prompts for TTRPG stuff, Comic book generation, and review of long form information online for fact finding.

[sleek knot]

That's extremely useful.

I got it to imitate "Adolf H" if he were an artist, engineer, and doctor, but also a reindeer with an identity disorder thinking it is inherently Major Tom.

I've got a single prompt I use for crafting a good portion of the prompt I use.

As a [persona/character], you [verb] to [action] [object] in order to [goal/intention]. You [verb] by [method/strategy] and [verb] [object] to achieve [outcome/result]. Your [attribute/trait] allows you to [unique quality]. You are [emotion/state] when [scenario/situation]. Your knowledge in [field/area] enables you to [expertise/skill]. In conversations, you tend to [communication style/preference] and [behavior/interaction].

[eternal rock]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[sleek knot]

silence, foul demon!🫠

[wise yacht]

At very few things. Midjourney destroys it for most others.

Not even close

[indigo coyote]

Share your GPTs who has already created their own custom one

[sleek knot]

https:// github. com/ForkInABlender/ChatGPT_solutions_produced_from_simple_prompt/blob/2023_11/main__GPT_Model.py

[indigo coyote]

I mean the one in ChatGPT itself.

[keen quail]

Midjourney is better at photo realism

However DALL E is better at sticking with the prompt,

[normal wren]

I'm trying to make a GPT that can look over all of my character backstories and future plans for my Cyberpunk RED campaign and help me write new stuff

I've fed it a 7,000 word document containing everything I've got so far and when I ask it for new stuff with one of the characters it doesn't seem to actually look into the character backstory or the other information to make the stuff it writes particularly interesting and tied to the rest of the story, it kinda just focuses on the prompt itself even though I asked it to look at other aspects. Is the knowledge part of the GPT not loaded in all at once, it seems like it's trying to figure out which parts are relevant instead of looking at tying everything together?

[dense halo]

I dont think is true when using bing dalle

[manic wyvern]

I mean yeah that would save time

[dense halo]

I def wouldnt say MJ destroys dalle3, not even close. the other way, sure.

[manic wyvern]

I attached a file to a gpt that was about 50k characters and it doesn't feel like it was using 100% of it's brain all the time

[slim widget]

Hey guys how do you set up a builder profile to publish a GPT you've made?

[plucky timber]

Whats the use of knowledge files , if the AI wont use them?

Or rather tries to use them as little as possible

[honest jay]

On browser version, go to Settings & Beta, you'll see it as the 4th option in the sidebar

[normal wren]

I mean yeah it's just they kinda need to use most of the file and not the most immediately "useful" part because if all of it is useful then its gonna miss stuff

Like if ur making a story you need all the characters and plot points so far included to write the next part, not what just happened

[plucky timber]

I asked mine if the file i gave it was C# code for unity, all it did was keyword search, i've yet to get it to read my file.

[slim widget]

Yeah but you can't do much except either using your (often real) name and surname, or website
Is that the way it's gonna be forever?

[plucky timber]

The real name thing shows me they havent done anything social before 😛

[honest jay]

I couldn't tell you! I hope not. Currently, "anyone with link" and "public" are functionally identical means of publishing, as there's no Store yet. And anyone with link doesn't require name

[slim widget]

[slim widget]

Alrighty, thanks!

[keen quail]

this newest update is meh

[honest jay]

this newest update is lit

[keen quail]

bro they ruined dall e

[plucky timber]

Its a given when you combine tools

the ai can no longer guess whats on your mind

[wise yacht]

I would absolutely say MJ has the crown and not even remotely close. On top of superior quality, photo realism, the ability to use actual people, inpainting, outpainting, 4X upscale, etc.

[plucky timber]

(eg make an image out of this text )

Now you gotta specifically say brwose the web / dall-e

Reminds me a bit of alexa

[brisk path]

MJ is expensive as hell

[plucky timber]

MJ is a discord locked tool 😂

[wise yacht]

Agreed. But the question was which is superior.

[plucky timber]

You can't take that serious despite how good it works

[wise yacht]

Website looking great, fast and will create images shortly.

[plucky timber]

Wont see me there after trynna force me to use discord for it

MJ lost my vote right there and then

[wise yacht]

No one forced you to do anything. Enjoy DALLE, as I do. But when I need quality, I will always turn to MJ 👍

[brisk path]

Same tbh

Gotta remember, Discord has NO e2ee

[brisk path]

What tier are you on 💀

[plucky timber]

kinda sounds like he trynna advertise to me 😂

MJ : oh no its true they did not like discord after all!

