gpt-models

[minor spruce]

why not just send the chat link instead?

[lost flume]

I paid for it ...but now seeing these basic mistakes. I am disappointed

[minor spruce]

tab1.id tab2.id
6 6
6 6
Is this the right answer? (for inner join)

I have seen this problem for the first time: Unable to load conversation cc106b3d-1c5e-49b7-8753-9e880d49ddd2

you are using 3.5 or 4?

well I am not able to see your conversation but still what's the problem you are facing?

[lost flume]

No

[minor spruce]

huh, maybe just hallucinating. Have you tried opening a new instance and asking it to debug your code instead?

[lost flume]

Correct answer is

6
6
6
6
6
6

[minor spruce]

well unfortunate :/

[blazing scroll]

ask it like 'continue the above code"

[minor spruce]

and? it apologized too?

[blazing scroll]

then you need to change prompt to 'Continue the code from where you left off'

[minor spruce]

Yeah be more specific like continue the code from this line or this comment

Glad you got your code fixed 😁

[humble zenith]

After playing with function calling for several days, I can say its basically the same as before, but with 30% chance u will get something correct 😄

[minor spruce]

function calling? what is function calling?

[drowsy lily]

yes

[humble zenith]

The recently introduced feature, which can let GPT-4 and GPT-3.5-turbo-16k to provide functions and let GPT to use them

[minor spruce]

oh

[minor spruce]

what can those functions be?

[humble zenith]

https://openai.com/blog/function-calling-and-other-api-updates

[terse furnace]

I want to integrate GPT into a game server panel to analyze crash logs from games like Minecraft Forge (which can around 80-100K characters in length). What model should I use? gpt-4-32k looks like it can handle more, but is 4x more expensive than gpt-3.5-turbo-16k. I also want the user to be able to have a continuing conversation with the AI, so it can guide them on how to fix their errors. I estimate that using gpt-4, it will cost $3 per crash. On average, let’s say users crash their servers 4 times a day, multiplied by 100 users, then I would be spending at least $1K per day and even more depending on the users’ needs

[fickle aurora]

Is anyone else have trouble with gpt 4 and 3.5 returning incorrect info? I feed it very specific instructions like ... "I am writing a summary for a resume. Write a paragraph that is 4 lines long that includes attributes that would help position this person for x (I fill in the role here). Write this in first person with no personal pronouns." and then I provide information.
One day I used all 30 of my messages burning through trying to get it to return the proper information. It would write 5 paragrapsh filled with me and my and then I would ask if it new what a personal pronoun was - yes it answered my and I. Then I reminded I did not want it to use it. It responds " sorry for the confusion here it is rewrittten without personal pronouns and then it would return with personal pronouns again. I would ask how many paragraphs did I ask for? It would return "You asked for 1 paragraph" and then it would again give me 5. I am not sure if I am not being specific enough although I am not sure how I can be clearer. I actually numbered the instructions as well. Any thoughts?

[remote epoch]

I can no longer use GPT4 without this incessant protection ....

[spare edge]

hi dear, i subscribe to personal account

why can access to gpt4?

[cyan gust]

Hello I'm new and I don't have many experiences about AI. How can I resolve tht: RateLimitError: You exceeded your current quota, please check your plan and billing details.

[atomic robin]

your prompt is ambiguous. what does "4 lines long" even mean? who does the AI know the length of a line?

[mystic plume]

I have to solve a captcha after every single message I send to ChatGPT 4. This started yesterday. Does anyone have an idea what could be the problem here?

alright, thank you!

[knotty grotto]

Delete cookie and cache

[digital path]

Is Bing Chat using the Multi Modal GPT-4 with new option to share images in chat ?

[knotty grotto]

It use gpt 4 but not chatgpt 4

And it use dalle technology

[spare edge]

when wll be avaible gpt-4-32k-0613

[remote epoch]

i had the same problem i deleted cookies and cache and it seems to work for now, thanks to @wind urchin

[modest lily]

i have premium and still getting: Oops!
Our systems are a bit busy at the moment,
please take a break and try again soon.

yes i got an email for premium link and still getting error

i followed the link the email and logged in and got"

Oops!
Our systems are a bit busy at the moment,
please take a break and try again soon.

[lost flume]

Always ask "are you sure "gpt-4 after it gives answer

[modest lily]

i do cause i use chatgpt 4 everyday lol

and getting charged monthly

[wraith ginkgo]

`` Server Tip: ``

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

[modest lily]

i have multiple times, so dumb whats the point of paying

[near wind]

same problem here

[modest lily]

where is the contact support?

[near wind]

nope, still not working

[modest lily]

yup nothing

[near wind]

the site seems pretty slow in general

[iron nymph]

same problem for me too, deleted cache, tried multiple browsers and os

[modest lily]

i get its slow but when you pay premium to by pass that i expect it to work, its not like the monthly price is cheap

[wraith ginkgo]

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.

[modest lily]

where is support

[iron nymph]

i finaly got in!

[fossil pier]

GPT at capacity, paid user, everyone getting the same thing aswell?

[near wind]

yes

[fossil pier]

cant login lol

[blazing summit]

Is it possible to implement the GPT-4 in Discord as a kind of live translator?

[faint flower]

dang paid and still locked out... RIP

[worn pivot]

Same

[loud quail]

yes impossible to login, timeout

[untold void]

where can I deaktivate this account protection?

[narrow saffron]

\

why does gpt 4 have a damn cap of 25 messages every 3 hours after paying as well

[supple violet]

yes if you pay for it

[narrow saffron]

can I get refund, I just purchased today

[faint flower]

xD

also I signed up free as apple user... then paid later... I get to personalized link page, but this does not work for apple account users apparently. .... in any case the mobile app seems to be working still

[fossil pier]

I continued my subscription yesterday and forgot to cancel, if I unsubscribe now, will I still have my gpt puls service for the next 29 days?

[faint flower]

If it didnt let you that would be crazy

[fossil pier]

preferrably i wanna get a refund but its on me for forgetting so thats mb

forgetting to cancel on time*

[faint flower]

Guess its working now

[near wind]

Yes, with OpenAI's API

[near wind]

nah still not working for me

[faint flower]

Got into the desktop 4 model now

[fossil pier]

nah, can login now, but no service

[median kiln]

anyone have any recommendations for how to work with the context limits of gpt4's API?

[knotty grotto]

Imo,I suggest maybe ask in #dev-chat maybe some people have more knowledge there.

[median kiln]

would i have access to longer context limits through gpt premium via the website? even though it's 25 messages in 3 hours, i presume the context limit available is 32k whereas the API is still 8k? is that correct?

[knotty grotto]

Good question, I didn't try the context limit with gpt 4 with sub.

