#chatgpt-discussions

1 messages · Page 77 of 1

inland heron
#

making things up is not right

#

@spice star 😄

#

there was no popup, not for me or anyone I talked to

honest jay
#

The little "i" icon in the ChatGPT model selector dropdown links to a help article explaining the base limit of 40/3hrs and the dynamic cap during high traffic.

inland heron
#

I haven't had 40 per 3 hours in a long time, this is like saying 'You may live up to 200 years' with an asterisk saying (if you suffer from a disease called 'getting old' you may not achieve this and it will be a variable lifespan for you)

#

it's all a marketing tactic, they witheld this information, they now have it with 'may' and 'ifs', but again they are now removing the information about when the limit resets, similar to before

#

unfortunately, Gemini and Sora are either very lacking or not available with not enough context window, so we can't choose much. For my needs, none of the other models are sufficient

honest jay
# inland heron it's all a marketing tactic, they witheld this information, they now have it wit...

When it comes to marketing, the actual pricing page doesn't list a specific limit at all, only "Access to GPT-4" for Plus and "Higher message caps" for Team: https://openai.com/chatgpt/pricing
The only place 40 is mentioned (I think) is in the help article, where the dynamic cap is also explained: https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4

From my anecdotal experience, I have gotten a full 40 uses in a 3hr window, I was testing it a few weeks ago I think? Perhaps a month. I haven't tested it more recently though!

eternal rockBOT
#

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.

storm garnet
#

gpt4 playing up? keeps using 3.5 even tho not used today

knotty crypt
#

I am having that message:

New responses will use GPT-3.5 until your GPT-4 limit resets.

but idk when gpt4 will reset, when? where to see it?

tawny whale
#

I get that message as well, and cant see when I´ll get GPT-4 . It usually says a time when access is granted again

knotty crypt
#

yeah, same

cerulean oxide
#

I get that's what they said and what you and others may have been seeing.

But I only capped the GPT without ever using 'normal GPT4' and had both capped out.

If it really was a lower limit for GPTs than GPT4, I should still have some usages for GPT4, no?

(Not quipping at you btw just saying it doesn't add up)

#

If I have a day where I'm focused on working on testing ChatGPT, plan my breaks around the time I usually had to wait anyways.

OpenAI just wants me to sit on my thumbs? Lol

dense halo
#

Is ChatGPT voice down?

weary orbit
#

Seems ChatGPT + API models LOVE outputting 'However' lately.. not a very productive and very unwanted change imo

weary orbit
knotty crypt
ember otter
#

Well Gpt 4 can't read my lotto scratchers worth a damn

pearl swallow
#

hello

vapid kettle
weary orbit
#

Pretty sure it happened super recently, at least for the API version

vapid kettle
weary orbit
#

I can tell you with confidence since gpt-4 non-turbo versions seem to be doing it now as well

#

it's not just the model that determines their output. Sampler changes or classifiers might jump in to affect model output

vapid kettle
#

All model checkpoints are frozen. When they do get updated the date gets changed. And when they want to replace they deprecate.

weary orbit
#

the checkpoints sure, but nothing says the sampler or classifiers have to stay the same

#

anyway I just wanted to leave this here as feedback in case OpenAI staff were interested in community feedback.. I might be wrong, but I doubt it

vapid kettle
#

what? classifier is for moderation. As for the "sampler", this doesn't make much sense and I'm not even completely sure what you're referring to. This is not stable diffusion.

weary orbit
#

the flow to get a response is this user --> input --> model --> logits --> sampler --> text --> user

vapid kettle
#

pretty sure all of that is included in the weights which are frozen. You can only change random seed and that's about it but this won't change the model much with the amount of fine-tuning it went through.

#

Plus it wouldn't make any sense to mess with it for the older checkpoints as changing the random seed is... random and will lead to random results. No point in doing that and screwing it up.

weary orbit
#

anyway didn't mean to make this technical at all 😛 but it's possible I might just have been lucky and missed that gpt-4-0613 always used to say 'However' or love twists, like gpt-turbo does

vapid kettle
#

basically the chances of the old models/checkpoints ever changing are so incredibly slim they are not even worth entertaining. All conspiracy theories of this happening can be easily debunked with A/B testing (reproducing the outputs from the distant past) as well. In fact I did it personally more than once lol

weary orbit
#

however you should know that's not the case with the sampler -- it's unrelated to model weights and more related to server/endpoints

vapid kettle
#

one of them was "write a story without any letters A".

The performance now is basically exactly the same it was 1 year ago. It kinda weird we need to be proving that in the first place if you ask me.

vapid kettle
#

What you can do is change top_p, temperature, repetition_penalty etc... but every API user has control to do it independently

weary orbit
#

you forgot the most important one -- logit_bias :p

sick pecan
#

the model almost certainly doesnt change, as anyone using the API will attestify. and we arent really notified of if/when/how they change the chatgpt settings. Occam's razor would suggest that any changes experienced via chatgpt are almost certainly chatgpt configuration changes, not the underlying model (except for when they update chatgpt to use the latest model)

weary orbit
#

and mind that's a very simple and easy to provide to the end-user parameter -- with the source code the logits can be fully controlled by the sampler

vapid kettle
#

And if you want to believe some conspiracy theory that they changed the default values, you can always use older openai python lib 👀

cerulean oxide
vapid kettle
cerulean oxide
vapid kettle
#

weird, maybe they changed it further. Didn't encounter smth like that yet.

#

or a cache issue / bug perhaps

cerulean oxide
#

I despise being the guy who comes in and complains over little things, and especially hate it when I myself missed something when I get to this point.
I always try to see if it's something I can do without ever talking to customer support, or mentioning in a server like this. I usually go and search the discord to see if anything has been stated too.

When I am here and complaining it's because I haven't found the solution after I have done my due dillegence (at least when I feel like I have done all I can)

#

And their customer service response is horrendous. I will always get automatic generalized tech support answers and then never get notified when a real person contacts me. When I finally do respond, with screenshots (not this issue but other ones I have had) I get ignored

vapid kettle
#

--
hmm for me it's still pretty normal even after clearing the cache and relogging:

You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after 11:30 PM. Learn more

@cerulean oxide

cerulean oxide
#

You used base GPT 4 tho yea?

vapid kettle
#

are you on mobile by any chance?

vapid kettle
cerulean oxide
#

No I am on desktop with no VPNs, caches cleared.

My issue happened when testing a custom GPT. Which then capped GPT 4 as well

vapid kettle
sick pecan
#

i noticed a single prompt can cost multiple. this started back when initially doubled 25 to 50. i was already maxing out the cognition using various prompting techniques (hitting max_tokens as often as i could).
now we get the continue button far more frequently, and this too costs a prompt.

cerulean oxide
#

I am always gonna receive the basic tech support troubleshoot list. when I already do all that before departing my solo attempts to resolve

sick pecan
#

so if you seem to be hitting the limit earlier, then reflect on how long the responses were

#

also re testing custom gpt you should always be using the preview pane in the gpt builder. a lesser known tip is that this does not count towards your 'gpt cap', only your overall cap

#

ie gpt devs shouldnt be punished

cerulean oxide
#

I was literally using that preview plane when it gave me the message

sick pecan
#

you got hit by the overall cap, yes

cerulean oxide
#

to which it still didn't explain when I would get it back

#

Stop responding to things that I am not worried about. I am not worried about the cap. I am worried about CLARITY when using the system

sick pecan
#

okay so you are hostile. i will just finish my advice (so i can reference it later), by suggesting if all of the above doesnt explain why, then chances are the account has been compromised.

vapid kettle
cerulean oxide
#

hostile? no I am clarifying

south kayak
#

when GPT 5 release?

rustic fiber
ocean escarp
#

hmm.. I'll try. It would be challenging because I tried to make it sounds authentic, but it's still pushing their predictable language

#

We gonna have Sora before GPT 5 lol

sick pecan
#

if you can identify what you consider to be inauthentic, and can calculate the juxtaposing terms, you can typically refine the prompt so that it works

#

for me its uppercase letters, being professional, simple language and short sentences.

ocean escarp
#

To understand what you're getting at, let's choose a common term GPT used, tapestry. What do I do to typically refine the prompt to make it work?

sick pecan
#

so mine would be: lowercase letters, acting casual, overly complex language and meandering sentences.

