#chatgpt-discussions

1 messages · Page 27 of 1

livid pulsar
#

I’m taking in general, but yup u r 100 correct, the thing is people are just scared of coding they just don’t understand or, are afraid to try it

winter elbow
#

can we at this moment make gpts public ?

brisk path
#

All custom GPTs, including the OpenAI built ones, use the January 2022 cutoff date

#

Plugins GPT-4 and GPT-3.5 turbo also uses it

livid pulsar
lone wyvern
brisk path
#

All in one uses the newer April 2023 cutoff though

sweet quartz
#

now it's fixed lol

winter elbow
soft turtle
#

Turbo is only available via API, correct?

fast forge
livid pulsar
soft turtle
#

hmm, not seeing it, not even in beta features

brisk path
#

I just had an idea

fast forge
brisk path
#

What if you made a GPT with all the tools + the Wolfram alpha api

hot tapir
#

guys how long are the threads now? Can they go on forever

soft turtle
outer seal
livid pulsar
#

@lone wyvern u have the new one with vision, browse and dalle integrated one right, what’s the context window for that?

near vine
fast forge
#

What i said about PDFs :
To transfer knowledge effectively in a format that is optimal for analysis by AI systems like me, text-based content is preferable. A .TXT file, which is a plain text file, is inherently more straightforward because it contains unformatted text. This allows for direct reading and processing without the need to interpret or disregard additional formatting, images, or complex layouts.

A PDF file, on the other hand, can be more challenging to analyze directly because it often includes diverse formatting, images, and other elements that are part of a more intricate structure. While I have the capability to extract text from PDF documents using specialized tools, the process can sometimes be less efficient than working with a .TXT file. This is due to the potential need to parse and interpret the layout and design elements to access the textual content.

#

So if y'all want to get your knowledge directly without problems, better copy your PDF and put it in the knowledge so he can scrapt some information from the source, idk if this features is the same as a DATABASE or not

lone wyvern
livid pulsar
#

@fast forge thanks for the tip

fast forge
crude slate
#

I don't have the ability to browse the internet or access real-time databases.

fast forge
#

I tried it sending s tons of text and it literally knew everything

near vine
# 
The message or response length for this model, based on GPT-4, is primarily constrained by the token limit, which is 8,192 tokens. This limit includes both the input (the prompt and any previous dialogue) and the output (the response generated by the model).

Here's how this impacts message or response length:

Combining Input and Output: If your prompt or the ongoing conversation is lengthy, it will consume a portion of the token limit, thereby reducing the number of tokens available for the response.

Long Conversations: In a long conversation, older parts of the conversation will eventually be "forgotten" as they are truncated to keep the total token count within the 8,192 limit.

Response Length: The model's response length can vary, but it's typically concise enough to convey the necessary information while considering the overall token limit. Responses that are too long might be truncated.

Token Definition: A token is roughly equivalent to a word or a part of a word, so the actual length in terms of words can vary.

It's important to note that this token limit is a technical constraint designed to balance between providing comprehensive responses and maintaining efficient performance. For most typical uses, including detailed explanations, conversations, and queries, this limit is sufficient to provide thorough and relevant responses.```

mine is saying this
crude slate
#

The what, did they break the search with bing

livid pulsar
crude slate
#

since I can’t even ask it to search now with the all in one browsing

fast forge
fast forge
near vine
#

okii

livid pulsar
small sphinx
#

apologies if already discussed, but gpt-4 is givne me the message along the lines of "my ability to generate images has been disabled"

near vine
#
  1. what is the name of your model?
  2. what is the message cap of this model?

and give us an update

livid pulsar
#

I’m probably gonna be one of the last ones to get this update

twilit ibex
#

finally got access to everything

twilit ibex
near vine
#

congrats

livid pulsar
#

I really like the new ui on ios

near vine
#

ikr?

wise anchor
#

what have people done with it?

fast forge
#

So for ex : you can copy your whole 100 pages PDF and send it as a single message, you can literally know that he will get everything from it (after all idk about accuracy but they said its more accurate)

near vine
fast forge
#

It is available in the plus users mate, you use the gpt 4 turbo

wise anchor
#

i think its more accurate because it doesnt have to pick and choose what info to return or look at. it can send more back to you so it can consider more of what you give it

near vine
#

wait so what the model can give is limited to 8k but we can fed to it is limited 128k right?

fast forge
#

You
But you can see more than 128K tokens right ?

ChatGPT
Yes, that's correct. While the immediate input and output for each interaction with the model are limited to a couple of thousand tokens, the model can access and refer to a much larger context from the conversation history, which can be up to 128,000 tokens. This allows the model to maintain continuity over a series of interactions and refer back to earlier parts of the conversation if needed.

near vine
wise anchor
#

i thought chatgpt was a user LOL

#

But from what i know the context window is only increased for the API is it not

fast forge
livid pulsar
brisk path
#

What if we unleashed ChatGPT 4 as an actual discord bot that acted like a user

livid pulsar
#

And people being able to add knowledge bases

outer seal
#

every time i refresh i get the old ui for a split second before it switches over 😅

wise anchor
near vine
fast forge
#

When he says 8K tokens : ITS HIS REPLIES not you, but for you, you can give him 128K tokens and he can read it !

wise anchor
#

although it doesnt just chat randomly it has to be prompted

brisk path
wise anchor
#

yeah i figured that after i sent that lol

brisk path
#

It would be expensive as hell but fun

near vine
#

its so good!!

brisk path
#

ny

fast forge
wise anchor
brisk path
#

But with 4 turbo with vision

wise anchor
#

because you can fine tune 3.5

near vine
#

but we dont know the message cap yet for plus user

brisk path
#

It might become a reasonable idea

normal quarry
#

message limit is 50 messages per 3 hours

eternal rockBOT
#
This is a Safe for Work Discord Server

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

Posting NSFW content will result in a permanent ban.
outer seal
#

would be huge for paid gpts if they bypassed the 50 cap

#

is the only thing keeping wrappers alive atm

near vine
#

or is it kind of a glitch

fast forge
#

They said they removed the cap in the API, we don't kow if they did for the plus users as well (ofc it would be awesome if we can talk as many as we want)

outer seal
# 
"message_cap": 50.0,
"message_cap_window": 180.0,
#

still 50/3hr

near vine
#

ooh

fast forge
outer seal
#

no

dapper isle
# fast forge When he says 8K tokens : ITS HIS REPLIES not you, but for you, you can give him ...

can you see 128k tokens?

ChatGPT
ChatGPT
No, I cannot process 128,000 tokens in a single interaction. The GPT-4 model, which I am based on, has a maximum token limit of 8,192 tokens. This limit includes both the input (the text you provide) and the output (the response I generate). The token limit is a fundamental constraint of the model, determined by its design and the computational resources required to process information.

In practical terms, this means that in a conversation or an interaction, if the combined length of the input and output exceeds 8,192 tokens, the oldest information is truncated to make space for the new input and responses. This limit ensures a balance between providing detailed and contextually relevant responses and maintaining computational efficiency.

why it showing me this?

kindred oracle
#

bruh its for the api turbo gpt4 model that has 128k

lone wyvern
kindred oracle
#

they doubled the API limit 🤓 not chatgpt (gpt4)

dapper isle
near vine
kindred oracle
#

not 4

left lily
#

What's the new effective token limit for ChatGPT-4 premium users? It doesn't seem like it's 128k

outer seal
#

lmao quit asking the model for info on its parameters. it has no clue and will just make stuff up

scarlet mulch
#

is chat gpt v4 limit removed?

near vine
scarlet mulch
#

i cant see the notification of 50 message limit anywhere

livid pulsar
fast forge
humble pagoda
#

Chatgpt is pretty much unusable right now. Even vision bugs out. Gpts so much too.

livid pulsar
#

It uses ChatGPT api

scarlet mulch
fast forge
#

Even if you ask GPT 4 Turbo if he's Turbo he won't know, the only reason to know it, it to ask him when does his KNOWLEDGE ends

kindred oracle
#

LISTEN huys GPT3 is turbo in chatgpt while GPT4 is not, it is only turbo in the API (not chatgpt) and is rolling out to DEVELOPERS 🤓

fast forge
#

Which is April 2023 !!