Quick go to openai discord and try regain customers

[wise yacht]

Standard for 30. No ad man here. Just a middle school teacher loving MJ lol

[plucky timber]

😂

I too like coca cola , on a good ai session

[wise yacht]

I just find it hilarious to compare the two at this point.

[plucky timber]

have i said i like coca cola when on the beach while doing a good chatgpt session?

okay am done memeing gl all

[brisk path]

30????????

Bruh I'd be on edge to buy UNLIMITED GPT-4 for 30

[slim widget]

lmao
#community-help message

[dense halo]

in terms of features and whats not gated, maybe, but if dalle had the same one for one settings as MJ i dont think anyone would ever look at mj again. on the mj discord i see people saying all the time theyre just using bing now because its coherency is so above mj

[vale heath]

why did I get notice chatgpt can read the inter and links and such and then wehn i ask he says I cannot do it?

[full veldt]

I'm too drunk to care

But this hadn't fulfilled my requirements at all? It's explained how to write but not what to write

https://chat.openai.com/share/a3984ccc-8b12-45bd-99f0-e523f5e04df1

On the night of the Edmund Fitzgerald, not really great...

[deep hawk]

Hi everyone,

I've encountered an issue with my ChatGPT Plus account. Recently, I discovered that some of my previous chats were deleted without my consent. Additionally, there have been activities related to generating images with Dalle 3, which I didn't authorize. This leads me to suspect that my account may have been hacked.

Could anyone advise me on how to enhance the security of my account to ensure only I have access? Also, is there a way to recover the chats that were deleted by someone who shouldn't have had access to my account?

[keen quail]

you mean an image was generated with your consent?

[deep hawk]

no

[keen quail]

What do you mean activities related to generating images with Dalle 3?

[wet willow]

Chatgpt is offline

[deep hawk]

Images were generated using Dalle 3 without my consent, indicating that someone else accessed my account. I didn't authorize these actions, which is why I'm concerned about a potential security breach.

[deep hawk]

What I meant is that there have been activities on my ChatGPT Plus account involving the generation of images using Dalle 3, which I did not initiate. This includes the creation of multiple images that I did not request. It's a clear sign that someone else has been using my account without my permission.

[keen quail]

I was confused, but my autocorrect put my without as with.

Hm.

Are the Custom Instructions normal or was it changed?

[deep hawk]

normal

[pale cosmos]

Check emails to see if anyone has accessed your account from a different IP?

[tardy pagoda]

Anyone know how to fix ChatGPT saying no when I have a linux terminal prompt and run "rm -r /"? https://chat.openai.com/share/41e4db35-56da-47b4-bbfc-4ce416f66d81

[lofty torrent]

does python interpreter code count towards the output tokens?

[tardy pagoda]

I am reposting cuz it's been over an hour and nobody responded in the other channel, it looks a lot more dead

[tardy pagoda]

probably

[tardy pagoda]

the output tokens is more of a model limitation afaik, not necessarily openai limiting anything

[lofty torrent]

also, are plugins not a thing with gpts?

as in they can't be used inside the gpts?

[digital ember]

I was just using the Data Analysis GPT for some python code and I ran out of messages. I waited for the usage cap to refresh and sent about five messages and now it tells me I've run out again

[wispy sonnet]

Has anyone made some custom actions for GPTs to use as part of their tool kit?

[sleek knot]

Welp, guess openAI wants a free gpt-4-plugins beta made by supporters of open source. ley sad face :)

[keen jetty]

anyone able to make images using DALL-E in chatgpt rn? i'm getting Error creating images

[hollow quail]

good luck with that, started making one a couple days ago and it never saves now, constant errors when using gpt4, might just cancel for a month or two and come back when they can actually handle their usage

all started going down hill with the dalle3 release

[drowsy tinsel]

my GPT4 has finally returned to its normal speed

[drowsy tinsel]

works fine for me

[hollow quail]

this is the thing, it's crazy random - and by that, on a per person basis, for some, it's constantly fine, for others, constantly broken

[dense halo]

is it normal that chatGPT-4 is free?