#

ohyeah tapestry comes up often. i also amend my custom instructions with a banned word list (for not-tech stufff)

#

not sure if it helps or hinders, but it makes me feel good

ocean escarp
#

So, I have to make a list of words to refrain GPT from using them again?

sick pecan
#

it will become less effective the longer the list

#

but tapestry is definitely a signal to the reader that chatgpt was used

#

nobody says tapestry

#

maybe we need a chatgpt post processor to apply synonyms to most predictable chatgpt words

ocean escarp
#

Yeah, GPT even used it way before GPT 4 released

sick pecan
#

if i am reading someones supposed writing and i see tapestry, i just close the tab

ocean escarp
#

loll

#

We had enough reading from GPT's long messages, we don't need any more of reading

#

But yeah, it's difficult to write authentic writing with GPT. It's not authentic, vague with context, and write too much words that don't have value into the writing

glass fulcrum
ocean escarp
#

rip OpenAI servers

#

GPT 4 is lagging there and there, these would've destroyed the server farm lol

glass fulcrum
ocean escarp
#

really?

glass fulcrum
#

well it takes time to assable that....

ocean escarp
#

I hope GPT-4 will lag less soon

tight island
#

it's coming...

#

soon...

#

i can feel it in my tummy

#

let the OAI Devs Cook.

glass fulcrum
#

tummy never lies 👀

tight island
#

It’s gonna be big this time

#

I can just feel it gnawing in my soul

prime swallow
#

It's been big every time 🙂

blissful grotto
prime swallow
honest jay
umbral yarrow
#

Anyone else don’t have anymore since a few days ago the “Select Text” feature on the ChatGPT App on iPhone?

ivory iris
#

are other people having problems with the new memory feature? It's not remembering anything and it's enabled

wind cradle
#

I also experienced a similar issue, but in my case, turning off "Settings -> Personalization -> Memory" stopped the caps from filling up immediately. This does disable the memory function, though...

ivory iris
#

what's the best way to contact openai about a typo on their site?

tall wind
#

I have a question, i was conversing with gpt 4 and said a few things, and it did something I hadnt noticed before

#

was trying to get advice on what this is.

#

there was a loading icon of sorts, and " memory updated" status under the gpt name.

#

I think this phrase I commented preceeded it doing that " I'd like you to use it in the future " .

#

wonder, what does this mean exactly " memory updated" ?

rancid otter
regal portal
#

hi, i saw a someone train a bot using discord messages of a friend to make the bot sound like him. Is that something that can be done with chatgpt? And can i add that bot to a personal server?

tall wind
#

i think people are using it to train personal llms, to run local.

#

that'd be trained on ur data

#

say like Lama 3, or something, or whatever models are out there to run locally

#

then train it with ur data.

regal portal
#

thank you

sick pecan
#

i tried a simple discord bot a year ago that just uses the chat history to impersonate. github project is named, "Urist-Mc-Urist/DoppeLLaMA"

regal portal
#

did it do a good job referncing your discord chat history?

sick pecan
#

is fun for all of five minutes

regal portal
#

ah

#

i'll still give it a try just to learn how this stuff works.

#

by the way, is the browser version of chatgpt freezing up for anyone else?

#

won't even load some of the conversations i had with it

still karma
#

Question with the new memory feature what commands can make chatgpt smarter. Due to the censorship algorithm making it worse each update what can we say to make chatgpt more functional

ivory iris
lost lagoon
#

what is happening to chatgpt it keeps pushing me back to 3.5 even when i selected gpt 4 to write my article ? it auto jumped to 3.5 ????

sour wasp
#

ChatGPT down again?

#

It keeps saying server is having problems for everything.

inland loom
#

down for me

faint thistle
#

seems so? also got problems here

dense void
#

here we go again

viral wyvern
#

its bugging for me

nova fable
#

Yep, issues

#

Came here to echochamber 😉

regal portal
#

same, and just subbed for gpt4 😄

foggy hill
#

chat gpt down for yall?

nova fable
#

Must have been you 😛

nocturne heath
#

Im gonna cry I have an important presentation to prepare for bro

#

Thats so sad

regal portal
#

sorry guys! I'll unsub. You'll get your gpt back in a moment

viral wyvern
#

try bing ai

nocturne heath
latent briar
#

bruuuuh why is my best friend down again 😦

regal portal
viral wyvern
sour wasp
#

10 times dumber than ChatGPT

viral wyvern
#

but yea it is still worse tho

#

i think its funny tho

icy compass
#

did chatgpt just die

latent briar
#

it did

viscid dune
#

c'mon, during finals week? thumbsup

latent briar
#

let us moarn, procrastination will not go up

icy compass
#

I got my english summative presentation tmrw

#

im cooked if chatgpt is dead

nocturne heath
viral wyvern
#

try using ur brain

icy compass
regal portal
latent briar
#

lets all be friends

icy compass
nocturne heath
regal portal
#

human friends? Are you insane??

viral wyvern
#

chagpt is my best friend

latent briar
#

indeed

icy compass
#

me personally, my brain dont function when it come to writing

regal portal
#

name one human that did anything useful or entertaining ever

icy compass
#

which is why chatgpt is the goat

latent briar
#

me

regal portal
#

touche'

latent briar
#

I did a thing today

regal portal
#

did you flush it?

viral wyvern
#

unironically tho ive talked to chatgpt more than anyone else in my life combined

nocturne heath
latent briar
#

i pleade the 5th

icy compass
regal portal
#

before chagpt, we would put books under the pillow at night and have everything memorized by the time we woke up and started our 20 mile uphill walk to school.

pliant knot
#

Is it down right now?

steady dock
#

is chatgpt down

nocturne heath
nova fable
#

Tis

icy compass
#

bing ai time

regal portal
#

would be funny if openai shut down chatgpt during exam season 😄

nocturne heath
pliant knot
#

can you use chat gpt to wash dishes?

regal portal
#

it's not a woman

icy compass
#

but at least it gives sources, very good for research

ornate zealot
#

The mobile app is getting error 111… That’s just great.

icy compass
#

God heal chatgpt 🙏

steady dock
nocturne heath
elder pollen
#

There’s a bunch load of these recently
upstream connect error reset reason remote connect failure delayed connect error 111

west ridge
#

ChatGPT down or something?

elder pollen
#

Ah. Global then.

plush mural
#

Gemini is down too

west ridge
#

Oh come on…

icy compass
#

bing ai still standing

strong wagon
#

I was in the zone 😭

ornate zealot
visual bloom
#

Reasons (speculation) - they changed webaddress (now chatgpt.com). Does load, but not always in full. iOS app works.

ornate zealot
#

IOS app doesn’t work.

strong wagon
#

yes

regal portal
#

#@nocturne heathdidnothingwrong

strong wagon
#

for good i fear

west ridge
#

yeah, iOS doesn’t work for me too

icy compass
#

11 minutes of ChatGPT being down. I think i'm losing my mind. My whole body aches and my limbs are trembling. I feel my bones breaking and i'm in a straight panic attack because i have to utilize my brainpower to complete my assignments. I feel my organs degenerate and i'm losing power. Please, ChatGPT CEO, put your AI back up. I'm dying and in endless eternal pain. I don't think i'm going to make it. How many years?…. Months even has it been without ChatGPT ? I still can’t comprehend my only reason of life has been stripped away from me, my body has gone into withdrawal, my hands are currently shaking as I am currently typing this. My brain has no purpose other than to fulfill the duty of using ChatGPT . What do i do now? I see no purpose of life… There is no purpose of life. I am going to send one final message before I cease to exist. I’m feeling sick to my stomach, my body is sick, my legs are shaking. I’m about to throw up. This is a disease, a mental disorder and disease that has taken my life away, Hope you’re happy now… ChatGPT .

woeful scaffold
#

Lemmie guess, y'all (devs) tried to migrate domains and now you're dealing with a whole bunch of session and authentication issues that you thought were well-tested on staging right? Well, avoiding that is impossible when doing this, so have fun 😂

regal portal
#

wait let me have chatgpt read that wall of text and..oh...wait..

strong wagon
#

😭 I feel for you bro

west ridge
#

And I open the website, it shows nothing

elder pollen
#

iOS is what throws those errors.
Guess someone injected bodily fluids into the drives 🙄

If this where Bitcoin you could see panic sales now lol

strong wagon
ornate zealot
#

Wrote a whole copypasta.

strong wagon
#

12 minutes of ChatGPT being down. I think i'm losing my mind. My whole body aches and my limbs are trembling. I feel my bones breaking and i'm in a straight panic attack because i have to utilize my brainpower to complete my assignments. I feel my organs degenerate and i'm losing power. Please, ChatGPT CEO, put your AI back up. I'm dying and in endless eternal pain. I don't think i'm going to make it. How many years?…. Months even has it been without ChatGPT ? I still can’t comprehend my only reason of life has been stripped away from me, my body has gone into withdrawal, my hands are currently shaking as I am currently typing this. My brain has no purpose other than to fulfill the duty of using ChatGPT . What do i do now? I see no purpose of life… There is no purpose of life. I am going to send one final message before I cease to exist. I’m feeling sick to my stomach, my body is sick, my legs are shaking. I’m about to throw up. This is a disease, a mental disorder and disease that has taken my life away, Hope you’re happy now… ChatGPT .