left lily
#

What's the new effective token limit for ChatGPT-4 premium users? It doesn't seem like it's 128k

kindred oracle
humble pagoda
outer seal
#

32k

kindred oracle
scarlet mulch
left lily
scarlet mulch
#

anyone knows if the 50 message limit is removed? i cant see the notification anywhere like before

kindred oracle
left lily
#

16k seems about right from my testing

kindred oracle
left lily
#

Basically, double from what Advanced Data Analysis was doing before

fast forge
scarlet mulch
outer seal
kindred oracle
#

yeh ik but not in chatgpt

kindred oracle
near vine
fast forge
#

I listened again to the OpenAI conference and at 19:15 Sam Altman says:

A small one, ChatGPT now uses GPT-4 Turbo, with all the latest improvements, including the latest knowledge cutoff, which will continue to update. That’s all live today.

kindred oracle
fast forge
#

Its not my words its sam's

near vine
kindred oracle
wispy iron
#

hi guys anyone figured out how to connect github account to gpts actions ?

left lily
#

Is the 50 message rate limit gone?

normal quarry
kindred oracle
kindred oracle
outer seal
#

no but your search bar is

wispy iron
#

so i have to setup my own git host ?

left lily
#

I think it's higher than 16k, I just tested it on 23k tokens and it's doing okay

outer seal
#

its 32k

kindred oracle
wispy iron
#

nice thx for the hint

kindred oracle
outer seal
kindred oracle
kindred oracle
#

the openai emojis are lit ngl

#

bye!

outer seal
left lily
#

BY MY TESTING -- it does not remember 29k tokens

outer seal
#

i love this guy so much

verbal heath
outer seal
left lily
#

BY MY TESTING -- it does remember 26,600 tokens

#

I'm not sure why, as long as I've used ChatGPT the real token limit is never the same as the advertised token limit

hidden ibex
#

I uploaded a big pdf and got over 12k tokens + intelligent review of what was in the pdf, trying to incorporate a feature from a paper into NanoGPT

oak flame
#

Good evening, is chat gpt 4 having issues with image inputs ? I always get an error when sending an image input after ChatGPT thinks for a long time without a response.

hidden ibex
drowsy harness
# oak flame Good evening, is chat gpt 4 having issues with image inputs ? I always get an er...

Sous Chef
I'm sorry for the confusion earlier. There seems to have been an error, as the description provided for the tuna dish does not violate any content policy. Let's try generating that image of the Seared Tuna Steak with Avocado and Spinach again. Please hold on for a moment while I take care of that for you.

Error creating images ⚠️

I apologize for the inconvenience, but it seems there's an issue with generating the image of the Seared Tuna Steak with Avocado and Spinach according to our content policy. While I can't provide an image, I'm here to assist with the recipe and anything else you might need. If you'd like, I can continue by providing you with a detailed recipe and a shopping list for this dish.

outer seal
#

i gave it slightly less than the max & it held up, also at that point you should be using retrieval but still very cool that it holds up

hidden ibex
#

I wonder if there might be some injected multi-shot examples now with the extra space, so 32k is the effective limit but they've got 4k tokens worht of moderation/consistency/internal only prompts

outer seal
twilit ibex
#

anyone else chatting with GPT Builder?
literally just having a conversation in the editor lol (idk if this counts towards the message limit or not (it thinks it runs on a modified version of GPT-4))

outer seal
dense halo
#

it does count just fyi

still loom
#

Plugins disappeared from my chatgpt .. does anyone have similar condition?

hidden ibex
#

should see a dropdown menu with gpt-4 , gpt-3.5 options, and a plugins widget below those

tardy plover
#

When will the GPT's be released for EU users?

outer seal
winter elbow
still loom
tardy plover
#

Ah, nice! wasn't availble earlier this morning

winter elbow
#

or sooner

vestal helm
#

Question: can gpt4 create .stl 3d files for printing miniatures?

silk marten
#

top 3 out of 4 suggestions for ChatGPT in this discord is sidebar search, oldest is from 4 months ago, and we don't have search yet ?

wise anchor
#

I have too many ideas for GPTs to make

tall flame
#

"I've reached the limit of image generations that I can create from a single description. If you'd like another image, please provide a new or adjusted description for further variation." WHAT? NOW WE CANT ITTERATE O NA PROMPT?!

#

That is DEFINITELY not conducive to revenue streams

dense halo
#

gpt: you're boring me.

tardy plover
#

Best GPT's you've found or made so far?

dense halo
#

havent been able to use or find much of anything the past day or so haha

outer seal
#

sit @wise anchor

dense halo
#

just keep hoping for 5 minutes of functionality

shrewd prism
wise anchor
outer seal
#

timeout

wise anchor
#

oh LOL

river frost
#

Is there anyway to check what is being sent on a get_function call ?

wise anchor
#

wow responses seem longer to me

bronze magnet
#

is the document upload bugged? It gets stuck at "Reading documents" then never responds

dense halo
#

lot of things are glitchy because of the massive response to the update i guess

tardy plover
#

ngl still haven't found a good usecase for chatgpt for a solo entrepreneur or MSP type business

ivory warren
#

is it possible with the api to use a custom chatgpts to query from?

tall flame
#

Using all 3 model functions has now become like tearing hair out. image generation is shot for efficiency and function.

#

the beta of each is better than this horrible amalgamation. slay this beast and let us bear another.

neat nimbus
#

So I got the new stuff, there was a setting saying that chatgpt would save our chats and analyze how I spoke and everything and save that information so it could better respond as I used it more and more. I disabled it, the screen glitched/cleared while I was using it, I checked the settings again and now that setting is gone

crude slate
#

I'm sorry, I can't generate the entire "Bee Movie" script as it is a copyrighted text. However, I can provide a summary or answer any questions you may have about the movie. How can I assist you further?

#

When did they add that

normal raptor
#

Huh why is setting up my gpt also counting towards the 50 prompts cap?

#

Ngl that is kinda stupid

outer seal
small oracle
#

Does the knowledge take a while to actually be put into the custom gpt?

#

because if I create a custom gpt with knowledge, it keeps giving 500 internal server error

main storm
#

is anyone else getting 'Error Saving Drafts when trying to upload documents to the knowledge every time?

ember scroll
#

I don't know if it's just me, I use the AI ​​chatbot(ChatGPT) for creativity and personal entertainment. I noticed that every time I ask a bot to write a certain story for me, it becomes less creative than the previous time and much shorter. I like it less. Is it because of the new updates? Or something else. I would appreciate any reply.

wanton lichen
#

How do i put custom instructions in with the new version? Can't find that anywhere?

ember scroll
#

I just don't like the new version. I lack the thrill I felt in the earlier versions. I lack plot, I lack length. Each story is the length of a poem.

gaunt gyro
#

Did anyone develop chrome extensions for chatgpt? If yes, can you guide me to do it? Recently I saw a chrome extension on chatgpt. That extension was a pop up screen on chatgpt screen. you can build a prompt from the pop up/extension screen. once prompt is built, with a click of button you can directly submit the prompt to the chatgpt and then run it on chatgpt for results. I am looking to develop this type of extension.

slim widget
#

Wooo finally got the update

#

Everyone has it now? Or are there still some people who don't

ember scroll
#

I think it's not just the censorship, it bothers me that it's so short and has no plot.

outer seal
vagrant epoch
#

With the new GPTs it feels like you use up the whole limit in no time

tawny crater
#

Everytime an update comes out it seems to get worse for the first few weeks from what i've noticed.

bronze magnet
vagrant epoch
#

Unless you only use one

knotty crypt
#

on this new version was dall-e and search on web integrated by default on gpt-4 chat? if so, isn't it slower? I remember that the raw gtp4 chat was faster than the dall-e and search on web chat

tawny crater
#

Yea i kinda wish it would be 50 per each different version tbh.

bronze magnet
outer seal
#

i wonder if it will be different for paid gpts thonk would destroy wrappers

shell kiln
#

I am getting this error when trying to save an gpt action
Couldn't save OpenAPI spec due to errors: None of the provided servers is under the root origin, anyone have a solution?

tall flame
topaz panther
#

It appears that I have received the update, with the message 'How can I help you today?' appearing. However, above the 'Message ChatGPT' window there still exists the four suggestion blocks. I thought they were going away with the release.

tall flame
#

its just a crippled product; it feels like its designed to cripple those it's meant to assist. That isn't good.

