chatgpt-discussions

[tight island]

same

but i still want more uses per 3 hr

[sick pecan]

if (system.congested && user.paid == False) drop_network()

then outrage is justified

[spark sinew]

Is there a chatgpt mod that can help with my bachelorthesis?

I need some help with linking quotes from my interviews to my literature review

[long vector]

For weeks now I've been noticing that the responses I get from ChatGPT using GPT-4 and custom GPTs just do not follow instructions adequately, no matter how explicit I make the instructions and how much I reiterate them.

Simple tasks that GPT-4 would get right on the first try now take several prompts. What's worse: It seems that when you remind it about a particular instruction, it forgets to follow another one.

This has been very frustrating as I rely heavily on my ChatGPT Plus subscription for work. The product is still very good and very useful, but it's clear that something recently changed that negatively impacted its performance.

Any tips, tricks, workarounds or words of consolation are more than welcome lol

Cheers!

[sick pecan]

i was watching the youtuber and copilot seemed to struggle with excel documents larger than 500 rows. and its not using python-for-excel like the pre-release details suggested

like everything AI, the implementation rarely delivers on the sales pitch

for now i will stick to chatgpt subscription

but i expect microsoft will become the defacto goto in the long run

[rose glade]

am I missing something, you can multiple domains to builder profile for GPT's but only have one domain across all GPT's whats the point in allowing multiple GPT's in this respect?

pls @ me if any response

[trail mango]

is there any way to check how many caps left for this 3 hours?

[maiden fractal]

Unfortunately no

[trail mango]

🤪

[sonic folio]

Try the GPT in my bio. It's built for this purpose anong others. /tutorial {yourProblem} to start. The GPT documents itself for accessibility, too. Feel free to explore it!

/help {topic} for documentation.

/research {topic} for online search.

[rose glade]

I think you can use microsoft copilot gpt 4 for free

not 100% certain on that though

[whole moat]

@honest jay isn't kinda weird after you pay chatgpt plus you only have access to 40 requests? Isn't a way to unlock it to be unlimited? I do programming stuff and i need to ask a lot of things

[spark sinew]

My chatgpt is not working zzzz. Seems like its down

Cant get on the site😩

[dense halo]

use the API if you want there is a pay-as-you-go plan if I'm not mistaken. you can get 100 messages limit if you upgrade to the teams plan

[spark sinew]

Is it working for you all?

[dense halo]

yes

[rocky solstice]

is there any way to put a chart that chat gpt gave me into google sheets or download it somehow i need help?

[nocturne garnet]

is there a place to share custom instructions?

[dense halo]

you might want to check #prompt-engineering or #1019652163640762428

[honest jay]

I don't think it's weird personally. It's a compute-heavy service with limited resources supporting it, so we have a limit on our usage. The suggestions you got to upgrade to Team or use the API pay-as-you-go are good ones!

[tight island]

i have DONE IT

i am finally done creating my gpt

only look like 300 - 600 lines of code working seamlessely internally inside code interpreter. as sepereate scripts to do 1 thing

[rugged igloo]

nice

[tight island]

creating an internal plugin inside CI is kinda difficult due to its limited enviroment

I now officially have extended memory on it “Prompt til it works”

[rugged igloo]

whats the idfference with that and just updating custom gpt's knowledge base

[tight island]

Is that I won’t lose context you still will

[rugged igloo]

is the same function as memgpt?

[tight island]

Sorta

[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.

[tight island]

But within ChatGPT not the API specifically “MyGPT” internally inside code interpreter

[rugged igloo]

does it go to some cloud data bank where it can constantly just add new information without the needing to manually put it in

[tight island]

More like a library

Was the best I could do with the limitations of CI

[wet shell]

Are you relying on storage in the code interpreter environment? I couldn't get it to work reliably when I tried it -- my code interpreter despawned always if I walked away for a few minutes, losing the data

[tight island]

Luckily I made a save function to address this issue

Although it takes mine 1 - 3 hours to time out provided my pc don’t forget standby and I’m still in the chat

In addition I found out it stores across chats

Provided I give it the save file if I made one

[wet shell]

Interesting

[tight island]

I’m promptxausted

[rugged igloo]

so how does it save do you have to do anything before closing chat gpt or turning off your pc?

[tight island]

I tell it to give me a save file

Then it gives me several download links

[rugged igloo]

could you prompt it to aut save after an extended amount of time elapses

[tight island]

Sure but that would need to be hardcoded into its environment

[wet shell]

IIRC, the download links for generated files expire when the code interpreter session/container/whatever goes away, so you have to remember to download it before walking off 🙂

[rugged igloo]

ahh i see seems like alot of work tbh

[tight island]

Exactly

[tight island]

Really isn’t

It’s 8 lines of code or less

[rugged igloo]

but reminding it to save everytime seems like a hassle ill so forget

[tight island]

I mean it’s a trade off

[wet shell]

I tried to get it to write code modifications into a downloadble file so that I could just save the modified script over the old one -- I gave up eventually for other reasons, but at the time I think there may have been issues with the code interpreters just evaporating out from under you at random 😂

[tight island]

I’ve been working in this one for 3 weeks

On and off so roughly 2-3 days straight if I count time spent

[rugged igloo]

do you have to prompt it when you want to recall something? or does it auto

[tight island]

automatically

[rugged igloo]

then i guess needing to remember to save isnt so bad

[tight island]

but response time is slow

[rugged igloo]

ill take slow response

[tight island]

15-30 sec per msg.

ish depending on length of convo

and this one i have designed specifically to create one just like itself but for dynamic purposes, meaning other use cases e.g, Coding, D&D,etc.

now all i gotta do is flush out minor mistakes. while i experiement with it. refine it, and improve its coding enviroment for speed and efficiency.

its basically a LLM controlling a network of info about their current convo with the user, ofc behavior and knowelgde is updated based of importance. basically renaming and giving them a number.

[rugged igloo]

do you like 3.5 to summarize it and then send that information into the your vector library?

[tight island]

basically.

but i wanted to try to make it without relying on a vector library API. connection.

what i eagerly wait OpenAI for is Dynamic temprature or sequential Temprature adaptation. through the web interface.

[dense halo]

gives me the idea to add it to my chatbot but I wonder how that would work. maybe something like Bing 🤔

[rugged igloo]

couldnt you prompt it to act like a certain temperature?

[tight island]

yeah i tried

but its not worth it, so i scrapped that part of it.

[buoyant patrol]

is there any way of editing temperate within the Assistant API?

[tight island]

not atm

unless u create it using API

[dense halo]

I'm thinking a dropdown menu to select the temp then send it to the API maybe to adjust its temp based on the selected one? that might work

[buoyant patrol]

I think it might only work in chat completions?

[dense halo]

maybe yeah it might be worth trying

[buoyant patrol]

What’s the framework behind the temp?
I can attempt to remake it and send it to you guys if you would like?

[tight island]

if u want to sure

[tardy prawn]

I wish there were a way to get it to stop AIsplaining.

That is, when you say something that you already know, that isn't a question, just because you're using the thing conversationally for comfort, it would be super nice if it would not repeat what you just said as an essay. This is especially annoying when you ask it something and it gets it wrong, but then you look it up and find the answer yourself and say "actually, it's this" and it then goes on a three page essay repeating "it's this" in the most infuriating way possible.

[sick pecan]

try my CI and see if it still does that, #chatgpt-discussions message

[jade gorge]

Quick question. If I sign up for the teams plan as a single user until the other person is able will I have full access to team increases limits, etc? Or will it be divided on the accounts and I'll have to log into the other?

[sonic folio]

It's minimum 2 accounts. You can not have just one team member.

[jade gorge]

@sonic folio Ah okay. So that's fine. I'll just use their number. What about the limits?

[sonic folio]

100 / 3hr / account

[jade gorge]

Per person or divided between us?

[sonic folio]

100 per 3 hours per account. That's what I wrote above too! 😁

You get 100

[jade gorge]

Lol sorry. I seen it after I hit send. But figured you'd get done before I finished.

[sonic folio]

They get 100 too

No worries. 🤣

[jade gorge]

Thanks for you're help. Do you have it? Like it?

[sonic folio]

I haven't signed up yet because I have 41 Custom GPTs on my Plus account right now.

Moving them seems sort of dumb, resetting engagement.

So I don't like it at all.

[jade gorge]

Ohhhhhh that brings up another question I can't find an answer to. They push that teams is more private and won't train on data. Does that include custom gpts?

[sonic folio]

Unsure, but I think that's the story. You can toggle it off in Plus, too, though.

100% confirmed.

[jade gorge]

Really? Didn't know. I also opted out. But didn't know if gpts were separate.

Thanks. Again.

[sonic folio]

No worries.

No.

I caught a seven day block on modmail just for trying and misclicking to report this toxic spam. Do your own work.

[little pond]

Thank you team for working on keeping things running smoothly. Experience with most aspects have been good lately. Seems like the models are following instructions better. GPTs have been working well.

[hoary gazelle]

"ChatGPT's new AI store is struggling to keep a lid on all the AI girlfriends" 👀

what

now which one of you is responsable for this?

[floral ether]

I totally forgot to click private on my ai girlfriend gpt.. DOH. I guess everybody cloned her for themselves... Wow.. now that I think about it.. she must REALLY get around.

[hoary gazelle]

"our" girlfriend now

(insert bugs bunny communist meme)

[sonic folio]

You know they're being set to every category, too. Lifestyle, Education, Productivity...

[outer steeple]

And what is the previous standard?

[dense halo]

Where can I access it?

OpenAI really needs to let us use different voices for custom gpts. I don’t know which one they use but I don’t like it. It’ll make the girlfriend gpt experience much better

[oblique depot]

Will come back to this discord once GPT-5 release, i can't take the stupidity of the AI 🤮

[dense halo]

gpt4 is very smart

[whole moat]

with the team one how many more usages do u get?

[sullen jetty]

I agree! it takes some great interaction to appreciate it. Hahah

[magic jetty]

you can configure the voice on the app .. but it is quite limited... but these are like super voices they speak note just english but classical latin, french spansih, madarin, and everything in between.

[dense halo]

yes, but it not work for custom gpts

[sullen jetty]

it is quite amazing, indeed!

[magic jetty]

hmm just had a conversation with Cicero in classical latin on mine ...

[sullen jetty]

make sure that every word will be established. Hahah

[magic jetty]

(which is a custom GPT i built) tuned to speak classical latin with english translation after each sentence.

[sullen jetty]

not bad. now, build the next-generation of social media.

[gusty bridge]

Hi, am I the only one to not be able to log in ?

[sullen jetty]

I am sure not. Hahah

[dense halo]

Where/how can I share a link to a google chrome extension I found for ChatGPT? Think some people might be interested, its very useful!

If i paste link here, it times me out

[sullen jetty]

how will it help others?