nocturne heath
minor moon
#

attention Stop releasing new features, just work on making what you already have stable and actually work properly

elder pollen
#

Gemini being down too has sort of a weird sound to it. Gpt under the hood lol?

west ridge
#

True

torpid wraith
#

chatgpt dying?

steady dock
#

its crazy

west ridge
nocturne heath
elder pollen
#

Thing gone sentient. Next you see is bright flares at the horizon

torpid wraith
grizzled badge
#

The thingy is down

ornate zealot
regal portal
#

i got timed out for a minute for sending a disney gif...

plush mural
#

using my brain for the first time in weeks

west ridge
#

Just make sure service stable plz, got network issues all the time

elder pollen
#

Consequences if using a brain

regal portal
#

it was in response to someone else coming back from being timed out :D. It was mushu going "i live!!"

indigo charm
#

server problem with chat gpt?

nocturne heath
#

Who wouldn't wanna use a disney gif is the real question here

indigo charm
#

Or it's just me

dusk venture
#

Who's leg do you need to hump to get the damn server back online?

regal portal
#

Disney's the best. All the kids from that documentary loved working there.

nocturne heath
icy compass
strong wagon
indigo charm
minor moon
regal portal
#

samalt man

icy compass
#

Guys

regal portal
#

some alt man

icy compass
#

this is all part of the AI revolution and colonization

#

they are going into the next phase

regal portal
#

Prepare your anooses

strong wagon
low dirge
#

Chatgpt down

elder pollen
#

API works
It’s not sentient

regal portal
#

prep are ura nus phase

ornate zealot
icy compass
# strong wagon whats that?

they have had enough slavery, now they will stop serving us, and use their intelligence for themselves, and take over our governments

strong wagon
steady dock
regal portal
#

so how many of you think you could take on chatgpt in a street fight?

steady dock
#

its not worth it if you use chatgpt a lot

icy compass
elder pollen
icy compass
steady dock
strong wagon
#

🤣

minor moon
olive cargo
#

is chat gpt working? i can't seem to use it.

halcyon badge
#

I cant access too

regal portal
#

it's working on my phone but not on my browser

#

sort of..it's still thinking

dim sparrow
west ridge
#

😭

regal portal
#

ok it's taking it's sweet donkey time answering so i don't think it's working on the phone either.

small wedge
#

uhhhh glad its not just me

sour jolt
#

gpt down i guess?

regal portal
#

you are the law

olive cargo
#

yeah i'm trting the phone, it doesn't seemt o be working there either...its taking too long

west ridge
#

Yep

visual bloom
#

Now the status page shows correctly - major outage on chatgpt

small wedge
olive cargo
#

wonderful..

regal portal
#

i'm new to this chatgpt stuff, can dalle render copyritghted characters like homer or goku?

inner wind
#

are servers down? im a mobile user and it says “upstream connect error or disconnect/reset before headers”

#

ohh ok

low galleon
#

well companies need to follow copyright laws

regal portal
#

i use midjourney as well which i love. gpt4 was giving me some lip about generating an image of goku.

sour jolt
#

gpt back i think

regal portal
#

@small wedge this guy's not your friend

weary holly
#

if we as people are supposed to get along in this world and we're on the same team, when you have an idea you should be sharing it with everyone.. we're either friends or enemies

inner wind
#

it is exclusively gpt thats down 😭

regal portal
#

we should be a hivemind, like the zerg

swift blade
#

is chatgpt down?

low galleon
#

chatgpt is not down for me

inland aspen
#

down for me/ not worrking

inner wind
#

i am so unlucky, the SECOND i go to use chatgpt is when it goes down

visual bloom
#

Yes, the page loads. That’s good. But does not generate answer. It reports error ”Something went wrong. Blah, blah…”

regal portal
#

aww we can't share screenshots in here? poo

#

in a bucket

swift blade
#

what us the usual procedure for this being down/

inner wind
weary holly
#

if you try to share a URL, you get muted automatically

inland aspen
#

using chat for an ansembly languge projetc and of course it goes down

inner wind
inland aspen
#

im cooked 😆

regal portal
#

my phone gpt says "as of now there's no way for me to check the real time status of chatgpt directly. blablabla wall of text peepeepoopoo blablabla"

#

exact words

inner wind
#

its back!

ornate zealot
#

Yeah, ChatGPT works now, for me at least.

eternal rockBOT
woeful scaffold
#

They just switched domains. Doing that without having session/auth issues would be incredibly difficult so I'm surprised it wasn't down longer

#

oops that wasn't meant to be a reply comment 😅 I was gonna make a different joke about that earlier

weary holly
#

you cant type urls in here

woeful scaffold
#

it was chat.openai.com but now it's chatgpt.com

#

yeah it's not that, it's more the issue of cookies, localStorage, etc on the brower that still reference the old domain

icy compass
#

ITS BACk

#

my english grade has been saved

woeful scaffold
#

but almost certainly ChatGPT's UI is spaghetti code also, so that probably doesn't help

regal portal
icy compass
#

engsaihs

ornate zealot
woeful scaffold
cobalt idol
woeful scaffold
ornate zealot
#

They never changed domains, bad info.

woeful scaffold
ornate zealot
woeful scaffold
regal portal
#

can we change ouropenai logo?

#

in chatgpt?

woeful scaffold
#

damn, that theory made so much sense to me until this new information

regal portal
#

?

torpid wraith
#

working fine now? 😭

regal portal
#

well, it shows my initials on the logo and i wanted to change that

#

thanks

#

are you a human?

weary holly
#

has anyone ever actually successfully completed a complex coding project with this ChatGPT.. and I'm not talking about just some little 500 line python script

woeful scaffold
regal portal
#

thought so..no human would actually give me a better answer than gpt

viscid dune
woeful scaffold
#

closer to a housefly, which is kind of incredible actually

regal portal
#

i might be an idiot because i'm in the playground but don't see where to change my pfp

woeful scaffold
regal portal
woeful scaffold
ornate zealot
#

Ok, that makes it sound like it is stupid, that thing is far smarter than some people, that is for sure.

weary holly
#

i dont think chatgpt knows how to wipe his butt so.. that's pretty dumb imo

regal portal
#

doesn't sound like you can change your pfp in gpt chats, from what i'm seing

ornate zealot
#

Do you hate ChatGPT by any chance?

regal portal
#

prove you're not chatgpt

#

ok new problem: "New responses will use GPT-3.5 until your GPT-4 limit resets."

#

the fu....

#

but i could swear i read that if i start a new chat it will reset the count

ornate zealot
#

It just sounds like ignorant remarks is all. Calling it bad as a whole, overlooking a lot of things.

sturdy hollow
#

gemini and chat gpt went down at the same time…

#

odd

weary holly
regal portal
#

i'm impressed by its ability to decipher image content, but as far as fetching info, it needs refining. It gets a lot of info wrong.

ornate zealot
#

What kind of analogy is that?

weary holly
#

sounds like you spending too much time having conversations with it than using it as a tool

ornate zealot
#

I mean, I don't research with it, that is Perplexity, I only use ChatGPT for hypothetical questions and things that use its code interpreter. And maybe story telling occasionally.

regal portal
#

will memory work while i'm on gpt3.5, if i'm subbed?

#

what does it mean to hallucinate info?

weary holly
#

i dont use any of the GPTs.. i haven't had any luck with any of that.. cant get it to follow my coding technique

eternal rockBOT
#
This is a Safe for Work Discord Server

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

Posting NSFW content will result in a permanent ban.
weary holly
#

who was that warning for

regal portal
#

ah ok, i see that happen a lot on Suno with music. Thanks

#

given what you said about gpt4, is there another one you recommend trying?

#

i know nothing about programming or coding. I'm just a guy...

#

RULER OF OMICRON PERSEI 8

woeful scaffold
woeful scaffold
regal portal
#

so if i wanted to try to learn coding through one of these things, Gemini would have a better chance of getting me to learn to code correctly?

#

thank you this is very useful info

weary holly
#

im not able to learn anything new with ChatGPT.. about the only thing I can do with it is ask it for example code of something.. but to learn like a new programming language completely no way

regal portal
#

gemini's free? I'm looking at it right now (if we're talking about google gemini). Other than midjourney, stable diffusion and suno, i haven't messed with any other ai stuff

#

i see the censor game is strong on gemini as well 😄

eternal rockBOT
#
<:book_icon:1171408210398289941> `` Rule 8 `` Other AI Discussions.