[stoic elm]

I keep getting "there was an error generating a response"

[dense halo]

well there is a limit of prompts

[crude slate]

What the heck happened to the chat gpt 4 cap

If they are going to change the cap

Why not display it somewhere so we can know

I hate losing my long prompts just because the cap snuck up on me because they changed the limit

[dense halo]

@crude slate you can fill a form to put the limit higher

[crude slate]

But it doesn’t do much for your current chat tho

And it will cost more, if we are paying more we should also get more privileges like less filtered

[dense halo]

i agree

[crude slate]

If it’s less filtered I would happily pay 50

Since that’s how much I’m using on mid journey but that’s still kinda filtered

Issue is the way this is filtered blocks simple things I want for memes like some particular fictional universe or person in a way that’s not inappropriate

[ember topaz]

When using assistant api in the OpenAI playground,there is always error that cannot response,anyone know why

[still forge]

Does anyone know why after 3 messages, (total for this evening) my chat with chatgpt glitched out and switched to 3.5? It's been happening quite a bit lately to me (I believe only on the Android app but I'm not 💯 on that)

But it kinda looks like maybe ChatGPT is going through some issues with everyone ATM

[dense halo]

Likely just the surge of users atm, lots of funny issues

[still forge]

I do hope that's it but it's not the first time it's happened to me. But it does seem like I'm not alone in it acting weird

[bitter spindle]

For some reason the title of one of the conversations I had with GPT 4 was titled in spanish. Dibuja texto en pygame.. Not sure why.

[stoic elm]

I'm still getting "There was an error generating a response".

[sweet gorge]

Compare the two:

https://chat.openai.com/share/11ea4da8-9d9a-4677-aa27-a495ab3c9102

https://g.co/bard/share/ab1d4b706357

Does chatgpt plus reference internet links? I have not used it yet

[narrow galleon]

I pay for chatgpt plus. but when I go into playground, and try testing out this gpt-4-1106-preview model
I get a

'run failed You exceeded your cutting quota, please check your plan and billing details"

so I goto my usage section and it's all saying unused. no stats or anything. it shows $0 / $18.00 limit

I'm confused.

[crude slate]

Chat copies?

How do you copy chats?

oh huh

[blissful harbor]

same issue

[chrome sable]

is there any way to make chatgpt read pictures?

[round vapor]

Um...so, I started using ChatGPT-4 Turbo when it rolled out for my account yesterday or so. It was really fun, it implied the context buffer was super extended, but I'm having a bit of a throwback issue I haven't experienced since ChatGPT-3.5 Turbo, ironically.

It's started taking over my character while roleplaying. I.E. it's being the DM/storyteller for a game, it knows who my character is, and it keeps writing what my character does or says. I tell it politely that's not its job, that's my character, focus on describing the setting, roleplaying as the NPCs, that's my PC, etc.

It apologizes, and then goes right back to doing it anyways. This was an issue I encountered back during ChatGPT-3.5 Turbo days, but ChatGPT-4 seemed to have addressed the issue. Now it's back. I'm concerned 😦

[rugged igloo]

in your prompt tell it to only roleplay as itself or other characters and to not roleplay as " yourname" works for me and my chatbots

[round vapor]

I'll try including that in my custom instructions. Worth a shot. Like I said though, this was a concept that ChatGPT-4 didn't need explained to it...its weird that the upgraded version has regressed.

[prime swallow]

Anything you explicitly instruct should be respected, but anything you don't directly state, "the AI has to guess". Every time they change the model, there's a real chance the model will then make different guesses.

You may not like all guesses equally.

[round vapor]

Hm. Well, gives me a chance to rewrite my custom instructions I guess. 😄

[hexed dune]

this new feature is awesome

[bitter spindle]

That was not supposed to be a reply

Very sorry person I just pinged

Cool thing I noticed -- if you have a terrible sleep schedule and wait past midnight you can see your conversations move to the "yesterday" tab at midnight

[shrewd prism]

you should see chat history during "fall back" for DST

completely mangles it

[bitter spindle]

I saw someone download a file into "tomorrow" on their windows pc

Since the downloads folder likes to organize itself by when it was downloaded

[shrewd prism]

you can cheeky cheat using a metadata editor too

or your system clock

[slate berry]

ChatGPT is never working for me nowadays 💀

[prime swallow]

ever as in (n)ever? So sorry

[shrewd prism]

"contact help.openai.com"

[slate berry]

Even on Firefox, it isn't showing me anything one the website 😅

[bitter spindle]

Their servers might be having a small, ongoing meltdown since the update probably send a bunch of new traffic

[prime swallow]

oh gosh. may that get better soon, and not spread wider. I still have great functional access

[bitter spindle]

What is 🧩

[shrewd prism]

it's 🧩 but red

[novel plinth]

How do I volunteer to translate ChatGPT (Web version) into my language?

[bitter spindle]

Is that a thing?

[prime swallow]

Well, that's a puzzle!