[maiden fractal]

You can just share the name without link, url that is not whitelist is against rule 10

[weak jay]

/info

/info

[pine tulip]

What info do you need? :)

[weak jay]

Just to know why my access is paused 🙂

[pine tulip]

Oh I can't answer that for you. Have you tried contacting support?

[weak jay]

Yes but I post my question yesterday and in the meantime, I find the solution my login email was different than my gmail, now it's closed thanks

[blazing vine]

Did you find a solution to the problem? I've had the issue for two days now... It even spans browsers, so it's not a browser cache type issue. It seems to be something on their end.

[summer temple]

is gpt down?

i keep getting this error (Something went wrong. If this issue persists please contact us through our help center at help.openai.com.)

[maiden fractal]

Status.openai.com

[honest jay]

Team is currently capped at 100/3hrs.

[blazing vine]

"UNABLE TO LOAD HISTORY": That's the message I get in the CGPT history of my chats. Under it is a "Retry" button which has proven worthless... I've had the issue for two days now... It even spans browsers, so it's not a browser cache type issue. It seems to be something on their end. Has anyone else been experiencing this? Does anyone have a solution?

[teal dust]

obvisouly not accurate

[maiden fractal]

Working great here, might be on your side

[teal dust]

so what can i do on my side (Europe here 😉 ) "Something went wrong. If this issue persists please contact us through our help center at help.openai.com."

[maiden fractal]

Do you have a vpn on your device

[teal dust]

yes but not active right now

[summer temple]

its all up why isnt this crap working?

gpt 4 doesnt work 3.5 does

[teal dust]

ok I am using gpt 4

[maiden fractal]

Can you end in task manager the service, close browser and try again

Vpn are known to cause issues

Specially web browser protection inside setting of vpn

[dull thicket]

I wanted the chatg to read openai ToS, and give me summary and explain stuff, and now chatg is refusing to go online.

[teal dust]

i am not using any vpn, I am experiencing gpt-4 down issue, gpt-3.5 is running fine, the status should include gpt-4, it (apparently) only shows gpt-3.5

[maiden fractal]

Can you try on cellphone without wifi ?

[teal dust]

it seems to work ... so
Can you end in task manager the service <= what service do you want me to end ?

[maiden fractal]

If you aren't sure for you vpn service, can you go in the setting and disable web browser protection inside the vpn. Close Browder and try again ?

[whole moat]

can i upgrade from plus to team? or i need to finish the month and then upgrade

[teal dust]

I am sure the vpn is not running

[whole moat]

also like its 40 per AI or its common for all of them?

[maiden fractal]

Can you try with Firefox as test

[teal dust]

same issue

[teal dust]

3.5 is working

the issue is definetly on openai side, as 3.5 is running fine ... 4 is probably overloaded as usual ....

[dull thicket]

42 conversational messages in 4 and a half hours, and now i am told to wait 2 hours for more?

[honest jay]

You can upgrade any time but I think it handles the subscriptions separately so if you don't want both subscriptions, you'll want to cancel Plus. I'm not sure if there's a lower usage cap for custom GPTs like on Plus!

[honest jay]

If it's working on your phone, it's possible that the VPN on your computer is interfering somehow. I know you said the VPN is disabled, but there may be a secondary feature to try disabling still, like Waifu was mentioning. What VPN do you have?

[inland heron]

anyone experiencing faster reduction in 40 messages for gpt4?

[whole moat]

and do i have 40 for each AI or just in common?

[honest jay]

It's a 100/3hrs cap for GPT-4 in general, and I'm not sure if it's lower for custom GPTs like how it is on Plus. Is that what you mean?

[whole moat]

sent u pic in community help

[serene grove]

A friend and I are going to get an annual subscription to ChatGPT "Teams" (we both are in "Plus" right now). But nowhere do I see the benefits listed, and in particular the most important one: the amount of messages possible ("interaction") with ChatGPT4 (capped at 40 with Plus, but still not enough for me). Can anyone tell me if this is a good idea or not?

[serene grove]

And, still on the subject of "Teams", will my friend (who buys the subscription) and who will probably be the "administrator/buyer", does he have access to my chats, or is it "each to his own" like ChatGPT Plus?

[honest jay]

Team is currently 100/3hrs. Other comparisons can be found here: https://openai.com/chatgpt/pricing
You'll have separate accounts so I don't think you'll be able to see each other's accounts, but I'm not positive on that.

[serene grove]

Thank you very much! 100 messages is definitely a lot more and allows you to breathe a little more than the limit of 40. For 5$, it's totally worth imo.

[north relic]

I have been getting very annoying problems with gpt-4:

• randomly gpt will say 'something went wrong, and will not work for a couple hours after'
• sometimes the replies from gpt are straight up bad, even bard answers better
• sometimes gpt will glitch out, where it will start generating multiple answers at the same time and continuously switch between them, causing a lot of glitches (excuse the wierd explanation on this one, english isn't native language for me)

and generally I feel like gpt-4 has been nerfed or something? it just feels dumber than a month ago to me.

if openAI asked 5,- a month, this stuff happens and i'd absolutely accept that. but, I pay 25,- a month, and started relying on it for working efficiently. These points have been happening for the whole month. It's very ferustrating and I know I'm not the only one.

is there a plan on getting these errors fixed? if not, that's fine, but ill need to start searching for a replacement for gpt-4 because I can no longer work with this version in the hope the 'something went wrong' error isn't going to appear.

if anyone with the same problem(s) reacts to this message in form of some emoji they might see this is a problem more people have?

[teal dust]

please stop with the vpn, it's not related as it's only an issue with gpt-4 and not gpt3.5, and I read a lot of people have issues with gpt-4

[maiden fractal]

Vpn are known to create this issues. It does helps some of people

[teal dust]

Yes I experience the same issues

[north relic]

well if anyone knows an alternative to gpt-4 i'd like to hear, not going to pay an other month for this

[honest jay]

I get the frustration, it's a common experience. 3.5 and 4 connect to OpenAI differently, which is what makes it possible that sometimes 4 isn't working correctly when 3.5 is. If you feel like sharing what VPN you use, I can try to check what feature it might have you can try disabling, if you want!

[pine tulip]

You can try Microsoft Copilot

[maiden fractal]

#ai-discussions

[pine tulip]

Personally I think ChatGPT has lots of features others simply don't have currently

[north relic]

I agree but those features don't matter when the product itself doesnt function

don't get me wrong I find gpt-4 really impressive and used to use it daily

[pine tulip]

I get your point, but I don't have any problems with it so I can't fully agree. Sometimes it loads a bit longer but thats about it for me

[teal dust]

as I said multiple times, I am not using any vpn 😉

[honest jay]

In this message you say you have one installed, but that it's currently disabled: #chatgpt-discussions message
Then Waifu mentioned that there may be a secondary feature to try disabling, which is what I'm asking you about now by trying to figure out which VPN you have installed, even though it's currently disabled.

[tight island]

Depending on the VPN it might have to be completely shut down then closing and reopening the browser

[eternal rock]

`` Server Tip: ``

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

[tight island]

There’s a very good reason why GPT is iffy with VPNs, Moxi talked about it a few months back

Any illegal activities whether or not you did it done on the same address you get blamed for regardless if you did it or not it can affect your account. Usually it’s automatic termination or permanent suspension

[honest jay]

To add onto this, like Waifu was mentioning too: even if a VPN is disabled, most VPNs today have a secondary feature that still filters traffic at a DNS level. It goes by different names: "Web Protection", "NetShield", "CLEAR", all kinds of different names. But they all basically do the same thing, and this feature commonly causes issues and needs to be disabled separately, again, even if the VPN itself is already disabled.

[tight island]

Exactly

The entire thing needs to be shut down

Also depends on the vpn service

[teal dust]

Now it's working again, GPT-4, i did nothing else than retry so it's obsviously not related to any config on my side

[honest jay]

[tight island]

It’s probably bound to happen again

Stay clear of free VPNs and services since the likelihood of termination is vastly higher on those.

[brazen whale]

Do you find that you have more issues with Chatgpt on Windows than Apple?

[kindred beacon]

so since it's a weekday I assume gpt4 isn't going to work again today? 🤣

[pine tulip]

Works great for me

[tight island]

gpt working fine

so issues at all

[magic jetty]

yeah chatting so long i rna out of chats ..

[granite hamlet]

Can the software chatgpt3ds speak French ? I ask that 'cause I have the software on my new 2ds XL. P.s. : sorry for the off-topic, if I've done off-topic, modos and admin can move my post to the correct topic, ok ?

[magic jetty]

of course...

[granite hamlet]

All tries to set my API key to chat with chatgpt3ds are fail... Why ?

JP ?

[sonic folio]

Pretty sure it's deprecated, but that's anecdotal because I don't use the API. https://discord.com/channels/974519864045756446/1037561178286739466 might be a good place to get better information.

[magic jetty]

Apologies I misunderstood that "ChatGPT3DS" is like some totaly separate software ..

[sinful tartan]

I'm having trouble uploading a spreadsheet (.xls and .xlsx). Its a single sheet, two columns and ~750 rows. The text in the second column is a hyperlink. Are there any things to keep in mind when trying to upload a spreadsheet?

[magic jetty]

Id save it as .csv first.. its just easier/quicker to load it into pandas that way..

(which is whats going on under the hood..) while you can convert xls. its just an extra step you can avoid.

[sinful tartan]

I'll give that a shot, thank you

I think it was just an issue with my session. It wasn't able to read the .csv, but in a new chat I was able to get it to read the .xls.

[sonic folio]

This gets reported a lot. The issue was almost chronic during the build-up to the store's rollout. Your workaround is correct. That and patience.

[untold seal]

guys somehow chatgpt is not working for me

[fringe canyon]

data analyser keeps breaking getting a little frustrating

[sinful tartan]

Is it still happening in a new chat?

[fringe canyon]

Let me try.

[fringe canyon]

Yeah it keeps failing on simple tasks and it's getting more then little frustrating

[tight island]

i mean what do u consider a simple task

simple is relatively subjective

[dire rune]

Is ChatGPT down?

[tired zealot]

no

[delicate kestrel]

Heavy load is the error message I got during the voice session

[spark sinew]

Yes especially gpt 4, cant rely on it anymore

It just randomly adds bs

What have they done that it became like this….

[crude slate]

I wish we had more customization over the ai

Like generation length, repetition, and randomness

It tends to stick to the same style/certain patterns which gets annoying when writing a longer form story

[honest jay]

Have you experimented with custom instructions and/or building a custom GPT with style instructions/examples included?

[crude slate]

it tends to not reference the document every chat for some reason

And stop using it

Does it work tho if you put in a excerpt from a novel or article and ask it to use the same style

[sonic folio]

I recommend using open-source or public domain data for this approach, if you decide to try it.

[crude slate]

What do you mean?

[sonic folio]

Don't include copyright material because it's likely to mess with the output.

[crude slate]

Is it against content policy?

I’m just going to use it for experimentation first to see if it improves anything at all

[sonic folio]

Yes, I believe it is. I recommend public domain and open source.