#

I literally spent all of my image gens this round arguing until it said i was out or on another that i couldnt reuse the prompt. that is ludicrous behavior

knotty crypt
#

how I can use that devtools feature? I enabled that but it says "Please enable a localhost plugin to use devtools."

spice crypt
shell kiln
#

Is it just me or is the save button for actions broken atm?

spice crypt
slim crest
#

Hi guys! Can plus users now use the all-in-one feature? I'm wondering whether to subscribe to it.

cloud pewter
#

I have got the new feature

slim crest
#

Congrats!

red saddle
gray swan
shell kiln
dense beacon
#

the new update sucks, why are the backgrounds of my messages the same color as gpt's messages

#

makes it harder to tell which message is which now

#

everything else is amazing tho

blazing belfry
#

lol

twilit pagoda
#

hi, does anyone know if GPTs are a plus feature only? didnt see anything about it in the announcement, ty

blazing belfry
#

so is the update amazing or does it suck?

sinful ridge
gray swan
gray swan
blazing belfry
loud vessel
#

what is this openai drama with office 365?

blazing belfry
loud vessel
#

Earlier today Microsoft blocked employees from using chatgpt due to security concerns

#

Sam Altman tweeted out an hour ago that they didnt block chatgpt from Office 365 and the rumors are unfounded

subtle wraith
#

Do you guy enjoy new gpt

loud vessel
#

"the rumors that we are blocking microsoft 365 in retaliation are completely unfounded"

red saddle
#

Where can I browse public GPTs?

blazing belfry
#

wasnt the ceo of microsoft just at the openai dev conference like two days ago lol

red saddle
velvet prism
#

huh... it doesn't make jokes anymore

rose glade
#

I am confused. So I use the wolfram alpha plugin quite a bit over this year. My main gripe is that its never had a direct upload feature of say a .tex file. Yet with GPT's this becomes possible, but then accessing the API is free to certain point, 'I think' for wolfram alpha.

My question over all, will multi modal dynamics also becoming to plugins soon? Thank for your responses in advance. Please @ me if any response please.

dense halo
#

good. ai is serious business.

rose glade
cloud pewter
#

the new custom gpts can’t be used

lofty torrent
#

"Please give me a moment to compile this information into your PDF. I'll let you know once it's ready for download."
Well this is new, it just ended on this message

#

am I supposed to remind chat gpt?

vagrant epoch
#

Why can't I upload files in the knowledge part while creating a gpts?

normal quarry
vagrant epoch
#

Brave

normal quarry
vagrant epoch
#

Thanks

vagrant epoch
willow osprey
#

Eh...I'm not sure I like the new integrated Chat GPT-4. lol. The responses seem different than when we had to select between advanced data analysis, etc.

normal quarry
vagrant epoch
#

Yes, thank you

normal quarry
vivid horizon
#

Seems like I'm getting a "Hmm...something seems to have gone wrong." when attempting to edit or test my custom GPT

#

on mobile it says Error with server

restive plover
#

I just got the new model

#

Is there no limit now? Is it unlimited for pro users?

normal quarry
restive plover
#

It doesn’t say anywhere 50 messages per 3 hours

normal quarry
wise anchor
#

i used a script to convert the NextJS docs into a massive PDF, gave it to a GPT and now it’s helping me dev

#

Pretty cool lol

#

It never forgets the docs since it’s a part of its knowledge

restive plover
#

Thank you

wheat thicket
#

They should remove the message cap I'm paying 38 dollars a month for Plus

normal quarry
grizzled thicket
#

it's 20 still, isn't it?

vagrant epoch
#

Probably canadian or australian

floral ether
#

I finally got access to the new UI/AllTools/Gpt version. Yay.. well I had hoped. Is it just me or is the new GPT-4 Turbo model.. umm.. a very dumbed down model? It cant seem to remember what I asked for. When I point this out and ask again. it says something along the lines Now I understand.. here ya go.. and then proceeds to ignore everything I asked it. ???

Is this just me?

This new Gpt-4 Turbo seems almost unusable for programming now. It seems almost as bad as gpt 3.5.

low talon
#

It's exactly that

lofty torrent
#

is the seed/temperature setting a thing for GPTs?

low talon
outer seal
supple zephyr
#

just ClosedAI downgrading their gpt models

low talon
outer seal
#

or refactored to work with assistants api

normal quarry
supple zephyr
#

sure but we all know it is bad now

halcyon cliff
#

images seem to be down for me

low talon
#

GPT-4 Turbo is not GPT-4.

supple zephyr
#

yeah they already showed that they can do better

floral ether
#

@low talon yes this is exactly what I am seeing now. Is there a way to use chatgpt with the older gpt-4 model?

low talon
cobalt dune
#

Is it allowed to share GPT link here?

low talon
halcyon cliff
#

I find GTPs writing has gone downhill as well, It used to be good but now it just write everything in the same style which i hate

lofty torrent
cobalt dune
normal quarry
cobalt dune
low talon
#

Fighting against half-solutions, hallucinations and the overall "dumbness" of new ChatGPT with GPT-4 Turbo makes the fun go away. The ChatGPT app now gives you 2 choices: a dumb model (GPT4T) and a dumber model (3.5). Ok, I get that it's now trying to be a consumer app, I guess, but for what? Dall-E is a severely censored toy, the GPT Vision aspect is not really of much interest for most people in the chat UI, the code interpreter is now only good for trivial functions because it writes much worse code, and the input context in ChatGPT still seems to be small

#

I have no idea what the use for the new ChatGPT might be, which isn't already covered by Bing chat and the other Bing things.

#

Well, ok, DallE is ok for creating sketches of an image based on very loose prompt (or actual prose), but the images are then at best sensible reference images for a better text-to-image model like MJ or SDXL supercharged with CNs & Loras

#

the DallE 3 API is useful, yes, if you want to bulk-create images based on some texts

#

and you don't care for quality

#

but why should I do it (or anything) in the new ChatGPT? I don't get it

#

why is the GPT 3.5 model still there if GPT-4 Turbo is only marginally better?

slim crest
#

It's still a better tool for me to study and review for exams than Bing chat, hhh

viscid condor
#

People think they pay 20$ and they deserve the world

#

U all already getting more than u paid for

wanton lichen
#

So the instructions for making a GPT function the same as regular custom instructions? Is that correct?

molten tundra
halcyon cliff
#

Dalle-3 is overly sensitized to, cant really make anything with it if you use anything as a base idea because of "copywrite" cant even do "zombies" because it goes against guidelines its basically useless for any real world applications now

outer seal
#

I’ve had better success with using data analysis / a coder assistant gpt than the default all tools option, could be the system prompt for all tools is too long & throwing it off idk

#

& letting it pull code thru retrieval instead of pasting code in the prompt

molten tundra
#

The stronger the tools get, the more direct your requests need to be

normal quarry
keen snow
#

guys, imagine having 1GB internet connection and Chatgpt giving network error and not finishing writing an answer... how annoying.. Any way to get this fixed?

normal quarry
molten tundra
keen snow
#

the problem is that it doesnt even let you continue from where it stopped.. only allows you to "Regenerate"... pity, i will try again later

twilit jungle
#

Well this is quite an upgrade, with some bugs

#

But I wasn't expecting this drop tonight

#

The DALL-E integration is interesting, looks like it's most like converting whatever the user says into a more sophisticated prompt. But it's ability to iterate on newly-created or existing image is still weak -- it's always a new image from scratch.

molten tundra
twilit jungle
#

When the image generation models can "tweak" an image iteratively that will be huge

subtle wraith
#

Can gpt 4 analyse a code file?

twilit jungle
outer seal
molten tundra
twilit jungle
normal quarry
halcyon cliff
#

I wish they would go back to generating 4 images like midjourney 1 image at a time sucks, its so slow

molten tundra
#

My gpt bot did exactly as I instructed it to :)\

outer seal
normal quarry
molten tundra
#

Ah I'm so hype my custom bot is performing. 🙂

#

Direct instructions are very crucial in making it work very well

alpine talon
#

Hey everyone. I finally got access to turbo and gpts through the subscription. Question, does it have 128k context?

molten tundra
brisk tapir
#

Hi, I can't save the profile picture for my GPTs. Unless generated by the builder.