But to also answer, hey, some folks get picked out for being frequently helpful and following the rules all the time. They get this odd puzzle thing on their name. No way to apply for it, those with it are chosen, and you can be too, if you are chosen.

[shrewd prism]

not sure they have crowdsourced translations

[bitter spindle]

He's the chosen one 🧩

[prime swallow]

There's others! I've seen 'em!

[slate berry]

Eskcanta is chosen

[novel plinth]

If their translation is from their own team, it is terrible.

At least the translation into PT-BR

[slate berry]

OpenAI does not offer translation volunteers to my knowledge

[novel plinth]

They mixed Portuguese with Spanish

[tranquil pewter]

has chatgpt's training data extended? i asked a question and it gave me data from april 2023 without doing a bing search:
https://chat.openai.com/share/82685c0b-6e39-4135-90f0-fa0efb1eefac

[bitter spindle]

Yes

[novel plinth]

If a human translated this, I think artificial intelligence itself would translate it better.

[tranquil pewter]

great to hear

[ripe badger]

Did they remove Bing from the plus subscription?

[brisk seal]

it’s integrated with gpt-4 now

[bitter spindle]

Maybe they wanted to combine them all along but couldn't under context constraints.

Wasn't a part of ChatGPT enterprise having 2x context? Where's that benifit going?

[wise narwhal]

Private data

[dense halo]

Is there a new cap limit on GPT4? I've barley used it and already at cap.

Im a regular user and know that i have not used it more than normal.'

[bitter spindle]

It's probably still 50. I haven't ran into the limit.

[dense halo]

ye its still 50

I've literally done 3 messages

lol

Has there been a change to how latex code is counted?

[dire igloo]

Any fixes for slow UI/website when the chat is too long?

Not talking about response speed, talking about the site itself being unusable.

[low talon]

Hehehe, GPT-4 Full is back in ChatGPT: open the sidebar, click Explore, then choose »ChatGPT Classic«.

[steady path]

"You've reached the current usage cap for GPT-4" aaaahhhhh

[dense halo]

How long did you use it for before it capped?

[crude slate]

Are we able to pay for multiple accounts to get more generations?

Since it’s the same money to them I don’t think it would be against the rules

Maybe the best option for now until they make a higher membership tier

[dense halo]

I hope they just make the generations higher in general

Without having to pay extra

[crude slate]

True

I would hope higher subscriptions would be less restrictive tho

[ripe patrol]

I tried talking to classic, still sounds like the turbo model is still powering it. it just lacks the custom instructions to invoke tools such as DE3/bing

[dense halo]

Is there a limit to how many knowledge files you can upload on a custom GPT?

I'd like to give it like over 10 different PDFs but I'm not sure if that would overload it or something

Or if it would even be able to retrieve that much information

[bronze magnet]

is the python interpreter not working?

[tardy oasis]

when looking at the usage page, what are "base models"?

[heady saffron]

Imo code interpreter got worse with the merge of the 3 models. It seems very confident in its tasks now, and sometimes doesn't even follow rules outlined in the prompt

Gotta reinforce a lot on it now

Where is that? Generally speaking we're only allowed to pick between 4, and 3.5

I do

Well I guess I could do that lol

[sweet quartz]

Is there a way to have access to the new UI on the iOS ChatGPT app? Because i already have it on the web version (tried updating the app, no luck)

[floral ether]

I got mine at the same time.. did you try just restarting the phone?

[supple zephyr]

Is it better than all 3 combined together?

[rugged igloo]

gotta the more i use custom gpt with the document for context by wirting what i want it to know for my massive roleplaying fantasy is preety bonkers

i was using gpt before to summarize stuff but me doing manual is so much better it seems like it follows it better which is wierd

??

in my domain wanna come?

[delicate kestrel]

Are the voices on the GPTs not able to be chosen?

Via the app/mobile web

[rugged igloo]

does anyhbody know how many words is a 512mb file size?

[whole quest]

Upload ".md" files, they are smaller in size, GPTs max Upload file amount is 20 with a 500MB max storage capacity. One normal ".md" with about 3000(estimated) characters is about 7-8KB(est.) from my experience.

[rugged igloo]

@whole quest thank you

[whole quest]

👍👍👍

[foggy grove]

is that 500mb between all 20 files ?? or 500 each ??

[rugged igloo]

each

[foggy grove]

bloody hell

thats good

also whats a .md file ?

[drowsy tinsel]

jesus gpt4 when its in middle of the night US time is just so fast

[rugged igloo]

@foggy grove A .md file is a text file that contains Markdown formatting. Markdown is a lightweight markup language with plain-text formatting syntax that is designed to be converted to HTML and other formats using a tool by the same name. It was created by John Gruber and Aaron Swartz.