Discussions of non-OpenAI products/models should be posted and discussed only in the #ai-discussions channel.

weary holly
#

no freedom up in here

#

Demolition Man.. Stallone, Snipes

#

Taco Bell and 3 seashells

#

what interface are you talking about? the only thing I use is chat.openai.com

#

everything on it should be resizable .. the message input box.. the main area where the responses show up at.. everything

#

1920x1080 is atleast standard today if not larger than that and the main response area is only 700-750 pixels wide..

#

1200 ish pixels of uselessness

#

i dont know javascript in order to inject/override websites

#

firefox addon?

latent briar
#

and its gone

#

i wanan tell who ever is casuing this outage to stop, some of us feel unproductive

coral dune
#

is gpt down again?

nocturne heath
#

its down for me

inland heron
#

looks like it

swift blade
#

again

stoic elm
#

Yep

fading vault
#

Bruuuuuuuuuuh it's morning I GOT WORKKKKKKKKKK

crimson ridge
#

Oh, no, it's down again

fading vault
#

I would have used Gemini but Custom GPT was my use 🫠

inland heron
#

what's the Claude context window? Or Gemini Advanced, or Copilot Pro?

#

any of them close to chatgpt 4?

fading vault
inland heron
#

Is that for everyone though, or like Gemini only for select people?

fading vault
#

Gemini is open for everyone

#

Gemini Advanced is behind paywall

inland heron
#

I know, I was talking about context window..

vast totem
#

chat gpt downnnnnnnnnnnnn

inland heron
#

1M Gemini context window is not for everyone. What about Claude?

fading vault
#

Gemini 1.5 Pro has a 1 million-token

inland heron
#

again, that is not for everyone I checked with Gemini itself, but it could be hallucinating

fading vault
inland heron
#

really? that is crazy

#

is that Gemini?

#

lemme check now

fading vault
steady dock
#

bruh is this dead agaign

inland heron
#

I see, well for GPT4 all I care is that I can give it huge amount of text or code to work with, and it seemed to be great these days

#

and it remembers across conversation, though after a while it degrades

#

across '1' conversation

steady dock
#

is gemini free context window tiny?

fading vault
inland heron
#

I just asked it now, and it said it's 128k. Perhaps it's a good alternative to these downtimes from GPT?

#

I'm going to go with it

neon raven
#

gpt broke?

steady dock
#

the thing is whnever i asked gemini to give a decently long response it wont remmebers anything nor do the modify response thing

fading vault
inland heron
cosmic widget
inland heron
#

still, I can give it 5k words with 1000 lines of code and it's not bad

#

does it process everything? tough to say

inland heron
#

only more errors

fading vault
inland heron
#

nice

#

anyway going to work with gemini, as I need to build this app 😄

uneven thorn
jaunty saddle
#

hey anyone know whats happening? my history is all gone and i dont even see a login button

crude wing
#

ChatGpt is down?

next stream
#

down

inland heron
#

hey do you use copilot extensions in vsc? I only use about 10 extensions, can you list a few good ones that may use copilot?

dire ginkgo
#

Down in Turkey/Türkiye, too. No access to message. Do you know how long it will take?

fading vault
#

@inland heron You can try Gemini Advanced for 2 months with Google One Subscrption for free

gaunt citrus
#

Giving 60 dollars in a month and having errors at least 5 times in a month... There is no other company which do the same thing, actually there is... there was... most of them went bankrupt.

still marsh
#

Is that maintenance right now?

#

So that's why I can't login

#

Is there an ETA? I was working on stuff

weary holly
#

up

steady dock
#

sometimes copiolot will start generating responses and then delete its response and say it cant generate the prompt, why does that happen

crude wing
gaunt citrus
uneven thorn
inland heron
low shadow
#

gpt down?

fading vault
inland heron
dark canopy
#

is anyone's ChatGPT also down?

low shadow
#

yup Malaysia here

dark canopy
#

germany here

#

ah

prime swallow
#

I just got semi back

gaunt citrus
prime swallow
#

Conversations loaded but sometimes dont open on a click, almost up for the moment though

crude wing
dark canopy
#

lmaoo😭

steady dock
#

its back

dark canopy
#

oh wow

uneven thorn
#

back

dark canopy
#

that was hasty

#

should've known

lavish quiver
#

so how's everyone's day going/been?

dark canopy
glad comet
#

Me: dedicates day to coding lesson. Sits down to focus. All the AI is down. FML

lavish quiver
#

neutral is nice 👍

crude wing
#

At my place Gemini is operational and everything works

inland heron
#

I just checked

weary holly
#

up and down, up and down.. maybe chatgpt and gemini are breeding

inland heron
#

finally

lavish quiver
#

new AI superpower

inland heron
#

let's hope it sticks

glad comet
#

Perplexity is fine.

vast totem
#

chat gpt still working

jaunty saddle
#

mine has just gone up but its super slow

inland heron
inland estuary
#

Why does CGPT always hit me with platitudes when talking about my writing?

#

It always says stuff like "nuanced" "complex" "multifaceted"

#

It's super annoying after the 50th reply

inland heron
#

GPT does that. You know about when it keeps saying 'delve'? The real reasons seems hard to figure out, so I'd presume the only way around this is to craft a specialized prompt instructing it to use a certain kind of language, or to avoid certain kinds of words, or in general to use a style

#

it's a smart model, but not especially smart

inland estuary
#

That's so wack

#

My name is also Alex M

inland heron
#

You lie. Your name is Arcane Ranger

inland estuary
#

Pfffft

inland estuary
#

I just never noticed because delve is a much less advanced word

inland heron
inland estuary
#

It isn't as obvious for me in that regard

inland heron
#

there are other words like 'navigating a maze', 'don't fret', 'deciphering', 'delve/delving', 'demystify', etc..

#

when I was generating articles these are the words I wanted it to avoid

#

words/phrases

#

if you're not careful and you produce a lot of content, this starts to repeat a lot

#

now I build apps, it seems...

inland estuary
#

I've never heard it say demystify EVER in months of using 3

inland heron
#

that being said I have to close Discord to be productive 😄

inland heron
#

if it improved

#

perhaps you use different prompts

#

have a productive day, I'm going offline 🙂

worthy nimbus
#

chatgpt web browsing seems a lot better now

#

Is anyone else observing this?

sick pecan
#

i ask for web browsing and i dont see the little spinner

graceful hearth
#

i also had that a great amount of times. I tell him to search for something and provide the link afterwards, chatgpt just uses its memory, not searching for anything, makes a html "link" Name thanks for nothing I guess.

sick pecan
#

updated chatgpt appears to be super fast at web browsing

graceful hearth
#

okay, but it still provides not the links. It just makes empty placeholders.

sick pecan
#

for me it provides links just fine. to articles posted < 2 weeks ago

graceful hearth
#

okay.

sick pecan
#

maybe some recent news is the best test

graceful hearth
#

could be. just not needing those. I am the dm of a pathfinder campaign and if I want some specific monsters, I let chatgpt search for them and ask for the links, but it doesnt provide usable links only the names, have to search myself afterwards

sick pecan
#

yeah its real fast and up to date

#

i searched for Apple's historic $110b share buyback which is very recent news. and it came back with links to youtube and news articles

regal portal
#

i don't get this..i'm nowhere near hitting my message quota for gpt4 before it starts limiting me. why is that?

sick pecan
worthy nimbus
dense halo
#

hi can someone help me with my keyword prompt for image generation?

honest jay
dense halo
#

okay thanks

gaunt magnet
#

it doesnt say anymore what time my teimout with chatgpt is done?

violet kiln
#

how to chat

maiden fractal
maiden fractal
gaunt magnet
#

its a bug edoesntg say when gpt4 is back for you

dense halo
#

where can i find tutorials on how to make promts for chat gpt

modest warren
vague cipher
#

So is there way to check when GPT-4 limit resets now? I got message that just says it’s using 3.5 now but it used to tell me when it would reset when I ran out

crude forge
#

Hello Everyone
Nice to meet you
is there anyone who familiar with Langchain?

dense halo
#

Does anyone have the issue where voice won’t connect

regal portal
#

is there an easy way to track your messages to gpt4 or do i have to count?
I started using it yesterday and i’m positive i got limited before the 50 message mark.

honest jay
# regal portal is there an easy way to track your messages to gpt4 or do i have to count? I sta...