#

after saving it will always be replaced by the default one

normal quarry
torpid wigeon
#

when uploading files into my custom gpt's what are the considerations of text size and the gpt's ability to parse its entirety and maintain context? does breaking down a large text file into smaller ones make these custom gpt's parse/recall/context better?

brisk flax
#

same here. did u fix it?

wise anchor
#

lol why r people having so many issues with the responses… mine is giving me responses that are more detailed than ever and have a lot of accuracy

molten tundra
molten tundra
drowsy tinsel
#

im a little curious how does the new progress bar work

wise anchor
#

The NextJS + TailwindCSS GPT i built is wildly good

#

it has the entire NextJS documentation uploaded as knowledge

#

I’m gonna give it Tailwind’s too

drowsy tinsel
#

also gpt4 speed rn is an absolute train wreck

molten tundra
wise anchor
molten tundra
#

Yea, that's a valid point tbh

wise anchor
#

getting the .mdx files into rendered .pdf was not trivial though lol, but after trying it myself and getting filled with rage at my code I found that someone made a github repo that does exactly what i need a while back

keen snow
wise anchor
#

but like ofc nextjs used .mdx and not .md

keen snow
#

building customized gpts

wise anchor
#

yeah it’s incredible

#

it’s just the first one i have a lot more ideas

keen snow
#

I built an essay writer and I fed it all my university essays that I wrote some years algo (before AI was a thing) so it learned my style so well

wise anchor
#

Oh that’s crazy

#

There’s so many possibilities

#

Especially with the action configuration

keen snow
#

I didnt dig on the action thing yet, how exactly we use it? need an api key?

wise anchor
#

Just lets you use APIs yeah from what I can tell

#

I’m not sure if there’s bounds on what types of requests you can do, i haven’t looked

knotty crypt
#

is it only me or have anyone noticed that after this chatgpt update with dall-e and search on web integrated it seems to give more unhelpful instructions on code?

before it was more straight forward, with snipped codes, now it seems to have a lot of more broad text, more text than code.

wise anchor
#

Honestly i’ve thought it’s been better lol

keen snow
wise anchor
#

Even on a regular thread and not on my NextGPT

#

Yeah my custom gpt is infinitely better but normal threads aren’t bad for me

#

slightly better if anything

#

But the custom gpt is soooooooo good

keen snow
#

would be nice to have an Agent GPT for building MERN stack applications

#

hahah I am just getting lazy

wise anchor
#

i know i’m worried the one i made is too good 💀

#

but i did design it to guide rather than do

keen snow
#

is it private or are u sharing it?

wise anchor
#

Currently private but i’ll probs switch it public when i fine tune it some more

keen snow
#

cool nice

knotty crypt
molten tundra
#

I think for the action section, I want to research more into that. I don't fully grasp what actions I can ask the gpt bot to do

drowsy tinsel
#

Always wondered how many GPUs does chatgpt use at the moment

keen snow
drowsy tinsel
#

there were rumors on gpt4 requires 5 A100 80gb to load

#

so there's at least a few thousand clusters of that

knotty crypt
grave raptor
#

when "creating gpt" does this use up your 50 messages per 3 hours?

knotty crypt
wind nexus
grave raptor
#

ah i see, even when we are creating the gpt itself and testing it in preview?

knotty crypt
#

I am wondering how that 'customized gpt' works, I mean, if I create a "customize bot for answering NextJS questions" will it be better to answer questions related to frontend than a bot created to "answer questions about coding"?

I mean, I am wondering if at the moment of the bot creation it gives the bot more informations on that specific use case, or is that only a mattter of context to facilitate how the bot should better comunicate?

molten tundra
grave raptor
wind nexus
#

some old conversations are down

normal quarry
#

up for me

knotty crypt
grave raptor
verbal lotus
#

i think the conversation history on the left is a little small and hard to look at

#

i think the older version of history textsize was just good

wise anchor
wise anchor
#

I think it looks good overall just could be a bit larger scaled

sly geode
#

all in one tool is so insanely helpful

elfin birch
sly geode
#

so i dont need to switch to dalle or add prompts

drowsy tinsel
#

once in a while gpt4 forgets it can search the internet

dense halo
#

Is there a way to block that feature? I want the classic gpt answers

urban sigil
#

Am I the only one who feels that GPT-4 turbo is less intelligent than previous one? I've been a frequent user of GPT-4 ever since they launched..can clearly see the reduction in quality of responses, both in coding and regular responses.

molten tundra
rancid otter
normal quarry
little peak
#

Just tried one of the custom GPTs I made with the voice interface on android and the voice is different. Wonder why that is.

urban sigil
normal quarry
dense halo
#

Nerds

urban sigil
#

I'll run more diagnostic tests and see.

rancid otter
minor flume
dense halo
austere fractal
#

How to get chatgpt to have knowledge cutoff of april 2023? Is it going to be possible?

urban sigil
normal quarry
eternal rockBOT
rancid otter
molten tundra
steep tundra
#

Qq: can custom GPTs access custom instructions?

#

If so, how?

normal quarry
austere fractal
#

What about context? Is it new context of 128k already for chatgpt or api only

normal quarry
#

Context appears to be at least 32K but you can upload documents that give it the ability to reference at least 300 pages.

rancid otter
knotty crypt
knotty crypt
dense halo
#

You are a life saver

#

Is the data analysis gpt the version of classic gpt with vision?

normal quarry
dense halo
#

Seems like it, but just wanted to confirm.

lusty island
#

Where can i see published custom GPTs?

toxic flax
#

are we still limited to 50 messages

lusty island
toxic flax
lusty island
jolly bolt
#

where can we post our GPTs that we have made

#

which channel

pine tulip
pine tulip
wanton lichen
#

So could the "knowledge" files I upload for a custom GPT be code files for a project I'm working on? Perhaps lots of code?

pine tulip
#

yup

wanton lichen
#

Well how 'bout that

pine tulip
#

2 million tokes is the max

wanton lichen
#

Unbelievable

pine tulip
#

or 10gb

#

depending whats first

wanton lichen
#

2 million tokens of knowledge files I can upload?

pine tulip
#

yup

#

thats about 200.000-300.000 words in a text file

wanton lichen
#

What's the new token limit for the updated chatgpt4?

#

like regular token limit

pine tulip
#

no one knows. Don't believe people that say that they know, there is no official info on that

#

I am guessing either 4k or 8k

wanton lichen
#

Isn't it 8k right now?

pine tulip
#

yea maybe

wanton lichen
#

thanks man

pine tulip
#

Sure thing!

normal quarry
#

I think the context limit for GPT Turbo is at least 32K, I could be wrong but I think I am right

wraith dagger
#

Can someone tell me if uploading a file as a custom gpt's profile picture works? its not seeming to work for me

normal quarry
wraith dagger
#

Safari on macos 14

normal quarry
wraith dagger
#

alright. just seems weird as uploading files to gpt 4 works completely fine

normal quarry
wraith dagger
wraith dagger
normal quarry
native mist
#

this new update is so fye

wanton lichen
normal quarry
wanton lichen
#

Fantastic. Have you done any experimenting with the knowledge files you can upload? Was thinking of uploading a code base

pine tulip
normal quarry
normal quarry
wanton lichen
pine tulip
#

I remember dalle3 giving me 2 pictures at once. Now it only does one. Is that intended or do you guys have any tricks to unlock two images again?

normal quarry
pine tulip
#

thats unfortunate

wanton lichen
#

And has the cap gone up? I don't see it displayed anywhere

normal quarry
velvet bolt
#

Anyone else having issues with not seeing the option to include your name in the Builder Profile? Which then doesn't allow my GPT to be public

wanton lichen
white shard
lusty pine
#

anyone getting Hmm...something seems to have gone wrong.