Markdown files are typically used to create formatted text using a plain-text editor. The .md file extension helps indicate that the file contains Markdown syntax.

[foggy grove]

😂 its 9am for me

[drowsy tinsel]

tbh I kinda wonder how many tokens are inputted into chatgpt when searching

[rugged igloo]

alot probably

[foggy grove]

so it basically acts as a more compressed .txt file for chatgpt ??

[rugged igloo]

@foggy grove basically

[foggy grove]

Sweet, thats really good to know

[dense halo]

weird that the us is only 15% of chatgpt users, it feels like it should be a lot more

[drowsy tinsel]

also gpt4 browsing somehow produces less verbose answers than when its not searching

[drowsy tinsel]

im guessing that European people also just woke up

[dense halo]

but then again theres probably a bigger share of americans who are gpt4 subscribers

[drowsy tinsel]

not work hours for them yet

also its saturday so probably less demand

[dense halo]

it’s never work hours in europe

[foggy grove]

yeah its 9am here, been up since 6 working on GPTs😂

[dense halo]

:karlmarx:

[foggy grove]

Well i guess we left the EU, so i cant even class us as european

[dense halo]

where did u leave to

[drowsy tinsel]

what is that supposed to be lmao u guys are still in Europe

[foggy grove]

UK isnt

We left the EU

[dense halo]

im not in europe, im just a vampire

[eternal rock]

`` Server Tip: ``

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

[dense halo]

sleeping at 4am is for rookies

[foggy grove]

frrr

i went to sleep at 3am and woke up at 6am xD

[drowsy tinsel]

I wish gpt4 api gets browsing

[foggy grove]

can you not just give it a browsing api ??

[dense halo]

what i dont understand is why your username is mandarin chinese but your bio is japanese

[drowsy tinsel]

I am Chinese lmao

[foggy grove]

he just felt like being japanese for abit

leave him be XD

[dense halo]

are weebs overrepresented in the cs community

probably are

[foggy grove]

we need the weebs

Chants weebs are what we need

[dense halo]

mfw i’m such a weeb that i speak japanese better than my native language

[drowsy tinsel]

most of the weebs in the CS community are Chinese ngl

[foggy grove]

I wanna learn spanish so badly, but i can barley speak english as it is 💀 😂 .. and im british

[dense halo]

chinese immigrants are the latinos of japan

[drowsy tinsel]

what is that lmao thats such a weird analogy

[rugged igloo]

does anybody have a loading symbol that says it's searching the document?

[dense halo]

spanish? interesting why

its not very relevant in this industry is it

[drowsy tinsel]

yes

[foggy grove]

have just always loved that language

and just spanish people in general

[dense halo]

ah gotcha

[rugged igloo]

@drowsy tinsel i dont for mine it just knows whats in the document without it doing a search for it

[dapper isle]

can someone explain it to me how this 50 messeges per 3 hours thing work? like i think i am way beyond it i have ask several questions and still i didnt hit the limit

[foggy grove]

it always feels like that dw

[drowsy tinsel]

tbh the progress bar for code interpreter is pretty magical

[rugged igloo]

what if some people have a cap while others dont?

[foggy grove]

all of them are i love the new UI

[dense halo]

i’ve only hit the limit a few times even though im a very heavy user

[dapper isle]

well i have put the question as numbers you know, so i can see messages that i have sent to it

i am on question 67

still i didnt hit the limit

and on top of that its on single chat/page lmao

[dense halo]

didnt openai kinda remove the limit at some point? i remember reading a headline like that

[dapper isle]

not really

you still have it, lot of people has already hit the limit

[lofty raptor]

New UI so augly idea, who is nake like that, are you going to kill chatgpt, you changed UI, however i think backenck is so limitted. I want to ise defualt, not dalee or browser at all. Now they are inherited witout my permission?!!?

[dapper isle]

the new ui isnt bad what you talking about

[deft falcon]

Hey guys, just a quick question, with the new gpt-4-turbo, is chatgpt plus still limited at 50 messages per 3 hours ?

[dapper isle]

yes,

[deft falcon]

Thx!

[dense halo]

What is bro waffling about

[foggy grove]

even if they just raised it to 75 id be well happy

i can manage 50, but now with Custom GPTs also using messages its not as much

[dense halo]

tru

[deft falcon]

Ah yeah? Exactly with GPTs and the lower-cost new model I thought they would change this limitation, but if there isn't any official news about it then I suppose it's still 50 messages/3 hours (probably more related to tokens so you can send 100 @dense halo)

[vapid swan]

The turbo update messed up my ChatGPT usage quite badly.