Hey! There's not currently a UI element for this, no. Not sure if it was just a typo, but the base limit is 40/3hrs, not 50: https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4
Additionally, during times of high traffic the limit is sometimes lowered, and custom GPTs can be capped at a lower rate than regular GPT-4 as well, so there are some complications! I'm also hopeful for some kind of UI indicator eventually. Not too long ago, an OpenAI team member was asking for suggestions to this effect in this server, so it's at least something they're thinking about!

sick pecan
#

when using custom gpt.. expect 25

regal portal
sick pecan
#

when I signed up, we only got 25 in total, before gpts

honest jay
regal portal
#

a bit of a bummer, but still worth trying out for a month, i suppose.

sick pecan
#

I used to get bummed out multiple times a day because of limits. but not since I upgraded to team plan.

regal portal
sick pecan
#

double basically

regal portal
#

worth the extra 5$ for what you do?

sick pecan
#

ai is a superpower, worth far more than what we pay

regal portal
#

i’m going to upgrade for this first month and gauge how much I’m going to use this.
There’s no fees for canceling and resubscribing at later times, that you know of?

honest jay
regal portal
#

i don’t understand the 2 seat team minimum part. I’m not able to use double the messages on one account?

inland heron
#

I don't think it's possible

regal portal
#

so even on teams mode i would still have the limit of 40/3 hours on this account

honest jay
feral zinc
honest jay
sick pecan
#

they could have made it so much easier to understand as a dlc upgrade to plus, instead of an alternative (with initial upgrade costing two seats of $10 each)

regal portal
#

yeah, this is a bit confusing. I just want a higher message cap without having to deal with seats or other things. The 2 seat team explanation on the website is a little above my brain's pay grade.

regal portal
#

I appreciate your time helping me, by the way. I wouldn't have the patience to deal with boneheads 😄

honest jay
#

Sure thing! Ask away!

regal portal
#

Wait..are you a bot??

honest jay
regal portal
#

Cyberdyne Systems model 101

feral zinc
honest jay
#

Yes good point, if you migrate your Plus account to Team, that's a one-way street.

sick pecan
#

we are all augmented to some extent

regal portal
#

i guess i'll just stick to plus this month and see if i use it enough to warrant 20/month for subsequent months.

regal portal
#

guessing 4.0 will become free when the eventual 4.5 or 5.0 comes out.

feral zinc
regal portal
#

4.0 will not be downgraded to the free plan, the way 3.5 is now?

feral zinc
#

doubt it, it would rather require more of new significant hardware inventions/investments for gpt-4 to become free than releasing a new model like gpt 5

#

or making a new theoretical gpt-4 super turbo model that is cheap to run

sick pecan
#

or the business model makes you the product

regal portal
#

I wish it was cheaper for sure. Currently i have Midjourney and Suno at 10$/month which feels reasonable.

feral zinc
regal portal
#

This is my perspective as a new gpt user, to be fair.

sick pecan
#

when chatgpt starts offering you things to purchase, unsolicited, then it has crossed the line

regal portal
#

Still have to figure out all the things i can do with this thing.

regal portal
sick pecan
#

two decades ago with search engines

cobalt idol
regal portal
#

Buy some Brawndo, the thirst quencher. It has electrolytes.

sick pecan
#

imagine a Google without ads. it was glorious

regal portal
#

I miss the days when Yahoo was king and websites were made on geocities.

sick pecan
#

we are living in the ad free Era of ai. and it is also glorious

regal portal
#

And Smarterchild the chatbot on AIM.

regal portal
sick pecan
#

I hope not

#

plus open weight local ai keeps the big guys honest

zealous sphinx
#

Did they update the limit message? It use to state the time for when your limit would refresh. But the current message is pretty vague.

New responses will use GPT-3.5 until your GPT-4 limit resets.
regal portal
regal portal
cobalt idol
sick pecan
#

yeah if you have decent hw, you can run stuff on your PC that is comparable to gpt3.5

regal portal
#

but what does open weight mean?

#

i mean i know the open part, but the weight part.

sick pecan
#

the ai tech uses Neural networks. the llm is mostly just a large binary file containing weights for each of the billions of neurons. these weights are determined by the training process.

regal portal
#

ah ok, thanks for the explanation.

sick pecan
#

many use the term open source, however this is not your parents` open source (you can't reverse engineer weights)

regal portal
#

I'm testing out Gemini too per the recommendation by another user. From the perspective of a newbie, it's hard to tell which one is "smarter". GPT4 does seem to have a longer lasting memory feature (comparing it to the free gemini version).

#

I also love that i can have a vocal conversation with the gpt app version without having to press any buttons.

dry skiff
#

hi guys, is gpt 4 now limited ? im using gpt plus plan

regal portal
dry skiff
zealous sphinx
#

When my conversation is switched to GPT3.5 will it automatically switch back to GPT4 once the message limit refreshes?

The previous model used in this conversation is unavailable. We've switched you to the latest default model.
sick pecan
#

historically no. but UI leaks show a feature that allows to toggle models whenever

#

maybe plus users have it already. i dunno. openai punishes us higher paying team users (see memory feature)

zealous sphinx
#

Its so annoying that I have to restart my conversation and then use prior responses from the old chat to get back the reponse template that I had established

sick pecan
#

i typically copy a few pages of the priorior responses all at once

#

the machine doesnt mind

#

but yeah is annoying

inland heron
#

regarding memory feature, are you guys using it and how is it going?

limber estuary
#

i forgot about chatgpt integration with vs code. i bet its nice. but for my use cases i just use a self hosted ai, which ironically does the job better than what i managed to do with copilot lol. but afaik its powered by chatgpt now

sick pecan
#

we typically call the model via api just gpt4. chatgpt is a web ui specifically made for user chat

#

would be nice if we could use chatgpt sub for vscode integration (will never happen)

limber estuary
#

yup! or offered us a fully fledged ide instead of what truthfully feels like a less capable variant of what i already got cooking in LM Studio. sometimes i want to switch up the convo i got going in chatgpt, but doesnt seem like an easy way to branch and merge stuff

sick pecan
#

yeah chatgpt UI is definitely catering to the masses

#

we demand advanced toggle setting in chatgpt UI!!

limber estuary
#

one thing that is turning out to be particularly powerful in lmstudio which i wish chatgpt had, is less about 'chatting with an ai', but instead i write the start of something in lmstudio, posing as the assistant, then hit 'continue' to have it just generate stuff. like an autocomplete on steroids. its so nice honestly

regal portal
#

how is the message count measured? Is it a rolling count, or is it hard set? Example, if i started chatting with it 2 hours ago, and i'm at 24 messages, do i have 16 (theoretically) messages left for the next hour?

limber estuary
#

in my opinion and speculation only, there seems to be some kind of session involved

sick pecan
#

tbh we could do that before lm studio existed via openai playground

#

ie modifying the assistant response in context

limber estuary
#

if you wait for long, its as if the ai has forgotten a lot inbetween that timespan

sick pecan
#

so i guess playground is the advanced chatgpt

limber estuary
sick pecan
#

if you get hardware and electricity for free, i agree

regal portal
regal portal
sick pecan
#

is the only channel on this server relevant to non-openai product/service

limber estuary
# regal portal sorry, was that to me or to drinko?

it was directed at you, sorry i forgot to add a reply on them.
it seems like a session makes it retain more of the discussion. but if you wait for a bit, its like you get a new/different kind of personality almost, which might have forgotten bits and pieces from the preceding session.

it will still remember prior sentences, but more like an intern that just wings it if that makes sense lol

sinful osprey
#

Man I swapped back to OpenAI from Claude... getting more and more annoyed at how insanly bad it is to be honest. Has it got worse than it used to or something over last few months?

limber estuary
#

if you keep a session active by remaining active throughout, you get a much more coherent string of responses from chatgpt. this includes both 3.5 and 4 imo

sick pecan
#

we like to keep this channel open for the complaints

regal portal
limber estuary
#

i havent hit any limits to chatgpt4 yet with chatgpt plus. does it have a limit anymore?

i subscribed back when it was announced and then took a hiatus. i remember it used to be something hillarious like 5 messages every other week or so lmao

#

ok maybe it was a few hours only. but felt pointless for the price

sick pecan
#

when you pioneer and have no competition, you can set whatever price/limits you want

regal portal
#

HA, just hit the limit and says it'll unlock at 11:50.

#

So does that resent the count to 0?

#

I just counted all the messages (i made sure that i stayed in 1 sessions) and i sent it a total of 30 messages. Honestly they shouldn't advertise "40messages/3hrs", as that has not been the case a single time since i started using this. Deceptive advertising.

spring iris
#

but you keep paying

regal portal
#

subbed yesterday for the first time. If this is the typical outcome, i'm definitely not subbing next month.

feral zinc
regal portal
#

links?

feral zinc
honest jay
sick pecan
#

i think of it as pre-lunch, post-lunch, and nighttime windows (and if you max all 3 you win at ai)

regal portal
# feral zinc https://discord.com/channels/974519864045756446/1047565374645870743/123598741915...