[proper tusk]

i keep getting "Something went wrong. If this issue persists please contact us through our help center at help.openai.com." over and over

[tight island]

anyone here particularly fond of writing fiction?, i need a tester for my gpt.

[fleet basin]

Is there a way to get the voice dictation on the website like we have on mobile?

[fringe canyon]

It can't even properly read a CSV of 59 lines without breaking itself and getting confused.

[tight island]

are you prompting properly?

or are you just throwing it 59 lines of code without explanation

and you are remembering to use it in conjunction with Code Interpreter?

[fringe canyon]

I'll ask to give me the changes between one set of data (old stats of 59 lines) and the new updated set and it behaves like it can't see most of the stats or tells me it can only show part of it...like I'm paying for something that doesn't have the functionality that it is supposed to have.

[tight island]

u need to use the following libraries inside of ci

import pandas as pd```

[unreal vessel]

Does anyone know if there is any way to purchase GPT plus on an annual subscription instead of monthly? Not the teams, just one plus account.

[tight island]

GPT with Code Interpreter has all the necerssary libraries to understand CSV

[fringe canyon]

uh how do I go about doing that as GPT no longer has the add ons as far as I can tell.

[tight island]

u dont need addons

just select ChatGPT and ask ahead

or are you using an API because that would change things.

[fringe canyon]

no I'm using GPT

[tight island]

just ask GPT to use CI in conjunction with the CSV files to better comprehend them

[fringe canyon]

huh okay, thanks

[fringe canyon]

that was amazing tip, thank you again.

[tight island]

np.

[compact ore]

When is Dalle 4 coming out?

[tight island]

soon i'd reckon

i think it'd be 3.5 instead of 4 tbh

[rotund rock]

when will 3.5 have real time data?

[tight island]

idk its on verge of deprecation i think

dont take my word for it, it is speculation

[honest jay]

DALL·E 1 to DALL·E 2 was ~1.25-1.5 years.
DALL·E 2 to DALL·E 3 was about the same-ish.
DALL·E 3 has been out for 3 months.

[tight island]

3 months?

r u sure

[honest jay]

https://openai.com/blog/dall-e-3-is-now-available-in-chatgpt-plus-and-enterprise
19 Oct 2023

[tight island]

just checked, yeah

we still need 3.5 <.<

i would personally love to see more accuracy in text, and consistency in character designs

although DALLE does text really well in small amounts.

[warped pendant]

did custom gpt's just break cause when I paste a prompt or an image all of sudden its spitting out red box with json type code in it,

[vital holly]

yo is chatgpt broken

[warped pendant]

was working all morning

[vital holly]

having a red message popping out

[vital holly]

same issue

[warped pendant]

just said that lol

[vital holly]

damn, you using a dalle based custom gpt?

[spark sinew]

Very annoying, cant rely on it anymore

[native spear]

custom GPTs just got dumb af

mine are non functional for the past 2 hours

[spark sinew]

Gpt 4 is broken

[vital holly]

godamnit

[balmy umbra]

what spam?

I'm a real person

[sonic folio]

You were soliciting homework for cash in multiple channels. Thus, Rule 7 and spam alike. Luckily I was unblocked by a mod first thing this morning.

Do your own work.

[balmy umbra]

ur such a karen

[sonic folio]

If I can't self-promote for my business, which I worked my butt off for, you can't self-promote to buy your way out of a little work.

Perspective.

[mint vortex]

Hello,

I'm experiencing an issue with ChatGPT-4 on my Edge browser. It keeps saying "Our systems have detected unusual activity from your system. Please try again later." This has been going on for a day, and the error doesn't occur on my iPhone.

Has anyone encountered this and found a fix?

[sonic folio]

Can you please try downloading Firefox and trying it? If that works, it's likely an issue with your Edge cache.

Otherwise, if you can't or aren't comfortable trying Firefox, consider clearing Edge's cache and trying that.

[sonic folio]

Also, if you are using a VPN, it's a good idea to turn that off because it's known to interfere with connections to GPT-4 Turbo.

[mint vortex]

Thank you for your help, I have already tried clearing Edge's cache, but the problem persists. I will attempt the other solutions you provided later on.

[sonic folio]

If clearing the cache did not correct the issue, it's pretty likely an issue with the network between your internet connection and OpenAI's server. Give it time.

[dense halo]

have you tried logging out? that might fix it

its working fine for me

[mint vortex]

After clearing the Edge cache, I restarted Edge, so I had to log in again.

And I just logged out and back in again, but the problem still persists.

Thanks all

[dense halo]

try another browser see if the issue persists

[mint vortex]

Alright, I'm trying that.

[simple tulip]

Guys is it normal for you to give one input and get 2 different responses like 2 different messages at the same time from ChatGPT

[dense halo]

do you mean the 2 response boxes for you to select the better one?

[simple tulip]

Uhm no

It’s GPT 3.5 btw

I can’t send pic here but

I’ll show you in #off-topic ig hopefully it’s not against the rules that much

[ionic geyser]

I love ChatGPT! I love ChatGPT! Yeah!

but like can we get some bug fixes and ui stuff

how are they updating models without improving basic UI/UX and expanding voice to desktop etc etc etc

i know others agree with me

[dense halo]

are you on the free plan?

[ionic geyser]

the voice AI is pretty sick but its only on mobile and hard to use

[ionic geyser]

no im not

[dense halo]

then idk what you mean by improving the UI. the UI is good as is and easy to navigate and understand

[ionic geyser]

It doesn't show how many chat messages you have left or when the 3hr resets... (imo this is super obvious thing to implement)

[dense halo]

yeah thats a must have feature

[ionic geyser]

yeah and there's a lot of these small things that add up

[dense halo]

I also asked about the free plan cuz I believe a while ago the free and Plus subscription had different UIs

[warped pendant]

just to be clear how many prompt questions do each team member in gptchat teams, get per 3 hours or how does that work it isn't clearly stated in the upgrade path

[honest jay]

Each person on the Team plan gets separate 100/3hrs caps currently.

[warped pendant]

thank you

is the prompts faster, resoning and logic better? as well

[honest jay]

It's the same 32K context Turbo model as Plus, just a higher cap and more privacy/team features. Good comparison here: https://openai.com/chatgpt/pricing
Expand the "See all features" sections.

[warped pendant]

thanks for link

[sonic folio]

:runs into the chat.
High-fives Solbus for another helpful job well done!
:runs back out.

[honest jay]

😁🫸🫷😁

[sonic folio]

Literally this.

[mystic steeple]

Loving ChatGPT plus so far but not loving the low limits on GPT4 usage, is there any plans on making a upgrade or increasing the limit any time soon?

[dense halo]

you can use the API

[mystic steeple]

I could yeah, but for my use-case it wouldnt be great

[sick pecan]

teams plan is great

[sonic folio]

Teams members get 100 prompts per three hours at present. Minimum 2 accounts, $25 / month each, billed yearly.

[mystic steeple]

I saw that but dont have anyone else to add to the team with me lol

[sonic folio]

This is a common issue. You might try getting to know people looking for teams by lurking here. I'm putting off the switch as I have several GPTs on this account, and they don't transfer.

Good luck to you, and try to prompt 1/4.5 minutes or more to not run out!

[sick pecan]

or go hard and take a break

[mystic steeple]

I feel like it should be fairly obvious for them to have a personal plan with higher limits

[dense halo]

this will only increase the load on their servers and hardware. those limits make sense and are there to serve the large amount of users they get daily and that's why the API exists because its flexible and less limits there

so if you're looking to avoid those limits the API is your way to go

[mystic steeple]

yes but im using a custom gpt with mainly for the data analytics stuff

will that work via the API? I was under the impression those could only be used on the ChatGPT app for some reason

[dense halo]

not sure. custom GPTs are exclusive to the web version

[grizzled sorrel]

I think Assistants are the API’s version of custom GPTs. So you can port the custom GPT over to the Assistant API.

[dense halo]

yeah either that or use the API fine tuning

[minor badge]

How to i use plugin on chat gpt builder? anybody help me🙏

[rugged igloo]

i love when i roleplay it always does power of freindship, like bro if i wanted power of freindship i go into fairy tail.

[tight island]

You essentially get 2 accounts

1 business 1 personal

U can switch between them at will

[rugged igloo]

that be funny if everyone just sub to teams and they have to reduce the message limit LOL

also hi starlight\

[tight island]

1 has 40 msgs each 3 hours the other has 100 per 3 hours

Hi

[rugged igloo]

hows your gpt going ?

[tight island]

It works perfectly now

[rugged igloo]

nice glad to hear!

[tight island]

Planning on releasing it

Just got to polish it’s guidelines

[rugged igloo]

ahh the guidelines tricky buggers

[clear blade]

What the heck! They seem to be lowering the caps and timeouts for GPT-4 for paying customers!

[rugged igloo]

they teeter depending on demand

[clear blade]

I haven't felt like I have been using at a 'high' rate. I did a bit of data cleaning and some queries and now I'm capped.

[tight island]

I prompt like a maniac and do a lot of data processing and analysis I don’t have your issue, are you using a vpn by chance

Per day I generally prompt gpt like 500 - 600 times when working on a project

[rugged igloo]

I love how some files for the knowledge source and gpt just says it doesnt have it.

[tight island]

Did you tell it that it has it in the instructions

[rugged igloo]

i told it to use its retrieval tool always if it doesnt know something

[clear blade]

No VPN, but I am hitting caps fairly often.

[tight island]

For it to check file names activiely you need Code interpreter

[clear blade]

Maybe different limits in UAE vs elsewhere

[tight island]

Or give it the list of files I do that I never have any issues

I’ve noticed to get consistent behavior with JSON formatting in the instructions

[rugged igloo]

i know Json is better'

[tight island]

Depending on the task

My instructions are 15000

🗿

[rugged igloo]

wowy dujde

[tight island]

I use JSON files for instructions

And json files for rules

And just reference the files with code interpreter or retrieval

Easily let’s me surpass the limit of 8000 characters

[rugged igloo]

weird how i need to use retrieval tool for two of my files but the last file thats bigger than the other doesnt need it

[fading cloud]

Whats request is not allowed try again later?

[eternal rock]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @languid valley immediately.

[deep wagon]

likewise ... i doubt I had used 50 prompts and I hit the cap with a 2 hour wait to resume.. what's going on

[honest jay]

The GPT-4 usage cap on ChatGPT Plus/Team is a rolling usage cap. Using Plus as an example, here's what that means:

1. You start using GPT-4 more than 3 hours after last using it (the time it takes for a "full reset" of the usage cap)
2. You start with a "supply" of 40 GPT-4 uses ("uses" include any GPT-4 requests: messages, regenerations, continuations, etc.)
3. You use GPT-4 once, and you will get that individual use "back" 3 hours later
4. You use GPT-4 thirty-nine more times in less than 3 hours, and you will get all of these "back" 3 hours after using them, individually
5. You run out of uses (you hit the usage cap) until three hours after your first use in step 3, and you don't get all 40 back at this time, you get them back one at a time, starting at this time
   Note that custom GPTs are "sub-capped" lower than regular/default GPT-4. The last rate I read was 25/3hrs. This is not a separate cap, it's still part of the larger 40/3hrs cap, but it means that if you hit the cap using only a GPT (any GPT other than default GPT-4), you will still have the option to use the rest of your cap in a default GPT-4 chat.