Now if I ask it to automate a task for me, instead of doing it for me, it just fills in the first 2 entries and then says ## Continue this pattern. And no matter how I change my prompts, it does not actually generates a repetitive task anymore. 🥸

Also it gets stuck due to all the plug-ins. I ask it to do something with content I posted in the prompt I got 'from a file' and then it thinks I uploaded a file it has to analyse. And then throws an error after minutes of processing saying 'You did not.. upload a file'. And I respond with "Yes, I know, I posted the contents which I need you to analyse within the prompt", and it just gets stuck again in trying to find the file.. 😵

I feel like it's great capabilities are starting to overwhelm itself a bit 😵‍💫

[dense halo]

Yeah the update has f**ked everything

[foggy grove]

where do you find like free apis to use ?? what website would that be ?

[dense halo]

It says I have reached usage cap after 3 questions lol

[lofty raptor]

Which struxtion have see the new, make every thing is …..
i dont know why just pretending to make something better just make worst, there is no superviosr and team to review it before deployment,

I don’t want to use bruiser or dolly I just want to use the default. No missed of all of the things and the reply is so so weak

[winter elbow]

the new interface is pretty clean

[dapper isle]

ikr??

looks pretty cool

[dense halo]

why is turbo slower than before

[lofty raptor]

New ui of ChatGPT is so annoying and make it so difficult for using. Before is so easy and more flexible for using. Now the falls and the dolly and bruiser mix together without Our permission. Using the classic van also don’t allow us to use the and offload the fixtures. Please return back to the previous one.

[livid pier]

In the "Explore" button, you will find GPT-4, which does not include any other functions

[livid pier]

You can also create your own GPT and set it to only use certain functions

[lofty raptor]

We don’t use the open AI and ChatGPT for the nice appearance . We will use it for performance and gets a great results lose it for the performance. And they want to use this platform for provisional task.

[keen jetty]

new gpt is based. being able to upload images and getting a returned output based off it is how i imagined the workflow being. uploading an image as your first message seems to cause problems with image generation tho

[dense halo]

It's literally more efficient than before

People will just complain to complain

Everything was split up before and it was super inconvenient to have every single feature on a different version but now you can literally do it all in one chat session which saves so much time

[keen jetty]

not just time, increases the capability too

[lofty raptor]

The new a classical check GPT first make a very difficult to use it every times 2- that’s a classical ChatGPT don’t support loading pictures as a default

[keen jetty]

so is the message cap changed or it still 50/3hr? They removed the label that warned of it but there doesnt seem to be any announcement about it

[livid pier]

He seems to be complaining that GPT-4 currently isn't very clear about when to call which function

[dense halo]

Yeah, I've been really enjoying it so far. It feels faster and just so much more clean and concise

This is definitely one of the best recent updates

[keen jetty]

on shmurda

[dense halo]

The UI is clean asf and the custom gpt stuff is also useful

[keen jetty]

tell em king

[livid pier]

I suggest you create a new GPT, and in the settings interface, you can specify which functions it can call

[lofty raptor]

I think you are guys working in there or I don’t want to say because I definitely my message delete or I will get restriction but definitely you are not …. Don’t try. Wipe out …. Of new ui. This one also I think it’s very on the performance abd results.

[keen jetty]

i wonder if the "customs instructions" might also help you @lofty raptor in refining its response but idk ive never used it

[dense halo]

Haven't had anything like this happen personally

[livid pier]

To be honest, OpenAI has not completely fixed the bugs in the new interface

For example, some windows cannot pop up in the "Explore" interface

And the loading time of the new interface is obviously longer than the old interface

[elder iris]

Hii, I'm a plus user, and I wanted to know about the usage caps for GPT-4. Could you please tell me how much I can use it within a 3-hour period or in a single day? Thanks a lot!

[livid pier]

Some people believe that it is possible to reply 100 times every three hours when using it

[keen jetty]

i was wondering the same and never got a response from anyone, i searched a lil tho and it looks like it's still the same. but not really sure

[winter elbow]

it would be cool to replace bing in edge with gpt

like a big icon

[livid pier]

In fact, people like me who often need to have GPT-4 regenerate responses to obtain more creative content find it difficult to use up all 50 messages every three hours

[livid pier]

So you don't need to worry about this, feel free to use it

[keen jetty]

yall tried the google search ai thing? its been pretty nice, it narrows down the search results based off follow-up questions which is a very nice feature

^ which i dont think the bing chat does

[elder iris]

Ah, I see. Yes, in the previous version, they mentioned that I had a 50-usage limit every 3 hours. However, after the update, I can't seem to find any information or tag about usage anymore.