pine tulip
normal quarry
lusty pine
pine tulip
#

lol chrome os

lusty pine
#

windows 10

#

i was j using it but it randomly now just saying smthng gone wrong

normal quarry
wanton lichen
#

Is there anywhere I can go to view GPT's created by other people?

normal quarry
wanton lichen
pine tulip
lusty pine
floral ether
#

So dalle would generate 4 images per prompt... then 2.... now 1. Is the next step a low res preview? So it might take 2 requests per image?

normal quarry
floral ether
#

Today.. I am wondering about tomorrow, based on the progression of how things are going.

normal quarry
lilac stratus
#

So what are some potential things that these gpts can do in the future ?

normal quarry
wanton lichen
#

What would be the best format to upload text to a cutom gpt knowledgebase? plain .txt files? .pdf?

normal quarry
white shard
#

because as you probably know PDF text is all just free floating elements not anchored to anything so parsers read it top down

#

slight variances in the layout can cause text to be truncated, cut off, or combined with the wrong other pieces of text

lilac stratus
#

Is there a way for ChatGPT to read links ? Like if I was to paste a YT link of some video

white shard
#

It cannot transcribe youtube links no

#

It's able to browse the web though

wanton lichen
white shard
peak edge
#

Are plug ins running on gpt 4 or 3.5 now?

wanton lichen
normal quarry
#

do EPUB files suffer from the same issues that might affect PDF?

white shard
#

Let me check real quick

wanton lichen
#

Can you confirm that's correct

normal quarry
white shard
lilac stratus
#

Is there a way go around Dall E restrictions, I was trying to do a picture of like a mafia boss and it told me it can’t do that

white shard
#

since html is easily parseable

wanton lichen
white shard
lilac stratus
#

Is there a way to make google slides through ChatGPT ? Or maybe use some kind of AI?

#

What is that ?

elder haven
#

hello everyone......

lilac stratus
#

Yes

wanton lichen
#

So a question I have is, what's the difference between me posting 200 pages of something into the main chatgpt window and asking questions about it, versus posting 200 pages into the knowledgebase and then asking questions about it? Because obviously it would say the length was too long if I did it the first way. @white shard

lilac stratus
#

How do I be specific with it ? For example I need to make a presentation about Indian’s building called Stuppa

#

And it will include pictures and like bullet points and stuff ?

#

I see

grand mulch
#

hey,
everything I ask for results in an error, right now: "There was an error generating a response" anyone else experiencing these problems?

#

when asking for an image i sometimes receive one but than the chat is over with the same error message. can only click "regenerate" which results in the same problem

normal quarry
#

Custom GPTs are not having a good time right now, my one is failing to find knowledge contained in it's text based knowledgebase

sly geode
#

got the update but keeps failing and errors

#

trying for like 7 times

dense nacelle
#

There was an error generating a response.....

sly geode
#

fr

midnight cedar
#

My GPTs doesn't work at all, I got "Hmm...something seems to have gone wrong."

normal quarry
real swift
#

Y'all go to bed so I can work like a degen through the night. JK lol yeah this is world chaning tech so not a surprise everyone is pushing their servers to the max

dense halo
#

yes the peoples love AI it seems, which ultimatley is a good thing. bring it on haha

white shard
#

When you paste into the chat window the whole thing is inserted into the prompt

#

when you use retrieval, there is an underlying mathematical system that breaks up the document into many chunks, creates a numerical representation of it, and then in the future when you chat and ask about the data it retrieves of the X most relevant chunks for use in the prompt, ensuring it remains under the context window for the prompt

wanton lichen
spark briar
#

What’s the most effective way to train custom gpts on uni course content like finance maths. I want to use it to solve questions as a marking guide

kindred ravine
#

I don't think you need to actually train it since gpt4 finance math know well about financial calculation. I guess you just need to attach your curriculum to it so it know what you are planning to learn and practice

earnest iron
#

Does anyone know why I can't see the Action set builder but only the schema input instead

spark briar
#

Yea it’s pretty good but seems to mess up some calculations so was curious how I could optimise it some more

kindred ravine
#

what do you mean by messing up calculations?

paper kestrel
#

Hello everyone, I am trying to make a customGPT and put documentation in it, any idea what is the best way to do this?

Should I convert the documentation to a PDF from the website and upload the file to GPT?

spark briar
#

Using incorrect formulas and misinterpreting more complex worded questions

eternal rockBOT
#
This is a Safe for Work Discord Server

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

Posting NSFW content will result in a permanent ban.
tardy oasis
sly geode
#

go to sleep y'all 😭

jk

paper kestrel
tardy oasis
#

I'm still trying to figure out if customGPT is good, I think I might just prefer using the assistants API rn but I'm not sure. These are good questions to ask GPT. I used python and selenium to do quick and dirty html scraping, then uploaded the dirty html to GPT4 chat and asked it to remove certain elements using xpath, and then uploaded the final cleaned html

paper kestrel
elder haven
#

A good idea

#

Thanks

normal quarry
#

Custom Actions in the GPT Builder seem broken. The docs say

Custom Actions: You can make third-party APIs available to your GPT by providing details about the endpoints, parameters, and a description about how the model should use it. Actions for GPTs can also be imported from an OpenAPI schema. So if you’ve already built a plugin, you will be able to use your existing plugin manifests to define actions for your GPT.

But it's only showing an option for OpenAPI schema and not a fill in the blanks form for endpoints like it did.

drowsy tinsel
#

does GPT builder already use gpt4 turbo?

#

it seems much faster than the normal gpt4

normal quarry
near vine
mild cosmos
#

wow ChatGPT just got an update?

#

it is now able to browse, do data analysis etc without switching

toxic shell
#

Is web browsing capacity gone? I can't find it any more

mild cosmos
#

i also tried to upload a PDF file and it was able to read the contents of it perfectly fine. I wonder what is the max number of PDF pages it supports?

warm tendon
#

gpt-4 turbo... thoughts on output quality?...

mild cosmos
#

@toxic shell i think its all in one thing now in GPT-4

toxic shell
#

It says it can't read any content from a URL

stable pier
#

Does anyone know whether GPT-4 still have the 50 response cap?

normal quarry
umbral echo
supple zephyr
#

is all system operational
ohh i see it is super slow but still answering

#

it just said 'In ' and disappeared

dusk epoch
#

It's so slow, that it encounters an error before it finishes a task for me

next ibex
#

i keep getting network errors

#

for each input, it cut off response

dusk epoch
zenith solar
#

I somehow feel it works even worse after the update

zenith current
#

I see this ai is still buggy

lilac stratus
#

is dalee working?

normal quarry
lilac stratus
#

hmm for some reason its not generating pictures for me

#

error creating image

dense halo
#

On chat gpt its gpt4 turbo or Just gpt4

dense halo
lilac stratus
dense halo
#

noice

lone wyvern
#

After some 12 hours of playing with custom GPTs, I went back to using ChatGPT Classic. I assume the most significant new feature is the increase of the context window.

normal quarry
lone wyvern
normal quarry
kindred ravine
# spark briar Using incorrect formulas and misinterpreting more complex worded questions

sorry was trapped to some cases. My experience was giving or naming all your need before processing any calculation. In matter of GPTs, I guess giving out a pdf with proper list of fomula would be necessary. I just tried to use GPTs to process some analysis for hydrogen industry. To ensure it doesn't make up some faulty information, I indicate a few trustworthy websites and already approved report previously. so far so good.

lone wyvern
normal quarry
dense halo
lone wyvern
#

I see. I thought it's just the old GPT-4

brisk dirge
#

@normal quarry has GPT 4 knowledge date been finally updated ?