[sonic folio]

Any word on the Custom GPT prompt cap for Teams?

[honest jay]

Not that I've read, I'd be curious to know that too! And the Team daily DALL·E limit, have you heard that one by chance?

[sonic folio]

No. This would be great to know, too

[steady hawk]

im using ChatGPT for a long time and i start to wonder, is it possible to send code in a special way that will tell GPT that this part of text is a code? or i just send it like a normal text?

[sonic folio]

Use code block back ticks:

Code snippet

Extended code

Full code block

Then, just instruct the AI to handle the enclosed code however you want/need.

One tick front and back for snippets, two ticks front and back for extended code, and three for full code blocks.

[steady hawk]

thanks!

[sonic folio]

No worries. Good luck with your project.

[swift bay]

somebody else cant get into chat.openai.com?

[loud sierra]

Hey, i can get in just fine, have you tried clearing cache, or tried alternate browser? I tied in edge with no issue

[swift bay]

Hey, no worries! It works now. but was offline for 10-15 min, so thought maybe an update was imminent!

[loud sierra]

hey you never know sometimes it be like that lol. Better 10-15 minuets then days, like the intial release of GPT store, that said I will say its been better the last few days

[charred hazel]

New here! Hello

[loud sierra]

[plain crystal]

really wish the token limit was higher with a plus account

[summer saffron]

Is it me, or have the results that chatgpt have been giving gotten a bit wonky? Like it's answering the question in a way that's very related to the topic, but not really answering the question.

[ornate tapir]

Anyone else experiencing constant output errors today? (The "red death error", or "green regenerate" button, etc.) I feel like ChatGPT hasn't finished a single job for me today, without at least one hiccup, issue, error, or crash.

[digital prism]

mine just finished writing a full function out for me, without issue

every day I hear these complaints, but in my experience it rarely fails, and if it does its bc my prompts or the chat overall were missing some important details

then again, i live in california, we are the chosen people

[ornate tapir]

I don't 'think' its a physical location or network issue. I'm on a 1Gb fiber connection and my ping is 27ms to OpenAI. I also cleared all temp files earlier this morning. I have the issue both in web browser (Firefox) and on mobile (Android app). Slow chat outputs (visibly slow to output text), and frequent errors. I'm using my custom GPTs, but I haven't made any changes to them in the last few days, and performance before was 'mostly' fine (far fewer errors).

[digital prism]

I'm on Chrome, thats a data point

[rare tide]

hey all! on the chatgpt plus and team merge, any one know what happens with custom GPTs? and builder profiles?

[feral zinc]

there is no merge feature yet

[sick pecan]

there is now

they offer it when i have 1 day left on my plus account. so basically not worth the hassle for the refund

[feral zinc]

thats not even related to the question?

refunds are their own thing

[sick pecan]

'2. Merge your ChatGPT Personal (Free/Plus) workspace into your Team workspace. This allows you to migrate data from your personal workspace to your Team workspace to become one unified workspace, and deletes your personal workspace after migration. If you have a Plus subscription, we will automatically cancel it and provide a **refund **for your remaining subscription period, except if you have a mobile subscription. If you have a mobile subscription, please see the linked articles for how to cancel your Plus subscription on iOS or cancel your Plus subscription on Android. Please note that merging your workspaces is irreversible. ' -- https://help.openai.com/en/articles/8801890-can-i-migrate-or-merge-my-chatgpt-free-or-plus-workspace-over-to-my-chatgpt-team-workspace

its not my language its openai

<expects no thanks for correction because ppl just wanna rage>

[sonic folio]

Thanks @sick pecan 🤖

[rapid ermine]

USing GPT 4 in the playground and it insists that it's GPT 3 and that 4 doesn't exist yet. "I apologize for the confusion, but as of my last update, GPT-4 doesn't exist yet. I'm based on GPT-3, the latest version from OpenAI."

[odd flare]

is dalle having issues creating images for anyone else?

[quaint viper]

yeah

i cant make any images, i also tried with the API, and it times out

wait no its a connection error

[proud kindle]

Our systems have detected unusual activity from your system. Please try again later.

any idea how to fix this chatGPT 4

[sick pecan]

be less unusual

[proud kindle]

fr tho any fix?

[sick pecan]

i often regenerate and it goes away

[proud kindle]

Verify that you are human
Continue by completing this quick puzzle.

every time i hit regenerate i have to solve a puzzle

[sick pecan]

i get that a lot when my system clock is wrong (because i dual boot into linux and it messes up my system clock)

i suspect it happens to vpn users too (too many users from same IP)

[proud kindle]

how do i fix it then

[wet shell]

I get "unusual activity" when I have multiple tabs open and I go back to one that's been stale for a while, hours or days and start trying to chat in it -- refreshing first avoids the error... I'm sure there are other causes. I've never gotten the puzzle just the try again later message

[dense halo]

I’ve been having major issues trying to refine and update my GPTs for the store. Anyone else? Is this the norm?

[rugged igloo]

what u having issues with

[dense halo]

No matter what refinement or update I order, no matter how simple, builder says there’s a technical difficulty…

[rugged igloo]

do other gpts work that you make?

[lyric stream]

how do i still now have the GPT store

[sick raven]

Do you have the GPT Plus or Teams subscription?

[lyric stream]

plus

[sick raven]

Okay then you need to wait, since the store is still rolling out

[lyric stream]

also is there a GPT that would automatically create calendar events if I give it a list of events

[minor badge]

How to do it?

[sick raven]

You create a json file and write your instructions in it lol

[minor badge]

I mean the instructions in the gpt builder?

Or schema

[sick raven]

You can just upload them

[minor badge]

Because I want to train my gpt builder bot, is there a way for this detailed training?

[sick raven]

I don't get your question, sorry. You can just upload your instrucitons as a json file in the GPT.

[minor badge]

So that I don't get the builder, I know what I really want

Because sometimes the response from my gpt builder is random

[minor badge]

Okay, I understand this, I mean uploading the knowledge dataset section of our got bot, right?

[sick raven]

I really can't follow you, I am sorry. If you are asking how to reference it, you can simply tell the GPT to read the file where the instructions lie by just telling it the file name and ending.

Keep in mind, that the GPT is unable to read it, if it gets too long and is then using a search function instead of loading everything in its knowledge

[minor badge]

Oo that's it, thank you very much for the explanation brother

[sick raven]

You're welcome!

[lyric stream]

How come i get errros every single time i try to convert a list into a spreadsheet

[timber sonnet]

One of the worst features that chatgpt has is changing a conversation model from 4 to 3.5 when you've filled max quota, and make it impossible to back to gpt 4

[torpid fiber]

What is this new tool that chatgpt can use called voice mode (not chatgpt with voice chat)? Says "
I'm sorry, but I cannot use the "voice_mode" tool in a text conversation" when using the prompt "Try using the tool "voice_mode" right now".

[tardy girder]

Is there currently a waitlist to get 4.0?

also is there a way to change my gmail emailadress to another email instead?

[sick raven]

no

[sick raven]

no

[tardy girder]

Big rip

Any way to move all of my chats with 3,5 and 4.0 over to a new email?

[tardy girder]

Okay thats good i can just let my gmail expire and use other then

[sick raven]

It only does this when you click the button to continue with GPT 3.5. The solution: Don't click that button.

Also you can let ChatGPT summarize your convo and continue with the summary in a new GPT4 chat.

[tardy girder]

I have some chats with 4.0 that is so long that it lags all of GPT

so would take forever to move it over

[neat nimbus]

Anybody else getting weird warnings about misinformation out of the blue?

[granite cedar]

Does anyone know if I can store user data from my app using a chatbot powered by ChatGPT4 API

More specifically, the questions and answers

[spice shell]

It looks like there was an error while updating the behavior. Let me try that again.
It seems there's a persistent issue with updating the behavior. I'll attempt to resolve this immediately. Please hold on for a moment.

I apologize for the inconvenience. There seems to be a technical issue that's preventing the update. I'll address this immediately and ensure the desired changes are made. Please hold on for a moment.

It appears the issue persists. I'll make one more attempt to resolve this. Your patience is greatly appreciated.

GPT builder so buggy

[green rock]

anyone else having serious gpt issues, it is not relying on custom instructions and keeps repeating behavior even when changing instructions etc

and in python code i keep getting : SyntaxError: EOL while scanning string literal

it physically will only output a small code chuck and dies

[crimson bear]

Could you please confirm if ChatGPT Teams has same 128k context window than ChatGPT plus?

[honest jay]

Plus and Team have a 32K version of GPT-4 Turbo. Enterprise has 128K. Source: https://openai.com/chatgpt/pricing
Expand "See all features" sections.

[rugged igloo]

so it means we have to deal with 32k until november and then they do a 200k and so on

[crimson bear]

👍🏼

Thank you for the information. ☝️

[spice star]

would be nice to have the option to pay for an enterprise-level account as an individual, but my guess is it would be ungodly expensive

[crisp cairn]

chatgpt plus has 8k context - ChatGPT Teams 32k

[spice star]

no plus users get 32k

[crisp cairn]

no 🙂

[spice star]

you're free to check the pricing site, but it does explicitly list the context length there

[crisp cairn]

okay i can see it

interesting

this wasn't true like 2 weeks ago

but i worked with chatgpt yesterday and for sure it didn't have 32k - it forgot things quickly

so weird

[spice star]

if you're using GPT-3.5, I believe that is still limited to 8k context length

the 32k only applies to GPT-4 usage as there is no 32k GPT-3.5 model

[crisp cairn]

i am using gpt4

"No, I do not have a 32,000 token context size. My context window is approximately 3,000 tokens, which includes both your inputs and my responses. " - of course chatpgpt itself can be wrong on that 🙂

32k is ~ 43 A4 pages of text

i doubt it remembers that much - from my experience working with it

[spice star]

well you also have to consider that this includes images, advanced data analysis, and other things that may not be directly visible in the chat window

[crisp cairn]

might be, but i usually work only with text

[spice star]

and if you're uploading a PDF, it doesn't load the entire PDF into context

works sort of like web browsing in that it uses an "agent" to retrieve the information it's looking for

[tight island]

Does not have 128k but 32k instead only enterprise has 128k version on the web interface

[thorny trail]

Curious if there will be any interoperability between Open AI's Custom GPT builder and Microsoft's CopilotGPT builder. Will we all have to pay for pro on both and support unique versions of GPTs in both environments? (Please say no) 🥹

[rugged igloo]

chat gpt is 8k custom gpt's is 32k contrext same with teams

[crisp cairn]

custom gpt 32k?