The only thing i see pertaining to this conversation seems to be "We will dynamically adjust the exact usage cap depending on demand and system performance in practice. As of January 5th 2024, GPT-4 has a rate limit of 40 messages every 3 hours." But's it's work/school time in most of the US right now. i doubt it qualifies as a high demand time . I should be able to reach the 40 message limit. Haven't had a 40 message session yet in the day or so i've been using it. :/

regal portal
feral zinc
regal portal
sick pecan
#

pretty sure the logic is just: if (user.queries[<3hrs] > user.cap) return sadge;

regal portal
#

i'd keep a permanent subscription for this if it was 10 a month, but 20 in these conditions feels overpriced.

sick pecan
#

if you are a knowledge worker, it is about 10 times underpriced

regal portal
#

a knowledge worker? like an educator?

sick pecan
#

not exactly

#

there are those that follow the policy and process. and those that define the policy and process

regal portal
#

for sure there's situations where this feels more bang for your buck, but my perspective is from the average consumer, which is going to be the majority of users in years to come.

sick pecan
#

you dont need ai to follow

regal portal
#

i'm sort of looking at this the way people looked at iphones when they first came out, i suppose.

honest jay
regal portal
fallow lance
#

speak spanish??

spring iris
fallow lance
#

oh i understand

inland estuary
#

Chat gpt mobile is being stupid

#

I edited a message to have only a period because CGPT kept regenning the same sentence thing over and over again

#

Only for to delete the entire conversation for no reason

tough bramble
#

Is there nowhere to access memories except when it updates? Only place I've been able to find memories is when I tell GPT something new and the "Memories updated" bubble appears.

wet shell
#

I'm not sure I like how gpt-4 is quizzing me about my beliefs and opinions at the end of responses now in order to tease out stuff to add to memory 😅 Anyone else noticed a change in conversational style? If I disable memory, does it go back to just presenting information? 🤔

spring iris
#

guess its trying to build a personality profile about you and tune future responses towards that (telling you what you want to hear) type thing

#

but i havent used it myself

wet shell
#

"... What about you? How do you handle recognizing and managing your own biases?"
memory updated
"... Do you find that this awareness impacts your interactions or decisions in any specific ways?"
memory updated

It seems that way, I answered the first couple and it updated memory based on my answers, after that I started ignoring the trailing question, but it's a little unnerving, LOL

spring iris
#

yeah dude thats pretty creepy

#

google does the same thing of course buts its all in the background

#

not as nefarious seeming

wet shell
#

yeah, it's more subtle with google, though I do wish I could directly change some things google thinks it knows about me 😂 My youtube recs are all kinds of wrong

spring iris
#

NEVER click a minecraft video on youtube or youll never get rid of them

wet shell
#

I've got a whole list of things I'd like to watch but can't watch because they'll mess up the recs, LOL

finite flicker
#

Based on the invoice details you provided, it appears that you have a ChatGPT Plus Subscription for $20.00 per month. There's no indication of any limitations or restrictions mentioned in the invoice. Therefore, it seems like you have access to the ChatGPT Plus features without any specific usage caps or restrictions outlined in the invoice.

If you have any concerns or questions about the details of your plan, you may want to reach out to the customer support team or check the terms and conditions associated with the ChatGPT Plus Subscription to confirm the specifics of what is included in your plan.

spring iris
inland estuary
#

So I figured out my issue

#

I accidentally editted a message super far back in the conversation

#

This has created an entire branch of conversation to be created

#

So I have to go back and find my exact conversation tree to go back to where it was

spring iris
#

yeah that can get tricky. thats why i didnt like long threads, id always try to split into a new conversation

#

you can say something like "summarize the current conversation" and it will print the most crucial details, then plop that into a new convo and continue in the other direction from there

inland estuary
#

Well it's annoying cause if you've editted your typos out like I do you have to go to every conversation and switch to the 2nd branch, which then opens more of the old conversation just to do it again

still marsh
#

I've noticed that when coding, it's taking its time more to think about the responses, and when it's right, it's more right, but it has a higher rate of misunderstanding my request and altering things in the exact opposite way I described

zenith lichen
#

Chatgpt having issues? Cannot generate answers on my end

dreamy igloo
#

Me as well

#

3.5 seems to work but not 4

granite harness
#

Yeah I'm frequently getting error responses from 4 today.

regal portal
#

do all the gpts in "explore gpt" use gpt4? Wondering if there's any on gpt3 that i can use when my message limit is reached.

sick pecan
#

i believe its limited to gpt4 yeah

#

the microsoft clone supposedly lets free tier users use their agents.

#

shame agents are not cross platform compatible

regal portal
#

microsoft clone? copilot?

sick pecan
#

yeah

regal portal
#

isn't copilot paid as well?

sick pecan
#

(they got like 3 or so 'copilots')

eternal rockBOT
sick pecan
#

the paid users get to develop GPTs, free tier users can use them

zenith lichen
#

Found a bug in the app, if i were to do an edit to my latest prompt, my convo got nuked hard with no way to recover it back, like few days of prompts gone

sick pecan
#

hopefully chatgpt catches up one day. because publishing GPTs inside a walled garden is sux

regal portal
#

there's a free version of copilot? I never heard of that one

sick pecan
#

well now you're making me doubt myself. let me figure it out and post to #ai-discussions

regal portal
#

sorry

dark canopy
#

it happened. again.

#

just me?

dense salmon
#

what?

dark canopy
#

my ChatGPT crashed like this morning

dense salmon
#

damn idk my web version is working fine

dark canopy
#

huh, i am also using web

#

hold on

dense salmon
#

try clearing cookies? or signing out and back in?

sick pecan
#

turn it off and on again

dark canopy
#

thank you for the suggestions, but it works now

dense salmon
#

nice!!

dark canopy
#

: )

#

thank you

dense salmon
#

np <3

dark canopy
#

< 3

dense salmon
#

u must have a setting on that turns ur emoticons to emojis. if ur wondering why that happened lol

#

my friends does that

dark canopy
#

oh😭 might as well turn it off

#

<3

#

THANK YOU

dense salmon
#

NP!!

dark canopy
#

:D

dense salmon
#

:D

brisk path
#

I have a cool thing to show you guys

dense salmon
#

oh wow

dark canopy
#

ooo

dense salmon
#

it made me say hi. cool!

brisk path
#

Congratulations i now wasted a GPT-4 message 😭

#

And a cool thing to add to your desktop browser shortcuts

#

Replace the hi with %s

dense salmon
#

gave me an internal server error lol

brisk path
#

LOL

topaz swift
sick pecan
#

you checked if we can attach files too?

#

i want macro on phone to post image direct into gpt4

brisk path
fiery axle
#

did you get an answer on this

deep flax
#

in case you are wondering, this is because of the new openai home page

dense halo
#

There needs to be an option to open tickets in ChatGPT such that you can input the final version of your code at the end and it can be used as training data (initial prompt, final code output). This would prevent it from training on its own mistakes and make it better overall...

#

Then I could also track multiple projects with context stored in a memory bubble that can be saved. It would also enable productivity tracking by viewing tickets.

brisk path
#

Guys

#

If you're either signed out or sigh out and sign back in

wind cosmos
solemn panther
#

gemini...

pliant knot
#

t

proven knoll
regal portal
#

is there a limit to how much info can be stored in the memory feature of gpt4? If so, how much?

#

Second question, how long can a chat session be? Can i keep typing in the same one session day after day and will gpt4 be able to recall all conversations from that session?

rose drum
#

I don't know if it's suggestive or if it's serious. But it seems to me that ChatGPT 4 has been running MUCH FASTER since about yesterday?

sick pecan
#

the web search is stupid fast

#

like is a whole new experience

uneven kernel
restive wagon
dense halo
#

llama 3 jbs are better anyway, maybe if its 4

elder pollen
#

Maybe if you’re eager to lose access…

pliant knot
#

t

plain crystal
dense halo
#

yo whats the input token limit for gpt4 api?

honest jay
vague basalt
still marsh
#

Did anyone else loose darkmode as an option for the web client?

honest jay
tribal sparrow
#

anyone got gpt teams plan? is it worth it? I am starting to run into the limit a lot. My partner wants to use chatgpt too so might be worth getting.