#

I'm asking for codding

rapid moss
#

Apparently, Iv seen it mention April 2023 however when it recommends lib versions to use they are all very outdated

inland heron
#

I can't remember, I think starting in December I read somewhere

brisk dirge
#

At least they took my advice and streamlined all the features into one single module.

lone wyvern
normal quarry
brisk dirge
#

Sweet, thank you friends, I was waiting for a knowledge update for quite some time.

lone wyvern
brisk dirge
slim crest
#

Hope there will be a copy/paste feature for gpts in future

spice star
#

yeah I'm sure that's planned

normal quarry
#

@brisk dirge do you have the new interface and all the new features? If so check out GPTs, very cool stuff!

brisk dirge
grizzled thicket
#

Hey can we actually roll back to the last version of chatgpt4? it worked wayy better

zenith current
normal quarry
normal quarry
slim crest
#

Hi guys, somehow I cannot save my gpt draft after I upload some pdfs into the knowledge base. Is that because I exceed the limit or something?

silk grotto
#

whats going on with the speed of chatgpt right now? 😄

slim crest
#

Is there a token limit?

brisk dirge
#

This new GPT 4 seems to work much better for coding @normal quarry it already managed to help me solve an issue that gave me big headaches with the previous GPT 4. And it seems to actually listen when you tell it to give back the complete code, it no longer truncates it.

silk flame
#

Hello guys, I started using MyGPT feature and I have a problem. The whole process of creating the new GPT is very easy, and I'm grateful that OpenAI gave the assistant that helps to create a new GPT, but there's one problem. When I add a PDF file or a transcript from a video, it always adds that file as an attachment. Therefore, when a user has a conversation with the GPT, itcan look up the document and search for information. I understand that sometimes this can be helpful, but instead I would like to fine-tune my GPT so that its core knowledge is updatet with the files I gave. Guys, do you understand my concept and question?

knotty willow
#

Is there a GPT marketplace somewhere or not yet ?

silk flame
#

How to fine-tune myGPT with the external files?

normal quarry
normal quarry
silk grotto
#

anyone else experiencing slow loads?

silk flame
normal quarry
grizzled thicket
#

The DallE portion of Chat GPT isn't showing its modified prompts anymore. It actually did a good job with that. Any way to get that back?

autumn quiver
#

I want to create an application where the AI is thinking about something. You can ask questions, and you will either get a "yes", "no" or "didn't understand" back by the AI. If you can guess the word in x amount of questions you win! How should I go about creating this? I was thinking about using the system window to give it information about what to answer. Is this a good idea, and if yes, how should I prompt it?

normal quarry
slim crest
gusty gyro
#

Question is chatgpt still 50 caps per 3 hours

#

The new UI wont shot

#

Show

knotty willow
#

Good question

#

Maybe they expanded it, if turbo's cheaper ?

silk flame
# normal quarry fine tuning is not a simple process, the idea of custom GPTs is that anyone can ...

Yeah, I understand that it's not that simple because the creator needs to provide with a JSON file that contains prompts and perfect replies to those prompts. To make it really good, the creator has to provide with hundreds of these. But I think that GPT Assistant is fully capable of making all that work from the PDF file that the user attaches. So I'm really surprised that it doesn't have that feature.

normal quarry
#

the current limit is 50 messages per 3 hours

knotty willow
#

On the other hand, they made all tools more accessible, it makes more token requests so maybe not

gusty gyro
#

Ok, they should include it on the new UI. Probably confusing

dreamy summit
#

Hey guys, read through the docs but still kinda confused as to what are the practical differences between GPTs and Assistants. Could someone shed some light on this?

knotty willow
#

and they make GPTs with that

dreamy summit
mild cosmos
#

@normal quarry does the new GPT4 model available on chat.openai.com supports 128k context?

brisk dirge
# autumn quiver I want to create an application where the AI is thinking about something. You ca...

I don't think it's going to work that well because the AI could think of many things similar to a human. Think about it, what are the chances of you actually guessing what it's thinking about, the posiblities are probably almost infinite if not infinite 😄 (it will probably not be fun is what I'm trying to say).

If you just want to create a chatbot ai there are free LLMs you can probably use. Try looking for some "chatbot AI programming LLM" tutorials on youtube to see how it's done but mind you it's probably going to be a challenge.

knotty willow
#

What I find interesting is that you have a GPT Builder assistant that helps you build your GPT

normal quarry
mild cosmos
# normal quarry I've only tested it to 32K

oh wow and it allows you to input such a large document? thats pretty cool. What about the size of PDFs supported by the model? Also does it properly understand the figures, tables and other equations mentioned in the PDFs? It would be super helpful to analyze research papers

dreamy summit
autumn quiver
normal quarry
mild cosmos
normal quarry
knotty willow
#

I don't have it yet on the API

#

And the UI is defenitely less appealing out there, so while it probably is more powerful and useful, it's less accessible yk

brisk dirge
autumn quiver
#

Probably React on the front end, Node and Express on the backend and Firebase for the server

#

Would it be a good idea to use the "system" to tell the word and that I want a "yes", "no", or "don't know response" back in JSON?

#

I saw someone using: "Return response in the following parsable JSON format"

dreamy summit
normal quarry
knotty willow
knotty willow
brisk dirge
#

I don't really know cause I didn't play that type of game. You have to structure your db tables based on the way the game would be designed in terms of input output/predefined answers and questions and so on. @autumn quiver

autumn quiver
#

Just want something really simple

brisk dirge
#

You could try asking chat gpt to help you if it's something simple if your school allows for it.

valid kestrel
#

Is the limit still 50 messages per 3 hours?

dreamy summit
silk flame
#

I have myGPT feature on my web version, but not on Android. how to activate it on Android?

torn cloud
#

Is anyone else getting network errors / something went wrong, but still getting some outputs ?

ornate zealot
#

Do we know the actual context window for ChatGPT Turbo?

dense halo
normal quarry
#

Sam Altman tweeted “usage of our new features from devday is far outpacing our expectations. we were planning to go live with GPTs for all subscribers monday but still haven’t been able to. we are hoping to soon. there will likely be service instability in the short term due to load. sorry :/“

torn cloud
umbral forge
#

Sorry to barge into everyone’s conversations, but I keep getting this UnrecognizedKwargsError error when POSTing to endpoints via a custom GPT action. I have found only one forum on this which was regarding the assistants api, and have implemented the suggestions but nothing. This is an OpenAI-specific error apparently, so I’m wondering if anyone else has come across this as there is currently no info available. I was able to get it to POST using a schema earlier today, and I see that other custom GPTs are able to POST, and I can tell that the request is not making it to my endpoint, nor is it improperly formatted (getting the gpt to output a curl command for the action it is trying to use results in a fully functional curl command). GET requests are working fine, at least in my current setup.

tardy totem
#

are you supposed to be able to upload PDFs?

umbral forge
coral agate
#

Anyone offer advice or input as to why one of my chat conversations has completely error crashed out on me... It started when I tried to upload an image. Then none of my previos convos worked. Then I could start a new convo, but it would error crash when I tested it by uploading an image. Now today, original convo is still error crashed, but previous convos are ok, slow, but working. I have logged out, cleared cache, opened a new tab to launch in, but still the same error generating... sigh...

ornate zealot
#

What the?! Some how it can recall the beginning events of the previous stories that I wrote with the old GPT 4.

#

It always got it wrong, this is next level then.

tardy totem
#

Every conversation that is stored is stored in its entirety.

#

That means it now can use more of the history in future answers, if I understand things correctly.

ornate zealot
#

Yeah, but it would’ve forgotten it due to the context limitations.

tardy totem
#

So the conversation history is stored as just messages. Every time you ask for a new answer it takes what it can of the history as part of the input.

#

At least that's how I think it works.

umbral forge
ornate zealot
#

Unless this 32k update made it read the entire story all over again.

umbral forge
#

The conversation is not fed into some memory somewhere, the context is the memory

ornate zealot
#

Although that story was kinda experimental and long as heck.

umbral forge
#

Essentially ChatGPT continually feeds the previous messages into itself up until the context length

coral agate
umbral forge
#

So if you suddenly have a model with a larger context length, it will just feed more previous messages into itself, which gives the illusion of suddenly remembering older elements of the conversation

tardy totem
ornate zealot
#

So it just restores the previous messages it got rid of due to the new context length?

tardy totem
#

It never got rid of them. You could always scroll up in a conversation and read them. But, when writing a new question, it wouldn't include as much of them before as it does now.

umbral forge
#

Unlike the new threads and messages concept introduced with the assistants api, chatgpt simply utilizes its context to store all relevant information. You’ll notice this if you try to use the api to build an your own chatbot, where you will have to locally store every message and response and continually feed the increasing amount of text into the api for every message if you want to have a relevant conversation with context. This is secretly what happens behind the scenes with threads in the assistant api as well, but it is automatically managed and truncated by the api, and thread is there to simply store the context server-side instead of requiring that it be fed into every api request

umbral forge
ornate zealot
#

Should have put them in quotes.