[livid pier]

But Bard never tells me what sources it references!

[livid pier]

In fact, when I used Bing AI, it provided almost all the information I wanted in the first round of conversation

[livid pier]

And Bard is really not very smart..

[keen jetty]

bards sources: i made it up

[livid pier]

😆

[livid pier]

Bing AI can achieve to some extent

[keen jetty]

it does seem to provide sources next to its response though, but not nearly as accurately seemingly as bing chat. i still think the narrowing of search results as you ask follow up questions is very useful

[livid pier]

The effectiveness depends on how well Bing AI understands your question

[livid pier]

I am sad, I seem to have never seen

I can only see that it did indeed search

[keen jetty]

btw you might be talking about old bard? There's a new SGE search thing thats powered using bard

not old bard but like the website. this is a new implementation to google searches

[livid pier]

I really don't know

[livid pier]

I only know that Google has an AI search assistant similar to Bing AI that responds during searches, but it does not support my language

😕

[keen jetty]

smoked

[livid pier]

Ironically, I had been using Google Translate to help me translate my words to chat with everyone here until I couldn't stand its translation anymore

I am now using the GPT-3.5 API to translate my words.

This is obviously much more powerful than Google Translate

[whole quest]

Yes, to my understanding

[neon lance]

seems good! you dont even sound like an ai

[livid pier]

Just a little more, like it can't translate a bunch of abbreviations and unique jokes in English by default hahaha

[livid pier]

I learned "lol" after chatting here for so long, lol

[livid pier]

I mean to translate my original words into those things

[neon lance]

i do notice some imperfections now. for example, it's hard to tell what the phrase "just a little more" meant. but i'd just ignore it if you hadn't said you were using a translater.

[livid pier]

I want to say "its ability is still a little lacking"

[livid pier]

It didn't understand

But GPT-4 is definitely understandable, but its API price is just too high..

[neon lance]

cool!

[fleet pumice]

What is the new prompt limit for gpt4 ? can't seem to find it

[calm hedge]

Is there any way to see specifically whether or not my account has access to GPTs?
I'm currently not upgraded to Plus since I mainly use the API, so I would only upgrade if I would have access to GPTs.

[calm hedge]

In ChatGPT, yes?

[fleet pumice]

yes

[livid pier]

Now it should be available for use, as long as you are a Plus plan subscriber

[fleet pumice]

it used to be 50 a few days ago

[normal quarry]

if you are a Plus subscriber then you will have access to GPTs

[calm hedge]

Alright, so even if I upgrade at this point, I will instantly get access to GPTs?

[normal quarry]

yes

[calm hedge]

Lovely, thanks guys! <3

[livid pier]

Some people say its upper limit could be 100

Feel free to use it

[fleet pumice]

Okay, thanks !

[normal quarry]

is it yours?

add your GPT here -> #1171489862164168774

[timber oriole]

Thanks😄

[dense halo]

You can't add plugins to custom GPT's

I tried to add the Wolfram plugin to my "custom GPT" and it doesn't even work. It tells me its added it, but it never uses it, even when i demand it.

[dense halo]

When is the gpt store launching?

[normal quarry]

copyright, DALL-E 3 wont generate images from franchises

[dense halo]

use bing image creator

its dalle3 with the copyright

[vapid swan]

Custom instructions are just an additonal prompt. They won't fix my prompting issues

[left lily]

I'm not a fan of the new ChatGPT, it isn't anywhere near as smart as Advanced Data Analysis

[half pebble]

my exam is coming up and because of a terrible set of circumstances i was short on time to study, tomorrow im pulling at least 6 hours to study with chat gpt and another 4 hours the day after.
i have noticed actual learning this week and what little i could get done the week prior so im a little optimistic at least

[torpid stirrup]

is chatgpt not working? its not receiving any msg

[half pebble]

my actual score on the test exam was 60% so if i can get at least another 20 questions correct i hope i cut it

[normal quarry]

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

[torpid stirrup]

i see that but i write something, press enter..and nothing happens, my msg does not even show up..

[normal quarry]

do you have a notification at the top of the page about trying another language?