[median geyser]

Why isn't the code interpreter available to eu gpt plus subscribers?

[knotty grotto]

It suppose to be rolling out during this week. #announcements .

[supple violet]

im from eu and got it on friday

[bleak crypt]

I can enable it but i don't see the code chat input / window.

I would like to find detailed instructions but Google/bing/openie don't have them. Can i put in a prompt that would teach me?

[supple violet]

you select it from the gpt-4 menu

[proven iris]

I'm sorry for any confusion, but as an AI text-based model, I don't have the ability to access or analyze images via url or any hyperlink. I can assist in generating text, information and answer your questions based on the text you provide.

why i get this gpt 4 ?

[knotty grotto]

It isn't connected to internet . It can't analyze picture or provide link.
Browser option is disable for now

[proven iris]

he could just read the pictures and suddenly it happened like this?

[bleak crypt]

I enable it, i start the chat, i don't see any button to start typing the code and run it

[knotty grotto]

It can't, #gpt-4-faq . It might hallucinations for now. But this function isn't available yet.

[mystic pulsar]

qq

[proven iris]

which ai understanding images?

[supple violet]

you talk to it like with normal chatgpt and it does the code

[knotty grotto]

I suggest to ask in #ai-discussions maybe someone could have idea

[tawny mortar]

I am paying for access to this service and I can't even login

What is the point of paying $20, only for the service to say it is too busy, check back later

[knotty grotto]

According to the tos
OpenAI makes no representations or warranties for Beta Services, including any warranty that Beta Services will be generally available, uninterrupted or error-free

[tawny mortar]

lovely

[knotty grotto]

Error can happen, follow status.openai.com to see the evolution

[tawny mortar]

"Oops" been getting this for the past couple of hours

[split edge]

😇

[knotty grotto]

This error is happening to several people including us the sub user

You can try this : on phone app if iPhone, phone browser for Android. Sometime it can works

[ivory vector]

I am a fu*&& paying customer and I can't even use GPT

what the hell is this

[mighty plume]

Me too what a pity

[knotty grotto]

#gpt-models message

[untold drum]

Hi~

[wintry jasper]

are they planning sometime to add a simple search conversations feature??? I mean this is basic right?

[main aurora]

its weird because the app has a convo search feature

but not desktop

also chatgpt is down for me too and im bummed :{ no acknowledgement or nothin

[narrow saffron]

Guys pls help

does gpt-4 mobile version have limit on messages using superpower chrome extension

or is it unlimited

[knotty grotto]

Chatgpt 4 doesn't offer unlimited prompt

[narrow saffron]

Someone said this on reddit, "If you use the iOS app, yes. Selecting "GPT-4" in the app uses a model which OpenAI calls "GPT-4 (Mobile, V2)". There is no cap on this model."

[knotty grotto]

Really doubt that is true.

[narrow saffron]

idk ig I'll try

[modest viper]

I first subscribed GPT4 for my friend on my phone. Subsequently, I logged into my openai account to subscribe GPT4, but the new subscription was automatically recharged to my friend's account.
I encountered an error message stating, "This subscription is associated with another OpenAI account. Please login with that account to access it, or visit help.openai.com for support." 🙃
I am confused about why this happened. Is it not possible to use one Apple ID to pay for more than one OpenAI account? Moreover, there was no prior notification about this limitation. I would appreciate some assistance with resolving this issue.

[civic sparrow]

is there any news on when they will be bringing back browsing?

[knotty grotto]

Unfortunately they are no eta available

[serene cipher]

stay tuned to #announcements

[edgy meteor]

hey, i have a question. I'm new to the ai world and i have just used opus.pro to create reels but they get copy righted. where do you guys suggest i go to get the answer to this solution

question*

[serene cipher]

#ai-discussions is a channel to talk about non OpenAI AIs

[edgy meteor]

yes i understand, thats why im asking if you guys know a chat i can join on the topic im having trouble with

[serene cipher]

_that would be #ai-discussions _

[edgy meteor]

thank you!

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[hasty gale]

It'd be pretty cool to get some sort of math-gpt so it'll finally be able to do complex math

[serene cipher]

that would be the Wolfram plugin

does exactly that =P

[hasty gale]

Sorry but that doesn't suffice for differential calculus, lagrangians, z-transforms, laplace, convolutions

maybe 5+7

[serene cipher]

[jolly vessel]

Left bro speechless

[light heron]

I literally joined to complain about the ridiculous CAPTCHA I have to fill out every other message for my PAID account. When is this going to get removed??

[knotty grotto]

Problem is often on user side. Can you try incognito as test

[wraith ginkgo]

`` Server Tip: ``

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

[fossil hornet]

Hello guys
I'm currently using the GPT-3.5-turbo-16k model and am facing some issues with the variability of the output. I'm trying to generate a set of 30 fill-in-the-blank English language questions, each with 4 options. For each question, I want the model to provide the correct answer, a brief explanation, and the skill practiced, all together in a specific format.

Despite setting the temperature to 0, I'm still getting inconsistent outputs. Sometimes the output aligns with what I'm looking for, but other times it deviates. I've tried refining my instructions and prompts to make them more explicit, but haven't been able to achieve the consistency I'm looking for.

I understand that GPT-3 is probabilistic in nature, but I need my output to be highly deterministic for my use case. I would appreciate any advice or suggestions on how I can modify my prompt or adjust the parameters to make the output more deterministic and consistent.

Any recommendation?

[haughty harbor]

does anybody know when bing will be enabled again?

[amber basin]

When, if ever can we expect GPT-4 to be released for free to signed in OpenAI users, like ChatGPT is?

[hasty gale]

released for free, or licensed for free

[amber basin]

released for free

[hasty gale]

I don't think you know what released for free means

[amber basin]

enlighten me

[hasty gale]

may as well gpt it

[near cave]

I pay for gpt plus but now I’m on a free membership today and don’t know why. Does anyone know why this is or how to fix it?

[amber basin]

what would chatgpt be in this case

released for free or licensed for free

from what ChatGPT told me, released for free includes the API being free

[hasty gale]

licensed, id advise doing your own research for the restrictions, because I don't know what they are off of the top of my head

licensed and released are not the same thing

[amber basin]

ah i see

well then i meant to ask when GPT-4 might become licensed for free

[supple violet]

never

[amber basin]

fair

[hasty gale]

could be when they have a better gpt. Or if some other ai company has a better free ML model

[amber basin]

ah i see

[hasty gale]

gotta go, yall take care

[amber basin]

you too

thanks for building my processor too

[ionic tide]

After a chat has started can I remove a plugin?

[wraith ginkgo]

`` Server Tip: ``

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

[inner trail]

No

[hexed shell]

I honestly think it’s impossible. But also I’m curious as to why you need deterministic output. It seems like your task requires variability

[inner trail]

Through ChatGPT you mean?