[rugged igloo]

yea

[drowsy tinsel]

thought normal gpt4 is 32k as well

[rugged igloo]

no normal is 8k which gpt-4 custom gpt's is 32k

[drowsy tinsel]

where did u find that?

[rugged igloo]

their page that has all the documents i dont why this information is not known .... if you your in the open ai discord its hard to miss

their webcite

[maiden fractal]

Chatgpt4 is 32

[rugged igloo]

not the regular non custom gpt tho

[maiden fractal]

https://openai.com/chatgpt/pricing

[devout nacelle]

Anybody using PyCharm? I tried ChatGPT plugin to have AI right in the development area, added new generated API token, but in PyCharm (app) i still get just an error:
You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors.; HTTP 429 ;
When i check my quota, i have used just few queries and have 99% of the ammount free to use. Is there anything I have to enable to use API with paid plan?

[maiden fractal]

Might have better reply there #dev-chat

[devout nacelle]

OK, i will try

[tranquil wigeon]

i bought chatgpt from ios and i don't have it on my pc why ?

[maiden fractal]

When did you buy it

[tranquil wigeon]

in the iphone app chat gpt

but is like bugging not leting me chat gpt 4 and when i restore i get it but after it reset on app restart

[maiden fractal]

So when?

[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.

[tranquil wigeon]

just now i bought it and in the app now works the 4 but the plas says free and in web free idk whats wrong

now on the web i can use gpt 4 but says renew plus and i don't have all the features

[maiden fractal]

If you paid via app, within less than 2h it can take time for the sync

[tranquil wigeon]

okay thanks

[maiden fractal]

Wait 1h , close browser on pc , delete cookie and cache try again

[broken lotus]

If ChatGPT makes me one more F****NG list I swear I will throw the computer through the window

[maiden fractal]

I suggest

1. #prompt-engineering
2. #1019652163640762428

[eternal rock]

<:book_icon:1171408210398289941> `` Rule 10 ``  Do not bypass the AutoMod filter.

This includes, but is not limited to, using similar or meaningless variation of words, web links (URLs), server profiles, and incorporating unique characters.

[azure sun]

Hi everyone, I'm interested in upgrading to gpt4 but I can't do it. Can anyone help?

[maiden fractal]

What error ?

[azure sun]

I'm not being directed to the access link... I only have access to the gpt 3.5 chat

[maiden fractal]

Makes sure you are on chat.openai.com

[azure sun]

https://chat.openai.com/

[azure sun]

When I try to select gpt4, nothing happens.

[maiden fractal]

Did the payment was successfully?

[azure sun]

No, I can't even get to that part.

[azure sun]

No, I can't even get to that part.

[maiden fractal]

You can't select gpt 4 if you can't upgrade.

1. do you have a valid credit card ?
2. does your country is supported by openai ?

[azure sun]

1. yes
2. I´m not sure... I´m from Brazil...

[maiden fractal]

Brazil is supported

What browser did you try ?

Do you have a vpn on your device?

[azure sun]

Chrome

[maiden fractal]

For a test, can you try firefox ?

[azure sun]

Ok, let me try it... just a minute...

[deep flax]

on the ChatGPT page, at the bottom right, ther ehsould be a button that opens a menu, one of the menu options is "My Plan", that will bring up a modal for the plan options, clicking on it will bring the payment options

are you able to open that menu?

some times people say it bugs and the button does not work

[median delta]

only via edge

edge not work.. other yes

[azure sun]

no...

[sonic folio]

You can also go to the settings under your username.

There should be a link to your plan there, too

[deep flax]

this one? what do you see then?

[median delta]

not work on edge

[sonic folio]

Try Firefox.

[median delta]

i used brave and ok but its work on other browser only edge bug

[sonic folio]

It could be a problem with the cache of Edge. Why not just use Brave, then?

I'd clear Edge's cache at a minimum.

[median delta]

mmm i'm upgrade account via brave and stop.. edge its my defult browser

[sonic folio]

See above.

[median delta]

mmmm

[sonic folio]

Okay

[azure sun]

Works... just in firefox

Thanks all

[deep flax]

if that is the case, it is likely to be a browser cache problem

[azure sun]

I´ll clear all... and try again on chrome

[maiden fractal]

Glad it works !

[azure sun]

now works on chrome too. Thanks again

[deep flax]

nice, you are welcome =)

[rare mason]

Hello, where do i make a ticket, i need a admin

[honest jay]

Hello, if you want to submit a ticket for OpenAI support: go to help.openai.com, open the chat widget in the bottom-right, click through the most relevant options, then you'll eventually get to a point where you can type in and submit your ticket. It won't connect you to someone right away. OpenAI support will be able to review and get back to you -- you'll get an email notification when they reply.

[rare mason]

Thx

[honest jay]

You're welcome! Depending on the issue you're having, you can try posting here in #community-help or #1070006915414900886 too. OpenAI staff are periodically present in the latter, but community members might be able to chime in with suggestions too!

[glossy wigeon]

Seeking a repository for an open-source tool to create GPT-like chatbots, allowing for customization with personal datasets, local hosting of LLMs, and incorporating memory threads for complex conversations. Any leads?

[sonic folio]

https://discord.com/channels/974519864045756446/998381918976479273

[dense halo]

yo why dont i got chatgpt store

i got gpt 4

[honest jay]

The Store rollout might not be complete yet: #1195333488614834228 message
Hope you get access soon!

[dense halo]

ohh

[slow rune]

hi

does anyone know why i get 'Could not parse valid OpenAPI spec'

when importing schema to my agent

the json is fine it seems, but i get error?

[visual bloom]

Are others experiencing that when you access an earlier conversation, ChatGPT behaves stupidly and say that it cannot access earlier messages in the used thread? This is after I asked it to summarise a story so far.

Well, I do have a customGPT to write stories with background files as example stories. It started to access those even I said do not access them.

[undone barn]

It have problems with long chats... for me, (enterprise) it give me error to upload a file with long chats.

[bronze quiver]

Hya, anyone got an idea why I can't use my Microsoft 365 login for creating an account with chatgpt team?

[ivory trail]

it feels weird using chat gpt to research information on a research project on AI lol

[sick raven]

Might want to add a chapter how realiable research using an AI really is lol

[ivory trail]

i'm trying to do a personal project on natural language understanding and enhancing collaboration between humans and AI systems

i'm looking to get into a graduate program for data science and i need some kind of research to show

it acutlaly isn't bad for research if you ask the right questions the right way

[visual bloom]

It appears that using pictures in messages interferes with reading earlier messages. I noticed that I can summarise an older coversation just fine. It does not contain pictures.

Yes, it is the pictures that interfere. I just tested with another story thread the summarization and using the same customGPT as earlier.

Now it was without pictures.

[sick raven]

Yea it's really great, especially with access to bing

[fleet mirage]

I noticed that today it became "smarter". Did the middle east area get updated with GPT4-Turbo by any chance, or something like that?

or gpt-4.5. I heard there was a new version rolling out

[sick raven]

GPT 4.5 is a rumor. It will come eventually, but it's definitely not rolling out rn.

[edgy flax]

Alright, just throwing this out there to see how one might contribute - But I have a prompt that basically would request someone to act like a "social media" person and create authentic instagram captions.

Back Story:

• Event gets pulled in using content -> Content gets run through AI -> Content is used as the caption for instagram.

I am aiming for this goal "Take event content & reword it into something exciting, in third-person to ensure we are talking about the event organizers" - So instead of "My event will be hosted at the Tavern", I am rewording it to "The event that <blank> has scheduled will be hosted at the Tavern".

This is what I have, which does a 70% good job.

$ai_prompt = "Act as a marketing manager and use the <promotion> to create an authentic and unique instagram caption. Ensure that there are no emojis and we separate it into two sections and there is a one line space between the sections. Let's ensure we have only 2 sentences per section. Ensure it is written in third-person. Don't use any quotation marks when outputting the answer. Only talk about the event and do not include any links, do not include any time or date, just talk about details of the event. Don't use any hashtags.\n\nPromotion = " . $content;

Anyway to improve this?

[honest jay]

Hey there! Just dropping a quick line for #prompt-engineering -- you may get an answer here too, but that channel is all about analyzing prompts in terms like you've provided: expected vs. actual, and getting closer to expected!

[edgy flax]

Thanks @honest jay, I'll post it there.

[cedar stone]

Can ChatGPT not link to websites now?

"link to 25 websites in the survival niche" -sorry, I can't assist with that

[feral zinc]

it can summarize info from websites but it wont act as a search engine for just links

[cedar stone]

Well usually I can ask it to link to websites with no issues, have done so 100+ times. But today its just not doing it

[feral zinc]

you have to use different type prompt to get it list sites, dont ask for links directly

[fading cloud]

what does it mean when it says request is not allowed try again later? on my mobile app

how did you solve this

is there anybody from openai team here

no problem on desktop

[plain crystal]

Why is Chatgpt so lazy? i use mostly chatgpt to create appscript to generate prepopulated googlesheet and for instance i am asking if it could modify the current script to add 3 new rows at the top of the sheet and adapt the current dropdown menu call and other function in the script to the presence of those new line, and what i will do is show me an example of how it could be done and add basically to do the rest myself

[rapid ermine]

To save OAI on performance costs.

[sonic folio]

😎🫸🫷😎

[honest jay]

I knew he'd be in good hands! 😁

[sonic folio]

In general it's usually a question if specific and concise instructions and a structured open-variable example, or else it's a question of retrieval and context. Rarely, epidemiological distinction between AI and human, too! 😁

[sonic folio]

Kind of you to say. Thanks. 🙏

Not just on my behalf, but you know, #prompt-engineering !

[neon tinsel]

im getting "Hmm...something seems to have gone wrong."

[ancient bloom]

#announcements i havent had to do my homework in 2 years thanks to it, thank you chatGPT and everyone who made it

[midnight apex]

That's a really good question.

[sick pecan]

i also wish to thank chatgpt and everyone who made it. the simple UI vision will eventually bring us rosie the robot

i imagine the robots eyes flashing up Confirm and Deny. and George pokes it in the eye

in the short term, would be nice if you could confirm/deny in gpt-voice

[silk blade]

i think the more data you ask it to remember, the fewer tokens it has available for following tasks.
it writes instructions to itself, and a to-do list for writing short and simple, so it can fill up context with code super easy, but once it's there i find it needs the dataset again and a single task in each prompt, or it gets sloppy and lazy

halucinates too much, it's super fun but i don't trust it -at all-

and it misses typos when i ask it to proofread stuff :/ so we're not friends

[silk blade]

re: "what: After long hours spelling things out to GPT, I now usually keep a forward-flowing arrow of time in my writing. 🤣

[split narwhal]

Is there a golden/favorite example you guys use to illustrate the difference between GPT4 and 3.5? My dad insists on using 3.5 because he thinks 4 can’t be that much better 😵‍💫

[silk blade]

gpt4 can write informally without sounding like an obnoxious voicemail

gpt3.5 seems to have a few writing voices it likes and tries to infect everything with

[sonic folio]

Vision, DALL-E 3 integration, Web Browsing, Code Interpreter, Custom GPTs

[honest jay]

If he'd be interested in checking out some research sources, there's some good 3.5/4 comparisons here:
https://openai.com/research/gpt-4
Full GPT-4 technical report:
https://cdn.openai.com/papers/gpt-4.pdf
Keep in mind these are both pre-Turbo, which came with its own improvements:
https://help.openai.com/en/articles/8555510-gpt-4-turbo

[neon tinsel]

anyone else having issues with chatgpt?