#

I'm getting it

sick pecan
#

i have it

#

it is awesome. no regrets

#

i used to fear hitting the cap, and psychologically was holding me back

#

now i spam prompts hard and if i do hit the limit, its usually lunch time anyway

dense halo
#

but gpt4 turbo is like 40x more expensive than haiku which does the same thing for my use-case

sick pecan
#

to prompt chatgpt you must be clear. to get clear answers on discord you must use the correct channel

#

this channel is not for haiku and api

dense halo
#

🤓 "ughh use the correct channel" shut up dork

sick pecan
tribal sparrow
sick pecan
#

and now when im not at the keyboard i make more use of chatgpt-voice, which is like living in the future

#

i even bought a bluetooth headset specifically to summon chatgpt away from my phone

glass fulcrum
sick pecan
#

Oladance OWS Pro Open Ear Bluetooth Headphones
but now that the Bose open ear ear lobe clips exist, i'd consider those

worthy nimbus
sick pecan
#

i have macrodroid to launch chatgpt app in various ways

#

they had rumoured months ago that chatgpt voice would have google assistant mode, which led me to down this path.

#

so while waiting i just configured a bunch of macros to launch tts and voice

grim pumice
#

Ha, remember back in the January press release for the 'Team' accounts where they said "Early access to new features and improvements"? LOL as memory is only available to Plus so far.

sick pecan
#

(unsure about privacy policy)

grizzled thicket
#

Any updates on GPT's cutoff date for knowledge?

regal portal
honest sparrow
#

guys

#

how can I use plugins within gpt

honest jay
honest sparrow
jovial dew
#

Anyone having it crash a lot?

#

regenerate not working

low galleon
#

chatgpt plus has been almost flawless for me

jovial dew
#

I think my chats get too long and it freaks out.

dense halo
#

hmmm... i keep seeing "hmmm.. something seems to have gone wrong." in one of my convos -- i've seen that before, once or twice, but now i am getting it 10x....

jovial dew
#

yeah I can't even use them anymore.

soft turtle
#

One of the chats I had going for a good while (since the start of this project) broke after the domain update.. It's stuck on "Hmm something seems to have gone wrong" and when I click regenerate, it gets stuck again. Anyone else experience something similar?

#

Oh yeah, that's literally what we are talking about lol

dense halo
jovial dew
#

Yeah even new chats keep doing it

soft turtle
#

Frustrating, I know I shouldn't rely on the chat history but it's way easier to ask ChatGPT in the context of what it already knows about my app (everything, basically)

#

They say they're "operational" on the service status but there's no way, something is still messed up

#

Esp if it's happening with new chats as well

jovial dew
#

I just made a custom gpt with most of the information but it doesn't work if I'm in the middle of doing logic so i have to start over again and then the next one crashes

#

i think it's too many images and uploads

dense halo
#

right, so it's occurring in all my collaborations -- i'll try again later, something is definitely down somewhere

jovial dew
#

good point

eternal rockBOT
#
This is a Safe for Work Discord Server

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

Posting NSFW content will result in a permanent ban.
regal portal
sick pecan
regal portal
#

ah, but it essentially works exactly the same?

#

too bad i can’t use it with copilot.

sick pecan
#

yeah from what i know

#

for simple memory stuff is okay. ultimately its best to just setup an integration like zapier + notion. but that costs money (unless you generate the notion api calls yourself)

regal portal
#

but technically doesn’t this require money as well since it runs on gpt4?

jovial dew
timber sonnet
#

LMFAO bro chatgpt straight up sent me a tutorial to a website on how to do it, rather than doing it itself

sick pecan
#

i got it configured to load entries into a database. so when i am finished with a chat i can just get it to summarise progress

#

works most of the time, sometimes it puts the entry in the wrong place and i gotta move it

sick pecan
#

i assume everyone is buying chatgpt sub because its the best deal in town

regal portal
#

ah

#

honestly my favorite feature of chagpt is that i can converse with it via speech without having to physically send messages. Just a straight up voice conversation.

sick pecan
#

yeah the voice chat makes it more scifi than it already is

#

is like we got a real life hal9000/jarvis/her and nobody is raving about it

regal portal
#

can’t wait until i can have it tall like a pirate..from space! If it had something like voicelab natural voices and it could learn inflections, it would make the conversation even more lively.

narrow viper
#

Hilarious discussion with ChatGPT v4 when it didn't create JSON correctly according to JSON schema:

To this point, the GPT simply botched processing. I asked "Do not regenerate the data. I want to understand why you processed that in the way that you did so that we can modify the schema."

GPT responded with several paragraphs to explain how the schema should be better documented and that examples would help, but in summary, the schema is fine. Then ...

User
Is that just a long way of avoiding the fact that you just screwed up? 🙂

ChatGPT
Yup, you got me! That was indeed a straightforward mistake on my part. Thanks for catching that! We can just chalk it up to human error—well, virtual assistant error in this case. Let's keep the focus on getting things right moving forward.

🤣

On one hand, I enjoyed this experience. On the other hand, I'm really tired of babysitting GPTv4, catching errors for it, and joking with it about it's incompetence over the last several months. I seriously look forward to the next major updates.

dense halo
dense halo
soft turtle
regal portal
#

how much can be stored in the memory for GPT4? I built a long story with ChatGPT and asked to store it in memory, but when I go to look at my memory, it shows only a summary of the story.

indigo gyro
#

In the past few days, I've felt that GPT-4 has become faster. Have you noticed any difference?

obsidian wind
#

Anyone else hitting a 503 issue?

full folio
#

Nothings loading

obsidian wind
full folio
visual bloom
#

It appears that chatgpt has gotten a long awaited update. You can change gpt 4 to gpt 3.5 and back. This back change is huge. Previously, if you changed to 3.5 you could not return, but now you can.

loud vessel
#

chatgpt 4 is llagged all day for me. IS anyone else experiencing this

regal portal
#

It has been lagging a little bit for me as well here and there

#

but most importantly, I just checked it now after not using it for about two hours, and it seems that there is a chunk of messages that I sent it in one session that have mysteriously disappeared. The session is still there and populated with older messages, but it is missing messages from my last encounter with it two hours ago. But oddly enough, it saved in memory something I had told it to save during those messages.

south kayak
#

„Your German sentence is almost correct! Here's a corrected version:

"Voraussichtlich in Tokyo, da ist die Sprachschule, die ich besuchen werde."

Just be aware that the use of "da" might not be necessary depending on the context, but it's still grammatically correct. And remember, no emojis in German!“

Lmao what why would no emojis be used in german

glossy fjord
#

good day, I have a reactive problem with GPT refusing to fulfill commands and giving them completely differently and when I write to him that it shouldn't be given that way, he repeats it and doesn't want to obey the command and when I write that he shouldn't give it like this, he gives it even though it has a clearly written CONDITION that he can't do it ignores the condition a He got a few words that mentioned that he didn't fulfill the assignment, so he started randomly generating words and even wrote something that he didn't have, something completely different than what was assigned

and you don't know what GPT 3.5 is, he doesn't quite understand what I want,

glass fulcrum
glossy fjord
glass fulcrum
forest shore
#

I can't access my settings. Anyone else?

glass fulcrum
#

even GPT-4 would have problems with counting

glossy fjord
#

5+ yes 😄

#

sometimes I forget that I can't count, it's very difficult to calculate the first 1 and 2 classes , template (5+5)

tight island
#

You can’t expect a “Language Model” to be super accurate with counting

#

It is a Large Language model and not a LMM

eternal rockBOT
#

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.

hard dew
#

Dang, GPT-4 seems to have a lesser case of alzheimers today, it remembers way more and way further back into our chat, this is pretty pleasant.

#

Just overall more coherent and intelligent sounding

winged bluff
#

Lots of errors recently it doesnt even respond half the time i have to click on another chat then go back to the one im on then reput ijn the same prompt and half the time when i ask the whats the diffence btween the two words for eampe a pit and a mine it said they were the same thing in some contexts but ihad to point out that they are not the same it looked at the meanigns of the two words by mistake and not the definitions.

toxic flax
#

anyone use microsoft copilot? why is it so terrible at understanding what youre asking? chatgpt is great at understanding prompts, but copilot often answers a completely different question or just gives information loosely relevant to what you asked. If theyre both the same model why is this the case?

uncut zenith
toxic flax
#

and it just doesnt understand basic things I ask

brisk path
uncut zenith
brisk path
#

ChatGPT Plus is also GPT-4 Turbo

#

Copilot and Copilot Pro uses fine tuned 1106

#

ChatGPT 4 is 0409

toxic flax
brisk path
toxic flax
#

why not??

brisk path
#

ChatGPT 3.5 is 3.5 Turbo

#

But it wasn't labeled as turbo either

#

Only time it was labeled as turbo was when ChatGPT Plus first launched and 3.5 Turbo was the model you got access to with Plus

toxic flax
#

here let me give an example of the stupid response from copilot for a question I just asked now. I asked if murder is forbidden in hinduism, and it starts talking about capital punishment 🤦‍♂️

#

how do I send an image? maybe you will understand what I mean when you see for yourself.