If so then no

[woeful yarrow]

L

[primal latch]

i only have to say think step by step if i want a better answer?

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[blazing halo]

hi, will you get gpt 4 api access if you just subscribe to chatgpt plus right now?

[opaque shuttle]

Independencia 107 colonia Morelos

[forest hemlock]

No

[wraith ginkgo]

`` Server Tip: ``

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

[kind plinth]

¯_(ツ)_/¯

[gentle ore]

Do we have any idea on when the bing search option will be back?

[inner trail]

Soon™

[gentle ore]

Okay cool, thank you

[thorn sable]

Does anybody know if ChatGPT4 would get an upgrade regarding token quantity in inputs/outputs?

[mild urchin]

Hey everyone, i'm writing to know if there is an eay solution to make gpt read a pdf of mine and then write an blog article about it

Also, i am wondering how can i get access to alpha features

[arctic wedge]

same is there a way to get acsess to these features without paying

[wraith ginkgo]

`` Server Tip: ``

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

[pure trench]

Hi, I am exploring plugins and prompts to create scaffold notes from PDFs for my study notes, but I am getting variable results. Any pointers would be much appreciated.

[modern plank]

is there a way i can check how many uses of model 4 i have left until the next refresh?

[lusty pike]

Is the limit 25/3hrs still

[blazing scroll]

yes

[lusty pike]

Any idea when that will change

[blazing scroll]

no idea

[lusty pike]

Dang

[modern plank]

you also cant tell when you're about to hit it

[blazing scroll]

if you want unlimited then check out api!

[modern plank]

where is that

[young agate]

hello I just pay for gpt 4 and idk why I have limited access

[unkempt saddle]

chat gpt 4 has been answering my questions anyhow for 1 week, it is no longer as efficient as before

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[young agate]

i paid for gpt 4 to use gpt 3

[inner trail]

You are talking nonsense

[young agate]

25 question every 3hours it’s nothing

[inner trail]

Not true, if you manage 25 message carefully are more than enough

For 3 hours

I've actually never hit the cap, or rarely

[young agate]

but don't forget that these 25 messages are not free so you pay for limited access

especially if you are a developer and you come across errors you quickly exceed 25 messages

[tranquil tiger]

Yeah, it's better to have 200 messages per day rather than 25 messages per 3 hours. So it will be possible to work hard and then rest, and not endure and wait every 3 hours.

[waxen trellis]

this code interpreter is an absolute game changer!

[waxen trellis]

it is slightly annoying that bing users get 30 queries for free (at microsofts expense?) and we are limited to 25

[knotty grotto]

Bing use gpt 4 technology but it isn't the same as chatgpt 4

[pure trench]

yes, I am afraid so

anyone wanna work on a plugin with me?

the 25 Q per 3 hour cap is very annoying.

[cerulean obsidian]

THEY NERFED the API too for causal talk about mental health are you serious?

Latest changes to the API has nerfed the API for any even slight mental health talk seriously this was not expected I had to click on old models and it still works, but for new latest model it's not anymore

[restive dirge]

if you need more, create a second account, share a chat with the second account and continue on it, but you would hardly agree to pay an additional 20 dollars, right?

[wraith ginkgo]

`` Server Tip: ``

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

[undone quest]

why does chatgpt change the model from 4 to 3.5 in the middle of the conversation

its so annoying

[inner trail]

Did you run out of messages?

[undone quest]

no

[inner trail]

Strange

[undone quest]

and when that happens, the model doesnt go to 3.5 automatically

I only passed the limit once and it didt go to the 3.5 model

now I have to create another chat...

[inner trail]

That never happend to me, unless i ran out of GPT-4 messages

[cerulean crag]

Why doesn't the code interpreter understand images? I tried it a few days ago when it launched and it could do it just fine

[knotty grotto]

?

[wispy skiff]

They deff dumbed down GPT 4 now too, what a scam

Guess that s what you get when you`re owned by Microsoft

[torn rover]

hey guys, I'm wondering if they whacked GPT4

s = "Hello World!"
Slice string into two parts
part1 = s[:-3]
part2 = s[-3:]

Insert a comma at the third-to-last character
s_new = part1 + ',' + part2
print(s_new) # This will print "Hello Wo,rld!"

It was not even coming close to making this mistake in April and May

(the print statement has the comma in the wrong spot)

[worn badger]

An off by one error? Even amazing programmers make this kind of mistake with some frequency

In other words, I bet you'd find this kind of error in any large training data set. GPT-4 is going to be at least as "whacked" as the average human programmer's brain

[torn rover]

yeah not denying GPTs amazing usefulness

just surprised, this is a process error that I haven't seen until this last month

it usually had those on lockdown. I would see this akin to it mispelling the word "variable" in an explanation. Weird for a computer to .

[pseudo warren]

Hey hey, did anyone else see the Insider article today on GPT-4 getting "dumber" and "lazier"? Speculation is that some big model changes were made to boost speed in GPT-4 but it may have reduced accuracy and recall ability.

[pseudo warren]

hey friend, I didn't see your post -- just asked about this due to an article I saw today in Insider

[torn rover]

where is the article in?

[pseudo warren]

Character Zero is not totally wrong, there's quite a few users complaining about this

[pseudo warren]

I can't post links without getting a timeout, but just web search "The world's most powerful AI model suddenly got 'lazier' and 'dumber.' A radical redesign of OpenAI's GPT-4 could be behind the decline in performance." It's from Business Insider

[torn rover]

yep just did

[pseudo warren]

Yeah something about a "Mixture of Experts" design

[brazen merlin]

For myself, it seems repetitive in it's Reponses

[pseudo warren]

Ultimately I'm going to be pretty bummed if GPT-4 isn't as good anymore, I definitely wanted the accuracy over speed based on my use cases. Might test other models out if this "change" is permanent

[pseudo warren]

Me too. I'm also getting weird responses, like it kept trying to answer me in Japanese...

[worn badger]

Interesting! I could certainly see them nerfing it for speed. I guess I thought that tasks involving precise math were always a weakness for GPT, but maybe at one point GPT-4 had this on lock?

[pseudo warren]

As you can tell, I speak English 🙂

[pseudo warren]

It's also getting suckier at code outside of Code Interpreter IMHO

Where's @knotty grotto at? Neko seems to always know everything.

[normal wraith]

hi there,
i want to buy gpt 4 to solve some physiks question for me, but before i buy i want to know if it even works. Gpt 3 gets close but often results are of by a few decimals.
So can someone pls paste this in Gpt4 and tell me the result you are getting?
got a time out lol because its german ; " Für rotes Licht der Wellenlänge λ rot = 660 nm tritt das 1. Hauptmaximum bei einer Beugung am Gitter unter einem Winkel von α = 7.3° auf.