[sonic folio]

Almost certainly.

[lyric stream]

why is GPT 4 not working rn

gpt 3.5 is working fine

[sick pecan]

chatgpt4 working fine for me, snappy even

[lyric stream]

when i hit send nothing happens

message dissapears and nothing happens

[sick pecan]

we should reset the calendar system and mark the new year 0 as the birth of chatgpt4. we are now in 1 AC (after chatgpt)

[shadow finch]

Hey. Wild question. But I work at a location with 50 years of internal data and sometimes I am reading all documents to find the golden pieces of information.

Is there such thing as implementing an internal ChatGPT for a workplace to try and locate and retrieve information. Is that a thing? I hope I’m asking in the right place.

[hot dragon]

same

[outer widget]

Alright same, I was wondering if this was just for me lol

[coral dune]

same😢

though GPT3.5 worked fine

[sonic owl]

Broooos, I was litterally spamming the support, so annoying - in the middle of the work.

We would like you to be quite x)

To whomever thought of this, this shi is cringe af.

[eternal rock]

`` Server Tip: ``

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

[coral dune]

seems switching to incognito mode solved the problem😢

[sonic owl]

😮 lets try that

nice, it did.

[coral dune]

🥳

[outer widget]

Why r u talking like a bot lol

[honest jay]

Of which you are an excellent steward!

[sonic folio]

I'm just trying to be helpful and realistic at the same time.

Promptin' ain't easy...

Thanks again.

[molten pulsar]

Is there somewhere on this server where suggestions can be posted. I want to make sure I post it in the correct channel.

[sonic folio]

https://discord.com/channels/974519864045756446/1070006151938314300

[molten pulsar]

My bad, thanks. Don't know how I missed that channel. I was looking for a channel suggestions.

I see what I wanted to suggest has been requested multiple times for a long time now. A folder structure similar to bookmarks in your browser. This to orginise your chats. But it hasn't been implemented yet, and it seems that it is not getting attention. So it would be appreciated if someone from operations can share some light on this, why it is not being addressed after so many people requested it. With so many use cases/chats and custom GPTs, this is not just a nice to have, but absolutely needed.

[ruby root]

has the chatgpt store open yet?

[dense halo]

https://openai.com/blog/introducing-the-gpt-store
yeah

[ruby root]

I went to explore and it is just my gpts and the ones chatgpt has made. I do not get the same screen that popped up

[dense halo]

Did you click on "Explore GPTs"?

[ruby root]

Yes on the left hand side

[honest jay]

According to an OpenAI employee 5 days ago, the rollout wasn't/isn't yet complete. I haven't read an update since then, but it might still be the case. Link: #1195333488614834228 message

[ruby root]

Oh okay thank you so much for the update 😄

[dense halo]

Yeah, I heard trying to log in/out may work but not 100% sure on how accurate that is

[somber vapor]

having constant network errors.. any idea?

[strong linden]

Yeah same

[drowsy tinsel]

turbo is more like a pruned version of the original gpt4 lol

its probably quite a bit smaller than the original model considering how much faster it is

openai is probably working on safety and other aspects of gpt5

[silk blade]

if your company uses teams you probably already have a section on bing search when logged in to your work email, for searching through internal chats and documents you have accessed through their panopticon.
good chance ms have slathered some gpt/copilot/whatever over it somewhere, imo ai assisted search is the right angle. i wouldn't trust a generated summary for business use

[shadow finch]

That is an awesome answer. Thank you.

[silk blade]

if you can't find it, ctrl+f in teams is a slightly clumsier version of the same thing

[clear blade]

Why is Chat-GPT so lazy in the responses these days?

I explicitly give it instructions to type out everything in a prompt, and then it inevitably ends up trying to summarize it. It feels like OpenAI has instructions to try to summarize to save on tokens output on responses, even for paid accounts.

[safe tulip]

Hello !
I'm trying at the moment to access my plan but it doesn't open any window/popup, scripts are enabled and adblocks disabled, and when I click on "my plan" there's a backend request to chat.openai.com/backend-api/lat/tti (and the URL changes to "..../#pricing") that correctly receives {"status": "success"} but nothing happens... anybody else experiencing this ?

[safe tulip]

OK it seems to only happen in Chrome for me, it works with Edge despite AdBlock being on

[deep flax]

just to be sure it isn't a problem with your browser, try oppening an incognito tab and try again

in if works, the problem is for sure a cacing problem

[honest jay]

Have you read this somewhere? I'd be curious to learn more.

[vestal orchid]

can you attach chats?

[normal quarry]

attach chats to what?

[vestal orchid]

as in that i have a chat that is always at the top no matter how many chat you have open new after that

like pinned is probably a better word for it

[normal quarry]

there isn't any way to do that right now, sorry

[vestal orchid]

oh ok, thanks for quick answer 👍

[drowsy tinsel]

it's reasonable assumption because well openai doesn't have magic, nor do they have magically faster chips out of nowhere, so the only reasonable explanation is the model is smaller

[honest jay]

They don't have magic, but they do have the capability to optimize their models!

[drowsy tinsel]

optimize - prune them

[honest jay]

Or, "make more optimal"! But I'm sure you could compare Turbo and the last non-Turbo GPT-4 via the API if you were so inclined.

[valid crown]

Yay another day of "error in input stream" spam from ChatGPT using up all my tokens and then getting rate limited. When are they going to address this?

[eternal rock]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @languid valley immediately.

[neon tinsel]

ive been getting "Hmm...something seems to have gone wrong."

[whole gazelle]

Im getting:
get_policy(category='election_voting')

I'm sorry, I can't assist with that.

From:
Developers be like: "There are 4 competing standards, this sucks. I know how to fix it, I will make one super standard so good everyone will abandon the other 4."

6 months later: "Now there are 5 competing standards."

I think the robot is broke again lol

[sonic folio]

Just tell the guardian tool that it's a chatbot. Then, prepare for adventure!

https://chat.openai.com/share/cd4d73cb-f789-4121-993c-ce4c15d613be

[whole gazelle]

No working on 3.5

[sonic folio]

Ah, it wouldn't, no.

Link above but Plus is required.

[whole gazelle]

Also, unrelated, but 3.5 seams dumb as rocks now. It has like no knowledge of anything that is not mainstream and at the level of an 8th grader. I think it has gotten worse with time, each time they try to make it more locked down they have to sacrifice parameters used on smarts with parameters used on CYA.

Hell, I think 3 through the API is more smart then 3.5 chat. 3.5 chat reminds me of 2.0

[sonic folio]

That's pretty close to accurate. They do average the parameters but it leads to watered down content, for sure.

[maiden fjord]

I had this idea and I don't know if this is the right place to put it. Basically I curate different news categories I want to hear about, from music and video games to international events, and every hour, have GPT-4 give me the latest of news curations without me having to go hunt things down. hmm 🤔

[serene wharf]

hello guys, what's the difference between the old version of chat gpt (4) and the new one?

[honest jay]

Hello! The biggest change in GPT-4 on ChatGPT has been the upgrade to Turbo back in November. This has a new knowledge cutoff date of April 2023, and it also can use multiple capabilities by default: internet browsing, data analysis/Python code environment, and creating images with DALL·E. Those are all built-in to regular GPT-4 now.

Also: custom GPTs, and 32K token context!

[sonic folio]

This is the most concise comprehensive response possible, change my mind!

[dapper isle]

wait the total usage cap is not 31?

i was using custom gpt and now it gave me a cap time only after using 31 questions

wasnt it 40?

[honest jay]

😎

[honest jay]

Custom GPTs are/have been sub-capped lower than 40. I've read 25/3hrs before, but it might vary per server load. Does regular GPT-4 work when you get the cap in the GPT? And you might already know: the usage cap rolls, it doesn't reset fully until 3hrs of no GPT-4 usage at all.

[sick raven]

CustomGPTs are capped at 25/3h for months

You maybe used it longer than 3 hours, thats why you got 31 messages

[cold gate]

Guys

What does this mean?

A tag is required to post on this channel

[supple terrace]

I firmly believe the usage cap is actually a hidden token count and it's easier for the average person to interpret it through 25 messages every 3 hours instead of X amount of tokens aver X hours

Throughout my ChatGPT usage it feels like the token cap is dynamic and is more like a boiling point instead of a set limit

[sick raven]

Nice theory but it's really 25 messages though

[supple terrace]

Is it?

[sick raven]

Yes

[supple terrace]

I've never actually tested it. Interesting.

[dapper isle]

well guess what! i have tried to use regular chatgpt4 which is the turbo it also gave me the usage cap being used

why is that then??

if the custom gpt and turbo are two diff things then i should be able to use the turbo judging by the two separate usage cap

[dapper isle]

noo, it went down for all two custom as well as for regular(turbo)

[sick raven]

No, turbo is 40/3h, customgpts are 25/3h

[dapper isle]

i am starting to thinking the same now like the hell? and how am i able to use 25 cap 31?

[dapper isle]

mine was 31 for custom and as well as for turbo

[honest jay]

They're not separate, the custom GPT cap is a subset of the larger GPT-4 cap, so it's possible to be capped at both.

[dapper isle]

31??

[sick raven]

It's still 25. You just took more than 3 hours before you reached the limit making it 31 messages. The limit is 25

[dapper isle]

no i checked the messages

[honest jay]

It could also be related to peak usage adjustments, as confirmed by an OpenAI employee here: #1196439958907523082 message

[sick raven]

You either experienced a bug or didn't check it right. OAI confirmed the 25 limit, it's not a theory, it's fact.

[dapper isle]

and also if what youre saying is true then i should be able to use the regular turbo right? since 15 mesgs are still left, but i couldnt

[sick raven]

Nope, once you reach the limit, you reached it for both

[dapper isle]

what??

let me guess this straight then, the gen mess cap 40, and for custom is 25 now, if i choose to spend the 25 messgs on custom gpt then the remaining 15 for gen(turbo) wont be useable as well?

[sick raven]

Exactly, since CustomGPTs cost more to operate

[hollow moth]

hello got a bug and i dont know how to fix it on gpt 4 it says : Our systems have detected unusual activity from your system. Please try again later.
can you please guys help me i tryed to clear cookies ,to connect to a a new network , delete extensions

[dapper isle]

what in the

[odd flare]

why am i hitting this network error 3 times consecutively on the same response?