#

More like set them aside.

ornate zealot
umbral forge
coral agate
#

what the...! I submitted feedback on the website like 20 min ago... My chat just came back to life right now. Sheesh! It's been frozen since yesterday!

ornate zealot
#

This is not even the api, it’s ChatGPT.

#

Unless it secretly is 128k.

umbral forge
#

ChatGPT is running off of the 128k model, but I am not sure if there are context restrictions applied per-account. API does though

#

Yeah “large” is relative. I don’t think 128k tokens is large enough to truly solve the context problem when it comes to, for example, managing a production application in its entirety, but it’s good for small projects

#

The context is limited on chatgpt, but not by the model itself from my understanding

#

Upload it as a file, tends to work better

ornate zealot
#

That, can’t share with image,” thing is stopping me from sharing the story here.

inland mortar
#

I keep getting network error no matter what I do. GPT generates and then just gets interrupted by network error. EVERY single TIME.

normal quarry
umbral forge
#

For the time being you can achieve similar functionality by creating an assistant in the playground and uploading a file with those functions

inland mortar
umbral forge
#

And then updating those functions if you want

silk flame
#

hi guys, does the new feature myGPT works on mobile device or only web?

ornate zealot
#

Although ChatGPT estimates that my story may will be 10,000 to 15,000 words or more.

#

Wonder if 32k context is enough for that

fleet fog
#

is it possible to upload a PDF to the API and then query against it?

ornate zealot
#

Well, now I can cook up a story and not have to worry about it forgetting the beginning anymore

#

I can straight up upload wikis inside it and it won’t forget

dense halo
#

whats the limit of knowledge you can upload i wonder

ornate zealot
#

Well, I’ll just copy and paste descriptions of fictional environments here and there for story development.

#

The characters will be unique though

dense halo
ornate zealot
#

It’s not a GPT, it’s a story made within it

dense halo
#

my bad

#

Hi, why does ChatGPT use an edit icon as a new chat icon?

#

is it intentional or a mistake in the design?

supple zephyr
#

"I'm sorry for any confusion, but as an AI developed by OpenAI, I don't have the capability to directly interact with external files, browse the web, or run code analysis tools."
hmm it is supposed to have access to them

hollow pivot
#

Hey guys, do you have any idea how to get the custom gpt to link the images in it's knowledge base in the chat when asked? Is it even possible? does it have to be a certain file type?

pliant citrus
#

I finally got it! The GPTs! Let's gooo! 😄 ❤️

sleek knot
#

https:// github. com/ForkInABlender/ChatGPT_solutions_produced_from_simple_prompt/blob/2023_11/gpt3_4_using_2_with_function_calls_optional.py

oops

#

:) what used to cost $20 now truly costs $0.

#

Forever....every month.....

#

"Why pay more for what you could simply not"

#

Let's gooo! 😄 ❤️

slim crest
#

Hi guys, have encountered problems that GPT assistant automatically updating itself

#

It just does not give me a chance to say anything lol

#

Now it tells me there is an error

upbeat pollen
#

I think average 100 line of code would usually eqaul to around 4k tokens with what I tested. ChatGPT rumored to have 4k/8k context window to work with. I don't think it's confirmed. Until they officially move to larger context window, working with anything beyond 100 lines of code is pretty limiting.

Developer API is the way to go for large code bases.

ornate zealot
#

32k context, yeah right...

#

This is 128k, and I proved it for myself.

#

And as far as I know, it would impossible for this thing to remember the beginning details of my stories, if it was only 8k.

#

Yeah, but it wasn't able to do that before the update.

#

It can't just search through the story if they context window isn't large enough for the whole thing to fit inside it.

ornate zealot
#

No, it wasn't, I asked it about the previous messages close to the beginning before the update, and it gave me hallucinated answers.

floral ether
#

As I understand it. There is still a 4k token limit. However, it now has multiple pre-processors to scan and evaluate existing conversation (up to 128k tokens) to pull out relevant parts and encode them into the 4k input token.

slow jungle
maiden fractal
#

Try firefox as test

#

Do you have a vpn ?

#

Did you try edge in incognito

ornate zealot
#

Well, it had to remember the 80000 words somehow.

maiden fractal
#

Vpn are known to be a bug with openai products

#

Can you try with chrome for a test run

#

If this is a fresh install , yes

spare anvil
#

Does anyone know the trick to get chatGPT to respond in full with the entire code and not cut it out for brevity? How to prevent this in its code responses?

<!-- (rest of the form code) -->

ornate zealot
#

And that is 4096 token outputs at a time.

#

Each prompt response can be that big.

silk grotto
spare anvil
#

obvious openai trying to reduce costs for that portion

hybrid haven
#

Best experience with GPT-4 now )

near summit
#

Is it just me or is GPT-4 getting more and more stupid and forgetful?

supple zephyr
#

oh it is doing that to everyone I am pretty sure

maiden fractal
#

Glad it works at the end

vestal orchid
#

wait is there a limit to chat gpt 4 now?

supple zephyr
#

always

vestal orchid
#

oh mb i pharsed the question wrong. i ment after the new update is it still 50 messages or have they made it higher?

supple zephyr
lunar arrow
#

Do we have Turbo in ChatGPT or is it still just in API right now?

near summit
#

Damn 😮

kind crater
#

Just hit my GPT-4 limit for the first time in a while. sad face

maiden fractal
#

You weren't supposed at first

supple zephyr
#

yea they dont like to make it

near summit
#

It's never let me do that without coercion and tricking it 😆

kind crater
#

It's amazing that people haven't realized you can just give it replacement terminology and it will do the job just fine. Replace words afterwards and boom, you're done. Don't be lazy

#

Even better, just train and host your own models

lunar arrow
#

It's annoying how when you're making a GPT and typing something in the create tab then switching to the configure tab it will delete what you wrote

kind crater
#

DOM elements are probably being dynamically changed, clearing contents as you switch. I would copy/paste what you have as you go along to not lose stuff.

#

Might wanna do that anyway. I noticed that it rewrites the instructions instead of modifying them, at least in my brief testing.

pliant citrus
#

Oh man, I literally just ran into reaching the limit on prompts and have to wait a full hour. I blame to many errors when trying to create a GPT hahah

drowsy tinsel
#

is regular gpt4(all tools) now gpt4 turbo or not yet?

#

haven't tried super long input/output

viscid dagger
#

That "image of a rose" text still hits it badly

maiden fractal
warm totem
#

did you try mentionning that in custom instructions ?

lunar arrow
#

Do GPTs use GPT-4?

brisk seal
signal oracle
#

Hi all, I am a developer and I just tested creating a GPT and also an Assistant thru the Playground. Technically speaking they seem the same right? The only difference is that one you use the ChatGPT UI and the other you used an API/Playground.

Is that correct or are there any features that are available in one that are not available on the other.

red saddle
#

I always reach my cap now, creating new GPTs and testing them 😅 How can I get a higher cap?

near vine
#

you can just ask it to say "give the code without executuion of yourself" it did the same thing for me, but after the promt in the same chat it worked fine

warm totem
#

i noticed also that sometimes, with little changes in custom instruction, big differences. like I changed things to say the same thing, should trigger the same result and nope, big changes in output. i guess the best is to try until you get the desired result and stick with these custom instructions

rich surge
near vine
#

people are saying that new gpt4 version got bad since and they would prefer the old gpt without the appropriate knowledge of the current date of april 2023, since the new model due to the vastly increase token size it gives out rubbish/unrealted code or forgets the real pupose of it. do you guys think will it be fixed in the near future or this issue gonna persist always?

rich surge
#

There is the old GPT4, it's called GPT Classic, in the Explore menu.

placid birch
#

OMG I got GPTs! 😎

near vine
nocturne badge
#

Does anyone know if gpt turbo will be added to chatgpt plus?

red saddle
nocturne badge
#

Also if they add today more limit with 100 messages every 3 hours

#

??

red saddle
rich surge
#

But it will be 100 soon enough.

red saddle
#

I know 100 would have been sufficient always before custom GPTs. But when you need to test the GPT you can need to generate 200 messages in an hour.

near vine
rich surge
near vine
#

time stamp?