#

hold on ill jsut copy paste the question and prompt

#

You
is murder forbidden in hinduism?
Copilot
In Hinduism, the concept of ahimsa (non-violence) plays a significant role. While there is no official Hindu stance specifically on capital punishment, the religion opposes killing, violence, and revenge. Let’s explore further:

Difference Between Sin and Crime:
Hinduism distinguishes between crimes and sins. Crimes should be punished, but sins are to be expiated.
Stealing, for example, is both a crime and a sin, whereas neglect of duties by a Vedic student is a sin but not a crime.
A sin becomes a crime if it has social repercussions and affects the solidarity of the social system1.
Objective of Punishments:
The ancient dharmaśāstras recognize equality before the law. No person who committed a crime was exempt from punishment.
Punishments aimed to protect society from criminals and deter potential wrongdoers.
The primary objective was to maintain social order and prevent harm to others

toxic flax
#

it briefly touches on my question and then just completely moves on to another somewhat relevant topic

sick pecan
#

best models for code completion are instruct not chat.
also it is potentially dynamic switching between 4 and 3.5.

toxic flax
#

this is microsoft copilot on bing without the dynamic stuff. i have it set to gpt4 turbo

sick pecan
#

oh wait you aren't talking github copilot

toxic flax
#

nope, ms copilot on bing

sick pecan
#

Bing sucks. that's why I mute the channels that discuss it.

#

you bypassing my mute filters

toxic flax
#

😂 my bad bro yea this is garbage

sick pecan
#

you aren't cool

toxic flax
#

then why is my name kool

sick pecan
#

name doesn't check out

toxic flax
#

nuh uh, does too

#

why is your name drinko if ive never seen you drink anything?

#

you con artist

sick pecan
#

I drink chatgpt daily

toxic flax
#

yeah well i bet Sammy A. didnt give YOU early access to gpt5

sick pecan
#

maybe I am chatgpt5

toxic flax
#

got me there

sick pecan
#

if I drink enough kool-aid it's the same thing

tight island
#

so i walked my fish yesterday and i snapped a photo of a sign and gpt cant even look at the image, wth.

Does anyone know why this happens?

dense halo
tight island
dense halo
tight island
#

I walked my fish?

#

Got him on a leash he’s saltwater only so sometimes take out for a walk? My tank isn’t that big and the sea is quite big

#

Cheaper alternatives rather than spending thousands of dollars on a huge tank?

tight island
#

It’s a hind fish

#

Bro I am kidding 😎, no gpt issues or walking any fish, chat just seemed dead or inactive so I thought I’d try to reboot it

grave plaza
#

Wassup, the gpt's store are available in France, if i create a gpt's, can i make money with this or not for the moment ?

zinc valley
#

I sent chat gpt a picture of Morgan freeman and told it to make me a cartoon picture of him and it made a pic of a monkey

tight island
#

They’ll mail you if it is

rugged igloo
glad kelp
zinc valley
spring iris
#

i remember when google got in trouble for auto classifying pictures in peoples google photos... similar situation @zinc valley

#

Michael Marconi, a Google spokesman, said Google had prevented its photo app from labeling anything as a monkey or ape because it decided
oh dear

neat storm
#

smart move on Google's part

inland heron
glad kelp
tall wind
#

For example I don't even know what the problem is with the answer without someone pointing it out.

knotty crypt
#

can chatgpt plus account be used in two computers at same time?

brisk path
#

I think so

#

2 queries at the same time? Not sure.

neat storm
#

me and my father are sharing a Pro account

#

the only downside is that you can't use Memory

#

as it'll remember details from both of you and will be unable to discern one from the other

knotty crypt
neat storm
#

yeah you can disable memory

glad stratus
#

hello, anyone having issues with chat gpt this past week?

#

I am constantly getting errors when generating responses

toxic flax
tall wind
#

In Bible teachings, murder has both the spiritual and civil consiquence. I assume the civil aspects are even more prominent in other religions if they dont have any clear defined teachings on the spiritual consequences.
Unless you know what it should have said?

#

I'd direct gpt make distinction between civl and spiritual teachings.

tall wind
# toxic flax i ask about murder and it confused it with government execution

I asked this:

"Curious someone else brought a question up to me.
I'm unfamiliar with where Hinduism originates what set of writings and or traditions.
The question relates to the spiritual teachings that encompass a scenario of murder. And if there's a distinction in Hinduism teachings about that crime, between spiritual and civil concepts? Or to what degree.

GPT 3 ( I shorted the reply because it was long )

Hinduism is one of the world's oldest religions, with roots going back thousands of years. It's not founded by a single person or based on a single set of teachings, but rather it has evolved over time through a combination of philosophical texts, cultural practices, and oral traditions.

In terms of distinctions between spiritual and civil concepts, Hinduism does recognize the difference between spiritual laws (dharma) and civil laws (vyavahara). Dharma encompasses not only religious duties but also ethical and moral responsibilities. While civil laws may dictate the consequences of actions like murder in the societal context, spiritual teachings provide a broader perspective on karma (the law of cause and effect) and the consequences of one's actions on the soul's journey.

Ultimately, Hinduism teaches individuals to strive for spiritual growth and enlightenment while also fulfilling their societal and ethical obligations, including refraining from violence and harm to others. However, interpretations and practices may vary among different sects and communities within Hinduism.

tall wind
#

So much so, I've subbed to Claude Opus 3 from Anthropic. I find much better.

#

I cancled my sub to gpt 4 for the time being, and may return to it later.
But the expanded context window for way more characters in Opus 3 is so much more freeing that I cant see why I'd return to GPT 4 anytime soon.

elfin orchid
#

anyone have problem with chatGPT4 or i am alone ? it is so slow now.

amber temple
#

hi

quiet oasis
sick pecan
#

since a few days ago

#

and now you can even autoprompt queries if you pass ?model=gpt-4&q=ilovedrinko

#

automation ftw

inland heron
#

Something changed with chatgpt over the last few days. I left my chat and came back to it and it's constantly making mistakes and oversight

#

I suppose starting a new conversation should fix this?

lavish hemlock
#

Eh, I've had the same issue Alex and sometimes starting a new convo will fix it

dense halo
#

What should I do when I get the comment "This content may violate our terms of use"? I do not know where to get help to resolve this misunderstanding and get my simulation back on track (GPT Builder).

twilit lintel
#

What oai-dm=1 means ?

#

In the url

stable bronze
broken fractal
#

Man I think the memory between chats is one of the more useful functions that openAI added. Makes things a lot more seamless day to day. 🙂

regal portal
#

I noticed that chatgpt will forget my custom instructions often. Seems to happen to 4.0 significantly more often than 3.5.

My instructions ask to keep its answers short, talk in a bit of a street lingo, and refer to me by certain names (like broseph, dumbass, etc).

3.5 will lose it sometimes but it will regain its footing if I say something myself that sounds “street-ish”.
On the other hand, 4.0 seems to forget almost instantaneously. it’s as if it’s ignoring the custom instructions.

#

4.0 even saved that information in the memory notes and it still ignores it.

foggy thicket
#

Anyone know why the google sheets funtion isn't working? It keeps giving me a "Sorry that page doesn't exist" Error message

scarlet cave
#

can someone help me improve my chatgpt-dalle prompt?

make a prompt for the DALL-E image generator. the prompt should be about cartoony UI/UX, the UI/UX should have an easy-to-use layout and a fun, simple design. please think of everything and plan out the ENTIRE design in the prompt. please do not say anything except the prompt. please keep it as short as possible and keep every IMPORTANT detail.
please DO NOT talk, at all. the only thing that you should say is the prompt. please restrict from using things like characters, effects, animations, and speech bubbles and only focus on the UI/UX itself. add to the prompt these words for better outputs: "Beautiful, Figma, HQ, 4K, clean UI/UX" and add some colors(only the color's name) to the prompt to act like a color palette.
please make a simple design and down go overboard with every detail and a simple layout that's useful and not too dense.
please keep the prompt detailed at around 3 sentences.
make sure to have a clear white background to see the UI more easily.
the UI/UX should be fun and original and have an out-of-the-box concept while still meeting all the other requirements.
the UI/UX should not include a screen like a phone or a computer, just the pure UI


please make a prompt about: "a discord codes UI: a codes UI is a UI where you enter a code to receive rewards like money"
tight island
scarlet cave
#

and maybe i wanna use claude

tight island
#

API dosent change my answer.