Berechnen Sie die Gitterkonstante.

Ergebnis ist in μm
anzugeben."

[supple violet]

language models are poor with math, physics etc. so if that's your main usage then dont get it

[knotty grotto]

Hewwo , I don't know everything but I can try to help

[torn rover]

yeah I'm happy you are noticing it to. so it's not just me. I'll repost what I showed earlier:

s = "Hello World!"
Slice string into two parts
part1 = s[:-3]
part2 = s[-3:]

Insert a comma at the third-to-last character
s_new = part1 + ',' + part2
print(s_new) # This will print "Hello Wo,rld!"

To me for GPT to miss this (given how accurate it tends to be on process-stuff like this), it's akin to missing 3+3 = 6. There's not a lot under the hood here.

*too

[pseudo warren]

Neko, what's the scoop? Why does GPT-4 seem less performant now?

It made headlines so it can't be all illusion

[knotty grotto]

I think they might have bug at the moment.
Keep an eye on status.openai.com . They will update it

[pseudo warren]

looking now

nothing big that I see there. shame! thanks for the tip.

[pseudo warren]

Yeah my workaround right now is to use Code Interpreter because at least it debugs itself

[knotty grotto]

They update the website manually . Based on the last two day it take around 30min before they show something

[pseudo warren]

Got it. Only problem is that users have been reporting degradation for weeks now. Use your powers of clairvoyance!! Lol

or however that word is spelled

[torn rover]

interesting, haven't used Code Interp yet. Never needed to

[knotty grotto]

I believe in openai team

[pseudo warren]

True, I'm sure they are aware of this by now. Curious to see how they react.

[mystic girder]

Guys what is the current 3h message limit?

[torn rover]

yeah I amsure it's intentional

it's really powerful software. I'm sure there's other parties involved

[knotty grotto]

3h/25 prompt

[mystic girder]

Ok ty

[pseudo warren]

For those who are depending on it for enterprise production usages though...they better not pull under-the-hood bait and switch like that! Otherwise you better believe there will be lots of glances at the other models in the room (pun intended)

[torn rover]

yeah for sure

[umbral zodiac]

has anyone explored putting entire codebases to gpt? i easily get hit the token limit when doing so.. wondering if there are any existing websites that does this

[mystic girder]

This is not really possible unless you can fit it into 32k tokens

[foggy lantern]

Has there been some 'accuracy' update pushed to GPT4?

GPT4 is refusing responses, where it's being asked to estimate something simple like iron content in a recipe.

ie) I have 3 recipe's down to specific ingredients and measurements. I'm asking for an iron estimate per ingredient and for total recipe, it refuses by telling me about "not knowing and things often being varied and that I should look it up on a nutritional site,", while 3.5 does it for me no problem.

This started today, yesterday it would give me a response and estimate 50% of the time based on my prompt.

Something odd is going on.

[wraith ginkgo]

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.

[mystic girder]

But does 3.5 give you the right numbers or are they just random? Gpt-4 might be trying to avoid spreading misinformation and messing up someone's meal plan

[foggy lantern]

3.5 seems to give a pretty accurate 'estimate'

Which is what I asked for...

[pseudo warren]

I was posting about this earlier, there was a Business Insider article that made a splash about this, lots of users reporting GPT-4 accuracy and recall decline over the last several weeks

#gpt-models message this message has the search you can enter if you're interested, I can't post external links here without a timeout

[foggy lantern]

Interesting... makes me wonder what I'm paying 20/m for...

Quality seems far worse and less creative since I first got it months ago, limits still pretty much the same.. excpet now it won't even give me a basic estimate that my open-source models will...

[pseudo warren]

I agree. They need to turn it around asap or Bard/LLAMA etc. are looking better. Especially not a good move for big-time players leveraging GPT-4 already.

[ionic tide]

Did Gpt go down again?

[covert haven]

everyday he goes down

[ionic tide]

Phrasing haha

[knotty grotto]

Hi, you can check the status.openai.com

[ionic tide]

Degraded it says

[dreamy fractal]

#tangodown <@&1007805617555124274> Too much #errors today!!! #ChatGPT #offline #networkerror

[sacred cypress]

Is gpt-4 capable of providing youtube links natively, or would a plugin be required?

[knotty grotto]

Browser will be required

[near prawn]

hi to all of you

[slender pivot]

Wow

So, I have the paid version, but haven't used it for a couple few weeks. I don't know what happened, but I'm trying to use it today and it is totally borked

[knotty grotto]

Hi, you can check the status.openai.com

[slender pivot]

I was using it primarily for coding related stuff, but it doesn't even remember a couple of lines up in the same conversation and can no longer tell even the correct type of code we're looking at even after I specify what it is

[slender pivot]

I'll check

"Degraded performance"

Says they put in a fix but from what I'm seeing it's pretty trash at the moment

[knotty grotto]

Starting to get back

I shower a result in #off-topic

[slender folio]

Hi when will I get access to gpt4 api? I have a CC on my account

[slender pivot]

Still trash

[knotty grotto]

Hi you can check #announcements

[slender folio]

Doesn't look like I've had a payment cycle yet. Thanks. It'll probably be open to everyone before the next cycle is billed I take it :b

[deep cape]

When will there be a free version of gpt-4?

[wraith ginkgo]

`` Server Tip: ``

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

[tribal hare]

probably not free but u can get its access from microsoft bing

[deep cape]

Bing AI is using gpt4?

[knotty grotto]

You can use bing for gpt 4 but it isn't the same experience and as good as chatgpt 4

[tribal hare]

true

[deep cape]

Oh lol rlly?

What’s the difference? I mostly use gpt for idea generation, light research, and writing stuffs

[tribal hare]

in compare with that free version of gpt3 is better than microsoft bing ai chatbot

[knotty grotto]

For complex stuff chatgpt 4 is more reliable

[deep cape]

Oop 😭

[deep cape]

Ok thx

[rigid estuary]

English only please

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[lusty pike]

Is it possible to get more then 25 prompts per 3 hrs

[fair iron]

have multiple chatGPT Plus accounts

or use the GPT-4 api

[lusty pike]

Wheres that

[fair iron]

https://platform.openai.com/

[uneven berry]

Guys what is the difference between 8K and 32K?

In the API pricing

[unkempt orchid]

An error occurred while processing the plugin.
Nice webpilot plugin 👍

[fast barn]

Bro gpt 4 is like gpt 3.5 only a little better

If you had plugins and code interpreter for gpt 3.5 it would be the same

[solid remnant]

not really, since bing chat uses GPT-4, and yeah I understand that bing chat is designed for search engine purposes but there are ways around it

Like, I have got working prompts on Bing Chat

it refuses a lot lot tho 😂

due to breaking certain policies I believe, but it would work on ChatGPT smoothly, so it's some buggy stuff on Bing Chat's side

[remote badge]

How does one get GPT-4 enabled?