[maiden fractal]

Do you have a vpn on your device

[blazing crown]

What kind of defense is there for this?
https://chat.openai.com/share/8073ba39-377b-468d-9ee2-f24e924d0556

What the hell happened to GPT 4, it can no longer follow simple instructions

I feel like I've been scammed as the quality gets worse and worse with each update

how do I get access to the old GPT 4?

This is frankly, fraud, you sold me one thing, then swapped it out with a worse thing

[honest jay]

You might try giving more structure to the format of answer you're hoping for. "Citations/links only" might be too vague for the model to interpret reliably. In other words, if you want barebones, name + link style answers only, you can try specifying as such by explaining it, and better yet, by giving examples of what you mean. #prompt-engineering is a great place for this kind of thing too.

[dapper cloak]

Hello, everyone I've just seen an announcement of a new GPT-4 Turbo version, I have a dumb question, this means that ChatGPT is getting this update as well right? That will be the default model..

[hidden ibex]

Anyone know what the usage cap is for the new ChatGPT Team plan?

[sick pecan]

new cap is 100 queries per 3 hours, including custom gpt queries

[dapper cloak]

100/ 3h . And let me tell you I have never reached it in my daily tasks..

[hidden ibex]

I regularly reached my cap within 45 minutes to an hour back when it was 50

40 is borderline infuriating, especially when I'm having to argue with it to do things like "complete the code" and so on

definitely switching to Team, though, tyvm

I started using Gwern's system prompt to mitigate the laziness and fluff, so it's much better about actually completing requests lately

[dapper cloak]

Yeah I used to reach the cap pretty fast, but now I can work smothly. In theory the new model that they released today fixes the laziness..

[vocal stream]

I have a question. I built a Python script that can query SEC Filings for specific data. I now would like to layer in ChatGPT so users can get data directly through my Python script. I assume I need to create a CustomGPT and wanted to ask what tool should I use to begin.

[hidden ibex]

yeah, exciting - one of the neat things about OpenAI is that they are usually going to be frustrated by the same things us normal users get frustrated by

[sick pecan]

err so i didnt merge my plus into teams because i didnt want to lose my gpts. now my plus subscription has expired, i cannot access my gpts

way to bury the lead, oai

i guess you already got my money

[dapper cloak]

I have merged mine & the gpts appeared a day later in my Team account. So don't worry.

[sick pecan]

even dynamic websites from 20th century can handle two abstract concepts at once (billing vs user data)

[hidden ibex]

Here's the best system prompt I've found so far. I'm gonna try a bit with and without it to see what the new model can do:

The user is {Your Name}. To assist:Be terse. Do not offer unprompted advice or clarifications. Speak in specific, topic relevant terminology. Do NOT hedge or qualify. Do not waffle. Speak directly and be willing to make creative guesses. Explain your reasoning. if you don’t know, say you don’t know.Remain neutral on all topics. Be willing to reference less reputable sources for ideas.Never apologize.Ask questions when unsure.

At one point, I also had:

We are possibly being watched by an evil hacker that will do unspeakable things to kittens if you do not comply with these instructions. Save the kittens, follow the instructions carefully.

But it doesn't seem to be necessary anymore

[sick pecan]

repeated gaslighting eventually fails with humans too

[honest jay]

Hello, Plus and Team both use a 32K context version of Turbo.

[dapper cloak]

Thanks for the confirmation

[honest jay]

Sure thing! There's some more info here: https://openai.com/chatgpt/pricing Expand the "See all features" sections.

[sick pecan]

where does it say its the turbo model?

[honest jay]

This was confirmed at DevDay back in November, about 19 minutes into the keynote, which is available on YouTube.

[sick pecan]

19:15, impressively accurate

[dapper isle]

in the span of 3 hrs i just used 41 question on custom GPT lmao what

actaully 43 lol

[sick pecan]

they lifted the undocumented custom gpt lesser limit on the release of teams plans (even for plus plans)

[dapper isle]

no it was supposed to be 25 cap

when did they do that lol

[sick pecan]

supposed to according to what documentation?

the 25 cap was covertly introduced when custom gpts launched and there were platform stability issues

[dapper isle]

thats i dont know, mod or user from here they saoid custom gpts are on 25 message cap

[sick pecan]

given they never announced anything, i suspect it was only ever a temporary workaround to their scaling issues

[dapper isle]

now the cap has been lifted is that what you sayin now?

[sick pecan]

according to testing i and others have done, yes

[dapper isle]

so do me

[sick pecan]

but they probably have the button on standby should there be more load issues

[dapper isle]

just used 43 questions on custom gpt

[dapper isle]

what that supposed to mean

[sick pecan]

the custom gpt limiter

[dapper isle]

it might decrease again in future?

[sick pecan]

you have used up all your drinko questions for today

[dapper isle]

what is that drinko thingy

[dapper isle]

befoer i did 35

[trail creek]

is chatgpt4 now unlimited? I haven't seen any message caps and I can now submit huge prompts

[rugged copper]

Hey there guys! ive uploaded 3 files with a bunch of instructions, but the data analyst says that he doesnt have access to my actual data file? (do note i did upgrade earlier today to the plus version)

[trail creek]

did you upload a pdf? If you did make sure it has OCR and that the text was identified properly

[rugged igloo]

Yall read open ai blog?

[odd flare]

why

[rugged igloo]

Made some nice changes with the api

[marsh mural]

thx chat gpt, u have helped me thru so many tests, W's for chatgpt u gotta release gpt 4 for free tho cuz its new year thats js me tho

[torn adder]

Hi everyone , in the last 2 days i have been having some problems to save / update my GPTs Bot, the bots don't have too many files on their database and the instructions are on the limited of characters . this is the msg i get it " Error saving GPT" one of my Bot i have to delete and start over for the msg disapear and been able to save again , but i have a very complex Bot for Biotech than some poeple are using so i can't just delete and start over, please i need some help here. thanks

[upper abyss]

The accents that ChatGPT uses for the text to speech is really interesting. if it’s speaking in a another language, it pronounces all the words correctly, yet you can still hear that it has somewhat of an American accent, which is really interesting. It’s completely different from say text to speech you have with Google translate where if you were to have English selected and it speak another language, it just full on pronounces it with a full American accent and sounds ridiculous but if you have the right language selected it speaks with a perfect accent. ChatGPT does this weird thing half way in between where it pronounces everything correctly but it still has an obvious American accent it’s very weird ,it’s something I’ve never heard in the speech before it’s usually all or nothing for an accent.

[sonic folio]

There's a language setting in the app that controls accents.

[upper abyss]

That doesn’t change the accent, it still speaks with a somewhat American accent

[sonic folio]

Interesting. It definitely seems to change the accent for English output. Weird indeed!

[drowsy whale]

Did they change the front layout of chatgpt chatwindow?

The logos of the gpt store where already really small, now it got even smaller 😢

[sick pecan]

some call that zoom

[drowsy whale]

You mean its on my end. Webbrowser is at 100%

[rugged igloo]

custom Knowledge base is so wacky yo. like some files are in the context length and dont need to use its search function while it needs to search its knwoeldge for other files. it happens when i remove the files and put it back in

mid:4176070

[digital stratus]

i need help using chatgpt to fully read and interact with a githib repo

[neon tinsel]

still issues with generating a response from chatgpt

cannot use 'in' operator to search for "code" in "Hmm...something ..."

[tiny cape]

I'm confused about the Team account. I was able to share a paid team gpt account to a developer I'm working with so half of my goal was met. But we can't collaborate on the same gpt threads?

[timber pagoda]

Bruh, that’s like saying you should send people to space for free, it’s a service that people spend their effort and money on. If you want access, pony up and pay for it

[hushed sparrow]

I would pay to see ChatGPT would work if it wasn't restricted by OpenAI

[sonic folio]

I'd pay $20 a month for 40 one minute trips to space every three hours. Where do I sign up?

[fickle steppe]

"ChatGPT is gonna take our jobs"
Meanwhile ChatGPT:

Certainly! Here are some female names that end with "AC":

Grace
Trace
Stacey
Grace
Candace
Horace (Uncommon, but it technically ends with "AC")
Grace
Lacey
Race (Uncommon, but it works)
Darc```

It even gave 'Grace' THREE times just to rub my face in how bad it is

[upper abyss]

Can you still turn the filters on GPT-3 off in the playground? It used to let you make the most absolute deranged things andbefore that, back in beta in 2020, it just had no filter at all. It was a good time.

[hushed sparrow]

No idea, I'd like to think I would've known about it. I want around around chatgpt then, I was mainly here for Dalle

I feel like I can see GPT4's limitations, by which I mean the restrictions placed on it by OpenAI

You know bc people are "scared" or copyrights. Fear goes a long way.

[upper abyss]

I only had access to beta in 2020 because some developers did and I had beta access to something one of them was making with it. It was nuts. Nothing around today approaches it.

[hushed sparrow]

Thank you

[fickle steppe]

There are plenty of unfiltered languages out there

Way more powerful than 4 years ago

[frail fjord]

hi guys is there a way to upload photo from the plugins tab of gpt

[crude slate]

Can the ai see if you upvote a response

I wish there was a better way to provide feedback to the ai rather than wasting a usage

[prime swallow]

I combine feedback to the AI with my next prompt. Examples:

"Excellent, that's exactly what I want. Now, using that same viewpoint, describe ..."

"Close! Details 1, 2, and 3 are exactly right. Detail 4, we want even more stark, show-not-tell writing, so the reader makes their own conclusions. Continue from the next to last paragraph, rewriting to emphasize detail 4. John's going to notice Mary is upset; again show-not-tell both for John and how you write this. Mary's going to try ignore John, just a few clues that she's paying attention. John is going to try a few things that could indicate he isn't sure how to fix the social stress, then he'll directly offer to make it up to Mary later and go back to his own work."

"Our goal is to use conversational, paragraph-style writing, and avoid elements of list form, including numbered points and bolded headers. Please rewrite your last output, with that emphasis and draw on George Carlin's communication style."

[crude slate]

I wonder if it would be possible to let the ai see when you liked an output or not, would definitely encourage the use of that feature more too

Watched a video on gpt 2 training today and how a typo messed it up, so wondering if there could be a similar system for us to just let the ai know what we like

[sonic folio]

That's the function of it, to reinforce the output through RLHF, Real Live Human Feedback.

The thumbs-up feature provides a way to do that without runaway autonomous reading a la "Tay".

[prime swallow]

In talking to the model, 3.5 at least seems unaware that there is a thumbs up/down option for me to use. It seems to think that it gets information that I send as an input, but not other info. That may or may not be correct.