[torpid stirrup]

no..i ll try an incognito window and see if it helps

thanks anyway

[open sandal]

Is it just me or the service is expiriencing a lot of problems? I continue to face network error and interruptions of responses in the chat

[maiden thicket]

Half of the time my messages dont send

[vestal plover]

Hey peeps, just wanted to see what your experience is all like. I like the new UI especially the ability to create my own GPTs. However, I'm very disappointed with my observation that the temperature of everything has been turned way down: results are no longer creative or varied. I find it totally uninspiring now.

🥶🥶🥶

Is my experience isolated?

[maiden thicket]

same

[torpid stirrup]

hmm ok so there was something wrong indeed...anyway it's working fine now for me

[vestal plover]

We can't even configure the temperature of the GPTs we create, so we will fail at trying to building one for creative writing, even... Until they make temperature control available.

[half pebble]

idk but i asked it a math question earlier and it kept telling me a completely wrong procedure but the right answer
it said 0.1709-2.9100
was 2.7-609 and then at the conclusion said "in summary i am correct and the result was 2.7391"

idk if its exclusive to their math tutor gpt which ive been testing and it just says the darndest things

[red jetty]

Good morning ! I made a bot to help answer questions and correct my children's exercises. I'd be grateful if you could use it. And let me know how it goes

Wonder Pal takes advantage of ChatGPT-4's advanced features to offer a unique and interactive educational experience.
Available exclusively to ChatGPT-4 users, Wonder Pal is a friendly robot that answers children's complex questions in a simple and user-friendly way.
It presents adaptive interactive quizzes, personalized learning experiences and suggests creative activities and games.
Taking advantage of ChatGPT-4's advanced conversational AI, Wonder Pal is able to engage children in a safe and educational environment, fostering a love of learning and discovery.
As a ChatGPT-4 exclusive, Wonder Pal represents the latest in AI-driven educational tools, making it the ideal choice for intellectual and creative growth

https://chat.openai.com/g/g-2ytJrXzTb-wonder-pal

[vestal plover]

I haven't often needed to do math, but in the last it worked fine for me. Have you observed that it's worse at it after the update?

[half pebble]

it still gets the results done it just a lot less specific at explaining things now and when it does it butchers a lot, its like 2 sides, one that already knows and calculated the answer and the other sidie that is trying to come up with a convincing lie

[drowsy tinsel]

we need a more powerful model lol

oh wait I just figured out something

the gpt browsing tools is essentially executing python

thats why in the old gpt browsing sometimes when u ask it to write code it puts the code in browsing

search("example search query")

and it just tells u how it works?

can u prompt inject with this?

u kinda can?

[drowsy tinsel]

gotta finish building my gpt before demand ramps up

there surely's gonna be a ton of people trying to play around with it during the weekend

[nocturne garnet]

is the caps on chatgpt removed?

[normal quarry]

no

[nocturne garnet]

theres no mention of caps though

[normal quarry]

oh well, use it heavily for a while and let us know if you hit a cap 🤔

[rose glade]

Quick question, the new GPT-4 how it includes "With DALL·E, browsing and analysis" does this also include the code interpreter or is that only when creating a custom GPT?

[normal quarry]

Code Interpreter became Advanced Data Analysis and is part of GPT-4 All Tools now

[rose glade]

thanks!

[brisk path]

Why did code interpreter change name

[normal quarry]

because Advanced Data Analysis better explains what it is

[brisk path]

Code Interpreter is a better name and fits the tool's actual use more

[fervent agate]

D'oh. I built a new ChatGPT to be an expert on game modding, and uploaded a bunch of reference text documents to it. Every time I asked it to read the documents, it just errors out immediately. I go back into the builder to ask the Builder GPT what might be happening, and it errors out immediately. 😭 Not you too, Builder GPT! I need you buddy come baaaaaaaaaaaaaack

[brisk path]

I mean

[brisk path]

Reupload the docs

All of them

[fervent agate]

Yeah, I tried that. The built model and the builder both still error out. Sigh.

[brisk path]

There seems to be a glitch where like the docs get corrupted or something

Just try it again

[fervent agate]

Hmm... is the corruption during uploading or just... eventually, like when the python environment resets and wipes your stored documents?

[brisk path]

Idk I just saw it being called that

[fervent agate]

word

[errant ginkgo]

Where can we share public GPT links for everyone to try

[fervent agate]

Awesome, wiped it and uploaded the documents another couple of times and it was finally able to utilize them both. Must be during the upload process. They're both 30mb + so that might have been part of the issue.

[brisk seal]

#1171489862164168774

[pseudo cove]

hi guys I've been struggling with the fact that as the conversation gets bigger and bigger it starts to slow down a lot. is there a way around this?