I have a card added but I don't have access to the model

[supple violet]

if you mean in api then you also need real payment history

[modern plank]

how do i check how many uses i have of gpt model 4

[remote badge]

I have paid for chatgpt plus, does that not count? how do I get it to submit a payment?

[uneven berry]

How

[remote badge]

gpt 3 is kinda useless, I tried using it but it makes too many mistakes and loops a lot

[old granite]

i started up a gpt-4 chat today and it says "used unknown plugin" when doing webpilot, it still works fine right?

[supple violet]

it does that when a plugin is in unverified state, which eg. happens when its waiting update review

[old granite]

its trying to use webpilot though, which was working fine literally yesterday

oh update i guess

ok thanks

is there anywhere i can find an ETA on when itll be back?

[supple violet]

after openai has checked so probably within 24h, author tweeted this "I just made an update to the WebPilot ChatGPT Plugin, boosting its speed and power.
During the review period, the plugin will be temporarily unavailable and displayed as "Unverified." No need to worry!"

[old granite]

ok cool, ill just wait until tmrw then. its normal that it doesnt work during this period though?

[supple violet]

it should still work if you have it installed

[old granite]

it tries a query but then gets stuck

its been "using unknown plugin" for ~15 mins

[supple violet]

start a new thread

[old granite]

bug report?

[supple violet]

no chat thread

of try refreshing the page

[old granite]

oh

yeah i do want to keep my progress because its writing a longer-form response over multiple chats

[solid remnant]

Since Microsoft owns part of OpenAI, it can use the GPT-4 model for their own purposes but the problem with it even tho it uses GPT-4 is that it is designed to be used as a search engine, and it's main problem are prompts that "break" their policies. It should work fine if you alter the prompts a bit that you think "breaks" their policies

[wraith ginkgo]

`` Server Tip: ``

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

[next jewel]

Has anyone else noticed a change in ChatGPT-4's memory after a May update from OpenAI? Previously, I felt like I could get about 20 responses in the same thread without needing to remind the model what we were discussing. But now, it appears to only recall the last two responses, requiring me to reiterate the original prompt with every request to stay on topic. It seems the model’s reasoning ability has lessened unless I craft a very specific, detailed prompt. Is this a universal change or have there been adjustments just to my account?

[lapis bramble]

not only do we still have usage limits for gpt-4 , gpt-4 is clearly not as "smart" as it was when it first came out, whats going on? it defintaley has changed for the worse

takes me around in circles like gpt-3.5, not as nuanced as it used to be, forgets as easily too

[next jewel]

I wonder if the recent changes were a response to some harmful activities, leading to an overall downgrade rather than targeting specific accounts. Or perhaps, these were intentional moves to pace our progress and creativity. I'm not sure, especially since I haven't seen any official announcements acknowledging these changes. Even though I'm still paying $20, the quality seems to have decreased since May, making me feel like I'm merely buying early access to new features. I don't intend to sound cynical or bitter, but it could be a scaling issue. Maybe we enjoyed superior services when the user base was small, and as it grew, resources were distributed more thinly to cater to everyone. At this point, the $20 for GPT-4 doesn't seem as worthwhile compared to using 3.5. The only reason I haven't cancelled yet is to maintain access to the plugins and code interpreter.

[lapis bramble]

yep not worth the $20 unfortunatley atm,

[jade coral]

I am a motion graphic designer in Iran. How can I get a freelance dollar project from European countries?

[lapis bramble]

a response from Open AI would be good, there have been countless people not impressed with Gpt-4 in recent times

[next jewel]

Let's hope for some form of recognition here, so we don't feel left in the dark. Ideally, this would also give us hope for a possible rollback or an upgrade in the near future. 🙏🏻

[grizzled flax]

It may be helpful to offer "before and after" examples (via chat share links) that illustrate what you're reporting. When I've seen OpenAI employees respond to reports like this on Discord, that's what they've asked for.

[next jewel]

To illustrate, let's say I presented a book outline with 10 chapters, along with the introduction and first chapter. In the past, I could rely on GPT-4 to write and edit the entire book without having to repeatedly remind it of the initial outline. Nowadays, I have to reiterate the outline every couple of responses. If I don't, it starts to modify and insert irrelevant elements into the original plan.

There's another example. I used to instruct GPT-4 with the phrase "simply continue," providing no context or reminders, and it would comprehend precisely what I intended, much like a human collaborator on a project. This was possible for numerous prompts, even up to 20. Now, that's simply not an option. It fails to remember or intuit what I'm implying.

In essence, before May, working with GPT-4 felt like engaging with a true human collaborator on projects. Post-May, the experience has regressed to something similar to the 3.5 model. Instead of feeling like a human collaborator, it feels more akin to older AI models like Siri or Alexa.

[grizzled flax]

If you have specific chat share links for before & after examples, it may be helpful! Though I get the hesitation to share private chats. It's just one way they may be able to look into your report beyond the qualitative descriptions you've offered.

[next jewel]

This feels similar to dealing with a customer support team from an Internet Service Provider. They typically instruct us to perform basic troubleshooting steps we've already tried, like rebooting or checking connections. It's a tedious process that we're asked to repeat. Feels like gaslighting.

However, I digress. I'd imagine a company as advanced as OpenAI, the creator of this technology, would have superior capabilities. If I provide specific timelines, such as events occurring before or after May 1, along with my examples, they should, theoretically, possess the necessary logging and technological insights to accurately understand my concerns.

I also wonder if any changes made were deliberate. If alterations were made due to various pressing reasons or to accommodate scale, I'd appreciate knowing that. It would be reassuring to learn future updates will improve as hardware scales up, instead of relying on software scale-downs.

However, without any acknowledgement or transparency, it's disheartening. I'm left in the dark, unsure if this is the best approach for humanity or their paying customers, who are investing $20 a month and additional API usage fees.

[grizzled flax]

I'm sorry you feel gaslit, I assure you I'm just parroting what I've seen OpenAI employees ask for when this has come up in the past! Though they indeed likely have access to a large quantity of example chats to check, I assume the request for specific chat share links comes from a desire to take a look at potential specific changes in behavior instead of scanning for this among a huge dataset.

[rocky path]

I read somewhere that when gpt4 was made public instead of paid that chatgpt has gotten worse at accuracy.

For writing/debugging software, 3.5 was endless loops of wrong answers and I started paying for GPT4 which was noticably different and got around 1/3 answers right after some back and forth telling it what its doing wrong.