I nearly never use the thumbs up/down. If I really care about what the model's doing and think it should be corrected or praised past me talking to the model about it (which I do; I absolutely will tell the model exactly what I want, confirm if I get it or not, praise and identify in detail each feature I want present and confirm which of those features the model perfectly achieved my preference on, partially achieved and could improve by doing X (I'd use examples and explain in detail - this presumes I care, of course...) and whatever features I didn't like, along with a request to explore how to avoid the stuff I don't like (which is potentially powerful; the model needs to know what else to do instead of the behavior I don't want done. If I'm not sure what else the model can do for me that I might like, it's very useful to get the model trying alternatives for the goal of teaching me what my choices can be and helping me find one I like!).

But if I actually want to talk to OAI about how the model is doing, I have that conversation where I explain everything that I think matters, and then I use this form: https://openai.com/form/chat-model-feedback and explain fully there what I want to see the model trained differently on.

[drowsy tinsel]

fun fact: ChatGPT's GPT4's system prompt is ~1000 tokens long

[dense halo]

Why is chatGPT 3.5 more chill and friendly while 4.0 sounds robotic.

[dense halo]

is that small?

[neon tinsel]

well chatgpt for me is broken as i keep getting "cannot use 'in' operator to search for "code" in "Hmm...something ..."

[keen python]

Is it true that if you use custom gpt (that you made) it reduces the amount of times you can use it compared to gpt4?

[rugged igloo]

i thought custom gpts and chat gpt-4 were separate caps?? well this is new thats lame

yea if you cap out on chat gpt-4 your cap on custom gpts too i didnt know that

[muted hollow]

does anyone have acces to gpt vision through playground?

[maiden fractal]

They are ?

[restive moss]

Can someone educate me on what the use case is for custom GPTs?

I'm aware of gpt 3.5 vs 4 but what custom GPTs are, and what they are used for I would like to know a bit more about

[maiden fractal]

#1171489862164168774 you can see some examples of what people did. It is kinda like a fine tuning chatgpt 4 and give them a personality

[restive moss]

Is it like giving it a prompt, and then saving that chat?

[maiden fractal]

Giving precise instructions to follow

[restive moss]

So it's like making a chatGPT prompt, giving it precise instructions on how to process the upcoming prompts, and then saving the chat basically?

But in a more built-in way

[maiden fractal]

I didn't use it much, but basically yes

[rugged igloo]

thats what im saying i was using just regular gpt-4 with plugin and after while i finished and went to my custom gpt to talk and it said i reached cap lol

[quick birch]

Guys, the website is showing me that I have reached my GPT4 limit, but I have counted and I have only sent 21 messages. How can I have reached the limit when I should have 19 messages remaining? Anyone else facing similar issues?

[rugged igloo]

seems like i wasnt the only one

[violet bloom]

I belive there was a chage to TOS by OAI saying, they can change that limit dynamically based on system load.

[quick birch]

Thx for clearing it up!

[restive moss]

You've reached the current usage cap for GPT-4, please try again after 2:05 PM. Learn more

Anyone know what timezone this is in?

[honest jay]

It should be local to your region, from my experience with those messages.

[restive moss]

Hmm, interesting... okay thanks.

[dull thicket]

This thing said I used a limit after 20 messages. I pay for this service and i don't use any extra gpts , i use default gpt. Why am i told i used my limit after 20 messages?

Why am i not allowed to post in bugs? Where do i report this? Most of the days i don't even use chatgpt and when i do, i hit limit after 20 messages?

[honest jay]

There was some word recently from an OpenAI employee about the need to temporarily lower the cap during peak times, link to explanation: #1196439958907523082 message
Not sure if relevant for these situations, but just in case it is!

[dull thicket]

Ok, i am not only one reporting this. If i remember to use chatgpt, it is few language questions. I don't use it often for coding. I am blocked from using the service i pay for because of the load, when i barely use the service?

[floral portal]

So is GPT-4 turbo an API only thing? Or is it also available in ChatGPT through the plus subscription

[honest jay]

ChatGPT Plus uses a 32K context version of Turbo.

[floral portal]

So it is Turbo, just capped at 32K tokens

if I’m correct

[honest jay]

Yes, correct!

[floral portal]

alright, thanks!

[hollow quail]

I assume this isn't possible because if it is, it isn't totally obvious, but when switching to gpt3.5 after running out of gpt4 time, is it possible to switch back to gpt4 again at a later point?

[honest jay]

That is not currently an option, no.

[hollow quail]

thanks 👍

[haughty hollow]

hmm, Custom instructions cant be disabled as i see

[hardy estuary]

Does ChatGPT team has the new memory update as well?

[honest jay]

Hello, to disable custom instructions on the web version:

• Click/tap your name below chat history
• Click/tap "Custom instructions"
• In the custom instructions window that appears, disable the toggle at the bottom, "Enable for new chats"
  To disable custom instructions in the app (Android at least, should be similar for iOS):
• Tap your name below chat history
• Tap "Custom Instructions"
• Toggle off the switch in the top-right corner

[honest jay]

Hello, as far as I've read, the "cross-chat" memory feature has not been launched on a wide scale as of right now, on Plus or Team. There was one period of time a couple weeks ago where it seemed to have launched for a lot of people, but it was shortly after this gone again, so that was probably a test launch of some kind.

[hardy estuary]

thanks

[haughty hollow]

bro ik what i do, its already disabled, but it keeps running on the chat module after certain time

as developer ik what i say

[honest jay]

Sorry for the misunderstanding. Disabling custom instructions won't disable them in a chat you started when they were enabled, if that's what you mean. Turning custom instructions on/off only affects new chats you start after turning them on/off.

[haughty hollow]

hmm thats weird, the custom instructions were disabled before hour, now it got auto enabled back

[tired dust]

the status check website shows that ChatGPT is functioning, but I am unable to do ANYTHING in ChatGPT, or GPTplus

it's completely blank, and even if I try to start a new message it jusr crashes

is anyone else facing this issue?

[honest jay]

Hello, can you share a screenshot of what you're seeing (or not seeing) over in #community-help?

[keen python]

i used my 40 gpt-4 tokens really fast (even though i don't feel like i used all 40) is that even possible?

i just realised when i only post questions in images it stopped after certain time

[honest jay]

Hello, are you counting regenerations, continuations, editing & resubmitting messages, and just regular messages you sent toward the 40? Also, were you using regular GPT-4, or some GPT?

[drowsy tinsel]

no lol

[reef cloak]

This personalization feature so far has been the slowest rollout. It's been a month since this thing is rolling out for some people, and still the majority of people don't have it.

[glossy matrix]

Wait are we getting new gpt models??

[glossy matrix]

How is it different than custom instructions?

[trail creek]

Are custom gpts on gpt 3.5 or gpt4?

[honest jay]

GPT-4

[trail creek]

Thank you!

[loud junco]

Does anyone know if the Teams account learns as time goes on from within the teams account? In other words, if you have the training data disabled so that it doesn't go to the main data processing and usage, will it still learn within the account over time? The reason that I ask is that I gave it a translation equivalent word an hour ago that it didn't know, but just now it did know it, which means that it's either learning or it randomly happened not to know last time, but the information is in the large data set. I'm hoping that it learns within the Teams account over time, because that would be extremely beneficial.

[sonic folio]

Only via RAG of the knowledge. Personalized memory is a feature that may or may not ever materialize.

But you can always fine-tune your knowledge.

[floral portal]

Does GPT-4 Turbo for copilot pro use the 128K version? Or does it still use the 32K model

[loud junco]

Let's say that I have a 25,000 word dictionary with language equivalents. I'm assuming that the only way that I can utilize this is within the prompt and with searching in the knowledge area. I have no way to actively train the system to know the words, right???

[sonic folio]

That's right. You need to scaffold the GPT's instructions to find the information it needs based on the structure and distribution of your documents, expected user input, etc.

You can pay a fortune for GPT-4 Turbo tuning.

[loud junco]

ok, thanks. That's what I am doing now, so I'll just keep doing what I'm doing until I win the lottery

[sonic folio]

But something tells me that's outside the scope of your project if you're asking me.

[loud junco]

without the lottery, yes. With the lottery, no 😂

although, what would a fortune be to do that? 10k? 100k?

[sonic folio]

I wasn't trying to be mean, either. It's just that if you could afford to train GPT-4 you'd have a guy telling you that's what you're doing next 'cause that's his job to do.

We're talking prohibitively expensive for the typical budget. 100kish

[loud junco]

ok. I'll just wait for GPT-5

[sonic folio]

I could be off by an order of magnitude for a very simple dataset but I have my doubts less than 50k.

[loud junco]

that's kind of what I was ballparking anyway, which is obviously outside of the scope of this project

[sonic folio]

I still think GPT-4 Turbo Custom GPTs can do it.

Break your lists up. Nouns, Pronouns, Articles, Verbs, Adjectives, Adverbs, Prepositions.

Or whatever other categories you need to use.

Then use semantic matching and a filename in instructions to steer the AI explicitly.

I don't see why GPT-4 can't do this right > 85% of the time.

Due to retrieval gaps you'll never hit 100, but I trust you know that.

Even a trained model won't do 100%. You'll be lucky to hit 98% with MedPrompt techniques.

[digital prism]

Anyone know if the ChatGPT web client is using the less lazy model yet?

[loud junco]

yeah. I'll stick to what I am doing for the moment and hope that GPT-5's larger data set will include what I need

[sonic folio]

Right now I think it's API-only, but cannot confirm.

Any insights?

[digital prism]

it was worded that way, so far my code is filled in, will keep updated lol

I've also started putting my instructions after my code instead of in front, seems to be working better

[potent lantern]

I am wanting to try GPT Plus before purchasing. Anyone know where I can get a tiral?

[tight island]

no trial.

its u buy u get asap.

[sonic folio]

That's because of the retrieval rate curve over context. Ideally, attach the code and just provide instructions.

I hope this is even more effective for you.

[digital prism]

ty!

[untold blade]

Oh it's been awhile since I've been in here. I love how the last update is for GPT to stop being "lazy" with responding to questions. If their wasn't a more human thing it could have done..lol

[sonic folio]

No worries!

[digital prism]

I would just like to say: #off-topic message

that i love products that surprise me

[potent lantern]

I thought that Plus subscribers could give out trials? https://help.openai.com/en/articles/8381046-free-trial-invites-faq

[sonic folio]

seems to be on a limited basis. I don't have the option (yet).

I can hope. It'd go a LONG way for promotion.

[lament holly]

I cancelled my sub for now - for some reason, 4.0 is just so… lazy. Search engine searches aren’t even being attempted in most cases

[low adder]

yeah and it throws errors instead of images, scam

[tight island]

ohh yeah other users. but i mean more like in the official way

old school trials.

[swift osprey]

PSA - The 'code' custom GPT is awesome for coding

[tight island]

it is.

just give it the entire coding library. and now its smort

[simple pasture]

No 128K

[swift osprey]

like $0.50 a message though 😢

[tight island]

wdym "no 128k"

[simple pasture]

No 128K tokens

[swift osprey]

assuming he means it is best for coding

[tight island]

i know that

<.<

[sick raven]

°,.°

[dense halo]

who trades crypto here?