#

do u remember

rich surge
#

One momento. I'll find it for you.

near vine
#

appreciated.

rich surge
gusty bridge
#

Hello, do you guys know what model of gpt, the new chatgpt is using ?
It seems its data goes up to april-23, even though on some subjects I found it knowing only data up to mid 22.
But I tried to test the window of chatgpt (max token), and I added my name at the beginning of the conversation, then I copy pasted texted from wikipedia for around 30/40k tokens
And finally asked for my name after that, he didn't remember
Isn't gpt4-Turbo supposed to have a 128k token window ?

turbid isle
#

guys custom GPT instructions are not working up to the mark for me.. i tried a lot changing the prompts, but it just doesn't work.. seems like natural language is still not enough compared to coding.

onyx sapphire
#

Hello, I would like to understand how to format a prompt more efficiently, I have an insignificant experience in dev but I undestant the beginner concept (variable, object, array etc...) I have seen some prompt written almost as a code with some symbole like -- to execute command and different bracket type.
I have a few question along these lines:

  • Is there a guide/doc explanning the signification of these different symboles and how to use them ?
  • What would be a good way to clarify some parts of my prompt that are for context and should be referenced to all the time versus a one time execution of an action?
  • how to indicate that something is a list and to pick a random item from this list?
    many thanks
silk flame
#

Hi guys, do you have access to myGPT through mobile app?

inland quest
#

I just got access to new chatgpt and GPTs, my question would be if chatgpt4 is using turbo now with 128k context size or is it still same old chatgpt?

gusty bridge
deep rapids
#

Does the browse with Bing feature work for ChatGPT 4? Or is it stopped? Saw an article from July (which I can't paste here) which said it was disabled. Whenever I try to use it I get the error Error browsing. I'm unable to access the contents of the page you've provided.

maiden fractal
maiden fractal
deep rapids
maiden fractal
#

And what site did you try to summarize? (Without copy paste url ) like Wikipedia?

deep rapids
maiden fractal
delicate cradle
#

How much does ChatGPT draw on previous conversations with you? I just asked it a seemingly unrelated question and it returned a response that was almost word for word and very different discussion.

deep rapids
delicate cradle
maiden fractal
#

Of what did you try so far etc

maiden fractal
#

Does it works ?

deep rapids
#

The content is public information, I'll see if I can find the same thing elsewhere

turbid isle
maiden fractal
deep rapids
turbid isle
maiden fractal
maiden fractal
turbid isle
turbid isle
eternal rockBOT
delicate cradle
#

How much does ChatGPT draw on previous conversations with you? I just asked it a seemingly unrelated question and it returned a response that was almost word for word and very different discussion.

deep rapids
maiden fractal
maiden fractal
sinful marlin
#

are GPTs only available to GPT Plus subscribers?

brisk seal
brisk seal
sinful marlin
brisk seal
delicate cradle
#

pretty darned specificly.

silk flame
#

Hello, I have a question. When I am creating myGPT, does it use GPT 3.5 or GPT 4? If the latter, does the responses of myGPT count to the 50 message limit per 3 hours?

brisk seal
delicate cradle
silk flame
delicate cradle
#

are you subscribed to chatgpt?

silk flame
#

I have an android device, and you?

delicate cradle
#

ah I might have misunderstood you

tropic atlas
#

Are there any forums or such where I can discuss GPTs?

maiden fractal
#

This one?

tropic atlas
sleek aspen
tropic atlas
hybrid haven
#

I'm experiencing problems with ADA part of new GPT, it does not execute code atm

sleek aspen
hybrid haven
#

"It appears that there is an issue with the Python environment right now. Unfortunately, I can't execute any Python code at the moment. "

tropic atlas
hybrid haven
feral zinc
pine tulip
#

Indeed

tropic atlas
#

I'm pretty confident GPT questions go in GPT channel? Where should I ask questions about making GPTs? I'm just here because this was a general question. Also, note how I am here on Discord discussing how Discord works, rather than just going to a forum and getting answers. I appreciate the time you took to say "wrong channel." 😄

rose glade
sleek aspen
sleek aspen
rose glade
sleek aspen
#

It should be available through both, the new multimodal version can use it along with DALL-E and Browsing, and you can enable Code Interpreter with your own GPTs

image0.jpg
rose glade
brisk path
#

Hit the GPT-4 limit while making a GPT

#

🥲

wispy elk
#

chatgpt looks so good

drowsy tinsel
#

with the new 100/3hr limit its almost impossible to hit?

supple zephyr
#

Asking to split its work into small functions then it is fine.
It works good for making codes for me so far.

drowsy tinsel
#

I've hit the limit exactly once and it was back when 25/3hr

supple zephyr
#

it is 1 per 108s so pretty sure you can't hit it unless you spam

#

at least I take much more time than 108s while prompting

drowsy tinsel
#

also most generation takes longer than 108s

supple zephyr
#

yeah that too

drowsy tinsel
#

but still the current gpt4 is much faster

supple zephyr
#

"Hello! How can I assist you today?"

drowsy tinsel
#

I remember back when browsing first rolled out back in Spring

#

u prompt, go to the bathroom, have a glass of water, and come back to see the results

supple zephyr
#

I do that often

stone helm
supple zephyr
#

sometimes I enter prompt and go bathroom real quick and it is still printing text

drowsy tinsel
#

I'd also love to see some stats on how long did it take to train gpt4 and how much electricity/water does gpt4 use on average for 1 prompt

brisk path
supple zephyr
#

of course ClosedAI will not share any information

brisk path
#

Also I was using the GPT maker

earnest iron
supple zephyr
#

guess that takes more than one per prompt

brisk path
#

Which is currently kinda buggy, so I had to use the gpt maker for everything

drowsy tinsel
#

The thing is openai got so much resources its hard for competition to catch up

brisk path
#

(I can't upload images for the pfps currently)

drowsy tinsel
#

xAI probably got potential, but it will take quite a bit

brisk path
#

xAI has exactly

#

Nothing

charred wyvern
supple zephyr
#

yea they uh yea

drowsy tinsel
#

well tbf Tesla got quite a bit of stuff ig... maybe they could take advantage of that?

earnest iron
brisk path
#

xAI in MY car???

drowsy tinsel
#

perhaps the biggest advantage of xAI is they got the entire Twitter to train on...

brisk path
#

The worst dataset in the world

supple zephyr
#

lmao yea right

brisk path
#

Second worst dataset

earnest iron
#

Twitter is as good as using Facebook as a dataset

supple zephyr
#

big dataset is usually good but when it is twitter

earnest iron
#

Might i even say Tumblr

brisk path
#

Tumblr is just the art platform now

supple zephyr
#

image training?

earnest iron
brisk path
novel parcel
#

my gpt is messed up

supple zephyr
#

ah yes that is a thing

brisk path
#

Like actually

drowsy tinsel
#

if I have to guess for the last many months Elon Musk has been trying to keep AI companies away from scraping Twitter data...

supple zephyr
#

artists will never let it happen

dapper isle
supple zephyr
#

at least never let it happen smoothly

earnest iron
supple zephyr
#

also the site policy and stuffs like that

brisk path
supple zephyr
#

bing lmao

novel parcel
dapper isle
brisk path
dapper isle
#

waht are you on mate

supple zephyr
#

doesnt chatgpt also use bing for web browsing

brisk path
drowsy tinsel
#

tbh if u use chatgpt enough chatgpt plus is probably a better deal than paying for the api

novel parcel
brisk path
#

Not GPT-fake

earnest iron
dapper isle
#

bing is far better than bard

drowsy tinsel
#

u don't have to make that many gpt4 requests to reach 20 dollars

brisk path
dapper isle
#

bro

brisk path
#

What are you on

drowsy tinsel
#

chat with bing was super good when it first got released