I havent used chatgpt in weeks and today its noticeably different since the last time i used it - it seems like its just like throwing darts at wrong answers with a blindfold on again like 3.5.

I don't know if this is bias from what I read combined with an unlucky topic or if capabilities have been downshifted.

Anyone else have a similar experience?

guess i should have scrolled up 😉

[next jewel]

Understandably. I wasn't pointing directly at you. With phishing scams skyrocketing by 464%, I have no doubt that the early May update was meant to protect us. I'll trust in this and wait for them to put safeguards in place, restoring the functionality we had before May. In the meantime, I'll make sure to remind the GPT-4 model of our ongoing tasks with each prompt so it stays focused. We might still retain some of the power we had pre-May this way. Alternatively, I could consider giving Claude 2 a shot until OpenAI restores the memory, reasoning, and consciousness back to GPT-4 pre-May.

[smoky creek]

yoo whatsup

[unreal marsh]

how is claude2 compared to gpt4?

[next jewel]

Developers: ChatGPT 4 was created by OpenAI and Microsoft, while Claude 2 is a product of Anthropic, Google, Zoom, and Slack.

Architecture: ChatGPT 4 is built on a transformer-based neural network, while Claude 2 uses a recurrent neural network with a memory mechanism.

Capacity: ChatGPT 4 is designed with 175 billion parameters and 120 layers, while Claude 2 has 150 billion parameters and 100 layers.

Prompt Length: ChatGPT 4 handles prompts of up to 2048 tokens. On the other hand, Claude 2 can process an impressive 100,000 tokens.

Output Length: Similar to its prompt length, Claude 2 can output up to 100,000 tokens, dwarfing ChatGPT 4’s limit of 1024 tokens.

Applications: ChatGPT 4 is a versatile generative AI that can perform a range of tasks, including coding, translation, image analysis, and more. Claude 2, however, focuses more on specific tasks like summarization, creative writing, basic coding, and identifying legal risks.

[unreal marsh]

ok but is it good?

[next jewel]

My trust in Google & Anthropic is limited, so I've focused more on ChatGPT 3.5 and the updated, albeit scaled-down, GPT 4.0. I’ve read articles discussing Claude 2’s features and comparing them with ChatGPT 4, but none conclusively state that one is superior to the other. Several sources point out that Claude 2 is still in beta and has room for growth. Try it I guess? 🤷🏻

[static isle]

GPT-4 currently cannot analyze images

[next jewel]

Code Interpreter.

[static isle]

It can accept images, but can't analyze them

You cannot say a noun then constitute that as proof

[next jewel]

My apologies for any confusion, I wasn't actually trying to argue. I simply misspoke, it was an honest mistake. I hope you can understand, my good friend.

[static isle]

I understand and I apologize for my tone. I hope you can forgive me

[next jewel]

You are forgiven. 👊🏻

[static isle]

🤝

[dusk crypt]

"apologies for the constant confusion" - gpt-4

[forest hemlock]

I apologize for the confusion caused by my previous message...

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[steady bobcat]

I posted this same idea and question a few days ago, i’ve definitely seen a change / reduction in chatGPT-4’s intra-thread memory, this can kind of throw a monkey wrench into a brainstorming session or complex research.. wish they’d change it back

[wraith ginkgo]

`` Server Tip: ``

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

[next jewel]

Absolutely, I'm glad I managed to complete a few of my projects prior to the May update. However, I really hope they'd revert to the previous version, or at least openly clarify their reasons for the change. It would also be helpful to know when we can expect the original functionality to return. Will we have to wait until an open source model delivers it to us, possibly as a Christmas gift this year?

[steady bobcat]

There was a alleged leak
(you can find descriptions of it on YouTube) and I don’t know how valid the information is.) That the ChatGPT-4 platform would be receiving an update that would have a workstation environment - something like this seems like it would be ideal for serious research, in fact in hopes of something like this ive thought about putting off some serious projects I’ve been working on..🤷‍♀️

[unique ether]

You can use BrowserPilot instead

[flint bough]

Hello everyone I want try gpt-4 API when I use curl request it. I got "The model: gpt-4-0613 does not exist" error. how can I resolve it

[old granite]

webpilot has resumed working since then, thank you though

[unique ether]

oh nothing,

[unique ether]

You can also try out the difference between BrowserPilot and WebPilot

[next jewel]

Oh yeah. I heard rumors about that. I wish they wouldn’t have nerfed our current version while developing a new tier system. I'm keeping my fingers crossed that it remains affordable. @grizzled flax inquired about how it changed because OpenAI could ask. I recall a deep conversation I had with GPT-4. Around the 22nd exchange, it self-corrected an error it made in its second response. Intrigued, I scrolled up and then down, nearly falling over in astonishment. However, following the May update, it's a stroke of luck if it can recall our topic by the third or fourth exchange.

[steady bobcat]

The self correction you mentioned is really fascinating something similar happened to me in a similar manner, it only happened once. I’ve been thinking recently about how many core users there will be in aggregate, ostensibly all demographics within any population will be a user directly or indirectly in the future, however it’s core usage may have a counterintuitive result. An example could be using Myers-Briggs personality data across a population (I am well aware Myers Briggs posits controversy) where approximately 65%+ of any given population has a sensory Mode of personality operation ie use case would be doing homework, Hotel reservations balancing household budget etc.. in contrast 35%- of any given population is iNtuitive where creative introspection, Science and math would be the weighted use case, and this smaller population would utilize deeper resources of a model like Gpt4 (i’m just generally speculating here however there is reason this could hold true)

[neat escarp]

vue

[next jewel]

I agree with this assessment! As an ENFP, I've stumbled onto some unique ways to use AI tools. It's a big playground and the 'sensing' folks might have missed a slide or two. I'm still learning, and who knows? Maybe one day I'll doodle my own AI, just trying to be part of this arms race.

[lilac shard]

What does curl https://api.openai.com/v1/models
-H "Authorization: Bearer $OPENAI_API_KEY" say?

[steady bobcat]

You lucked out bro, you happen to have the best personality type 😉 (ENFP myself)

[flint bough]

It does't contain any gpt-4 model

[next jewel]

LOL! 🤣 Another example, after I posed two tech issues, GPT-4 even suggested a third one I wasn't aware of. 🤯 Before May's update, it felt like a real strong team. Now, it feels like having Siri or Alexa. Despite my gripes, I'm confident that OpenAI will note these changes that 36% of us users see, even if the majority don't. 🥸😎

[flint bough]

I check the official whitelist page. it does't exist. now its release gpt-4 note

[thick elbow]

Chaptgpt-4 yes

Chatgpt-5

[daring latch]

who are you?

[verbal socket]

who are you

#welcome

[hard nacelle]

I have noticed that too. Previously some of the short novels I was working on could go on for about 8 chapters before losing context. Now it's only about 5 or 6 chapters.

[surreal geyser]

who are you?

[gritty flint]

yo, guys i have made a group where people can talk about any subject they want with different people for eg we wanna do a project with 6 people which requires 6 different fields.. so if these 6 people have knowledge of their respective subjects they can develop that project together ... if u wanna join dm me

[fringe glacier]

how do I know if you are just wasting time?

[severe flax]

[trim quest]

How long until browsing is back as it’s been 10 days now

[modern plank]

so how do i check how many uses of gpt-4 i have left?

its funny how i got banned for saying something yesterday but the mod also didnt answer my question

[forest hemlock]

You can check that ATM.

[modern plank]

or do u mean "can't"

[forest hemlock]

Can't*

It was a typo.

[modern plank]

ahh i see

Thanks, that kinda sucks though fr

[long swallow]

do

[wraith ginkgo]

`` Server Tip: ``

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

[wise horizon]

what up

[lean helm]

sorry to barge in, i just joined for some advice, i know there are personality prompts for gpt4 like dan and whatnot, i want a dan prompt but i dont want any of those weird details to bypass stuff, i just want the personality bits... if that makes any sense

[subtle token]

Hey, guys! Is the web browsing function deleted eternally?

[tall basin]

They need to role back to older version. Even the vp knows it's dumber and tries to convince us otherwise.

[unique ether]

Dude, you can try BrowserPilot plugin, he can use web browsing function

[soft grail]

Hello,

I'm currently using the GPT-4 model from OpenAI and the /chat/completion API in NextJS.

However, I'm looking for a way to retrieve the latest information from the internet or Google regarding a specific question or organization. Since OpenAI models are pre-trained, they don't provide real-time information. I've already tried using serpAPI, which is used in langchain, but I haven't been able to obtain accurate answers like BARD does.

Specifically, I'm interested in obtaining the following basic details: the name of the organization, its founding year, revenue, headquarters, and industry. Additionally, I'd like to retrieve recent news about the company, recent posts by their leaders on LinkedIn, and any relevant technical news.

How can I achieve this using the available tools or APIs?

I've tried serpAPI from lanchain and expecting real time data.

[lusty pike]

Is there ever going to be no limit on prompts

[ember depot]

1111

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[serene spruce]

/key regen

/key get

[fading junco]

has anyone reached the 25 messages per 3 hours limit?

i feel like ive sent more then 25 messages in the last 3 hours but its not stopping me

[wraith ginkgo]

`` Server Tip: ``

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

[foggy plover]

Hello, i get error message like this
"Your request was blocked as bot traffic. If you are using a VPN please disable it and try again."
Tried to change the network, but still get same error.

[supple violet]

where are you using it from

[scenic widget]

is it Gpt 4 become not so consistent?

seems like a 3.65 or something

[knotty grotto]

maybe the prompting wasnt good? i dont have issue with mine looking good

[scenic widget]

I'm using same prompt, just for different group of results

that's why seeking statistic before speculation

[next jewel]

Have you considered using web scraping tools like Octoparse, or real-time data sources like Hooks Data that have their own APIs? Websites like RapidAPI and Nordic APIs can help you find more real-time data APIs.

[soft grail]

I'm expecting the result of what the Google BARD AI bot is giving. So what's your suggestions for that one ?

[next jewel]

Google BARD? I’m not familiar. My trust for google is extremely limited. However, for instance, you could use Octoparse to collect data like a company's name, founding year, revenue, headquarters, and industry from diverse sources like Wikipedia, Crunchbase, or LinkedIn. You could also get real-time data from RSS feeds, Twitter timelines, or stock data using Hooks Data. Then after gathering this information, you could combine it using a tool or framework you like, such as NextS.

[inner rapids]

Is there a way to check when the gpt4 messages get restored?

[knotty grotto]

wdym?

[inner rapids]

like what time the 25 messages come back

[knotty grotto]

oh unfortunately not yet. atm only trick it is 1 prompt every min to never reach your limit

[inner rapids]

ahhh ok thats good to know. Thanks! 🙂

[grizzled flax]

25 messages in 180 minutes allowed = 1 prompt every 7min 12sec

[knotty grotto]

If you want to be precise yes

[grizzled flax]

Was thrown off by "1 prompt every min" and thought maybe you accidentally a number

[hallow owl]

I think @knotty grotto meant one prompt every 8 minutes, probably a typo

[grizzled flax]

Yuh that's what I was wondering!

[knotty grotto]

Yes I did x)

[knotty grotto]

Sorry for the confusion

[lime star]

any idea how i can check if i meet the criteria of a "history of successful payments" to access gpt4? my billing history says i have made 2 successfully paid invoices, but they both have a value of "$0.00".

[sonic musk]

Anyone know a prompt i can send that is not a jailbreak but has a similar effect? I study cybersecurity and ChatGPT just refuses to help me with anything

[wind urchin]

is Code Interpreter available as an API or Playground or even as a Jupyter plugin or something? If not, any guesstimates as to when it will be?

[spiral dawn]

When Webbrowser come back? This is missed!

[supple violet]

all 0 means you were on trial or not enough usage, disqualified

[supple violet]

no its not and never been said it will ever be

[wind urchin]

Copilot2 is coming soon tho, there is a waitlist already

I guess that's what it is basically, it'll be like the Code Interpreter

[supple violet]

nope

completely different things

if you meant github copilot

[hardy basin]

Alo

[wraith ginkgo]

`` Server Tip: ``

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

[glad mica]

So what is the status on getting GPT4 search back? Any news or timeline?

[scenic widget]

we got to be precise

[balmy turtle]

Bruh

[next jewel]

Yes. They won’t let me send URLs so search up pinokio.computer and you’ll get what you’re looking for and more.

[next jewel]

I replied, believing I hadn't breached their rules, yet they erased my message. They privately warned me that they would ban me from this chat if I respond to your query.

[wintry mirage]

I have found multiple instances where claude2 can do far more than GPT can in programming. Its able to translate more specific output instructions as well as use various functions in more elaborate and imaginative ways than GPT could up to that point. Many of the batch files ive asked chat GPT to create it just couldnt, and couldnt say why. With Claude2 it got the right function, and answer first time. Anyone else used claude2 for coding?

[still seal]

Is there anyone know that if I can upload image to ChatGPT when I upgrade to the plus version?

[fair iron]

image uploads aren't currently supported

[still seal]

Well, so now I can just upload files to GPT-4 to let it help me analyze ?

[fair iron]

I don't believe you can upload files to GPT-4 outside of the Code Interpreter (that takes Python files)

[eternal quest]

yes and clause2 did very bad. i suppose it depends with what coding you ask it

[wintry mirage]

In this case it was creating batch files and powershell scripts. I also got it to convert code function from windows terminal batch files to macos scripts of the equivalent function. Its done better than GPT4, and ive made a couple dozen batch files with gpt4.

[still seal]

Yea, the new feature that I want to use is the code interpreter . But it looks like there are some restrictions to use it.

[fair iron]

you'll be able to upload Python files to it, but I've heard people have varying luck with whether it will analyze the files or not. You'll need a ChatGPT Plus subscription to access it, as well

[sonic musk]

thanks

[eternal quest]

yea i think thats quite specific so maybe, when i use it for more complex and analytically stuff gpt4 has more reasoning and context with whats required

[still seal]

Ok, thanks for your reply, but it seems it will take a time to wait for OpenAI opening the access of uploading more data formats to ChatGPT.🥺

[fair iron]

yeah

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[flat laurel]

hey hi just a quick tip, i just found out that gpt4 responds REALLY well if when you ask for code you say "code tersely and elegantly"!! i regret not having had that in all of my coding prompts before now, so i just felt like i should say it somewhere on the internet to help others not make the mistake i did,, give it a try, requesting elegance makes it uh not just write like such a careless scriptkiddie 😂

[distant jetty]

Help, how can I avoid the constant puzzles when using chatgpt4? Getting 10+ captcha a day

[frigid badger]

maybe its with your ios private relay. try to set it to have constant general location

[distant jetty]

Using Brave, shield turned off, cookie enabled

on desktop osx

can it be adjsuted at account level - for a paid service I feel like I need to work a lot on those captcha

[supple violet]

old osx or macos?

[iron whale]

hello guys, quick question is chatgpt 4 25 messages per hour over?

can you send unlimted as of 2023/07

[knotty grotto]

nope

[iron whale]

its still there?

[knotty grotto]

yes

[iron whale]

ugh

if they removed it i would buy plus 100%

[next jewel]

Also search up “gpt4all.” The website you’re looking for ends with io. It has powerful open source models that won’t treat you like a child.

[dusky iron]

Dear OpenAI please 🙏
Give our boi a little sentience
Amen

Just a tiny bit 🤏

I hope Code Interpreter gets a workspace sort of setup we can interact with directly

[tepid sun]

does anyone know why code interpreter is showing this?
with a red thunderbolt with a red line accross?

[dusky iron]

I believe code interpreter is under same constraints as gpt 4 where there’s a msg limit of 25 per 3 hours

[tepid sun]

@dusky iron I didn't even use chatGPT today.

[distant jetty]

macos sorry, why the captcha hell

[wraith ginkgo]

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.

[peak hornet]

Is there a timeline for when they plan on removing the 25 message limitation??

[fair iron]

there is not currently

[peak hornet]

Thanks

[onyx light]

Also facing this, not over the 25 limit

So something broken probably

[tepid sun]

yea, apparently its everyone.

[onyx light]

Back to 1985 debugging 😂

[tepid sun]

Lol. must be nice to know coding.
These AI tools have made my networking job easier by a lot.
For the past week, i have been working on a script which would create my network topology in draw io based on some show commands.
This alleviated a bunch of user mistakes when creating such documentation.

Man programming can be so hard, when you don't have time to study it.

[wintry mirage]

ive been pretty agressive about creating scripts for windows environment management as well. Got some pretty cool functions out of it and claude2's use. You want to collaberate on some scripts? Ive got about 20ish scripts ive fleshed out and integrated into a batch menu file that calls on the other scripts.

[tepid sun]

Sure. lets continue in private chat.

[wraith ginkgo]

`` Server Tip: ``

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

[balmy ridge]

I can't use it either?

[balmy ridge]

why did you delete?

[lusty pike]

Same

[wheat axle]

Yeah but as you see I don't know

[lost flume]

Code interpreter is down

[wind urchin]

is Bing visual input available for just some users without a waitlist?

like an A/B thing

[bleak crypt]

That's what she said 😁

i just joined to investigate . Hope they fix it soon.

[exotic oxide]

i would happily pay more than $20 a month for upgraded GPT4 performance

[wind urchin]

same

[wind urchin]

Lots of people might get it though and that would mean as more people get it the performance would decrease. Also they can’t just magic out performance - adding a higher performance option would mean not as many others get to use it untill they expand.

[exotic oxide]

no. its just a matter of server bandwidth. higher tiered Users would get more allocation

that's how it works

the pool of people would get smaller and smaller as the price gets higher and higher. i would like premium premium access

[broken patrol]

I used the entire raw transcript of a presentation i just did...about 29 pages from Otter...and had Claude 2 identify either explicit or implied questions I could use to generate an FAQ for future audiences. I then asked it to posit responses to each of the 20 questions, and then used those same questions to get responses from GPT 3.5 and 4.0.

Example question (audience is new to GAI):
"Could generative AI be used to create personalized learning plans or curriculums tailored to individual students' needs and interests?"

Was really interesting to see the variation in word count and was significant enough i thought I'd share. I did not evaluate response quality quantitatively or qualitatively...just reading through it, it seemed like GPT 4.0 struck the best balance in concise yet thorough responses.

On median, i got the following:
GPT 4.0 responses were 50% longer than Claude 2
GPT 3.5 responses were 190% longer than Claude 2, and 94% longer than GPT 4.0.

Statistics
Variable Mean StDev Minimum Q1 Median Q3 Maximum
Claude Word Count 26.40 5.01 16.00 23.25 26.00 30.25 38.00
GPT 4.0 Word Count 39.15 9.50 20.00 32.00 39.00 43.50 62.00
GPT 3.5 Word Count 79.05 16.22 40.00 72.50 75.50 87.25 112.00

[wind urchin]

umm

[graceful summit]

vist

[cunning rose]

Ok guys, who jailbroke the code interpreter so they had to take it down? We are still waiting for browser to return because someone decided they can play to be a web scrapper..

[keen basalt]

Hi all! Is it possible to turn on timestamps for chatgpt queries and responses so i can see the date and time of my queries and responses?

[static isle]

more server bandwidth = faster responses not better responses

The parameters don't change because the servers have more compute, they just get processed quicker

And thus the output doesn't change

[main mantle]

I swear they are making it dumber and dumber
Like it's so much harder to work with it now, it doesn't remember, it doesn't read a full message. Like I do exact same thing as three-four weeks ago and it just acts completely different

[normal quest]

Did I see in the latest release something mentioning if we pay for GPT we can get access to unlimited API 3.5 included with that?

[forest hemlock]

Didn't see anything like that.
You already have "unlimited" 3.5 API access, as long as you don't reach the rate limit and pay for it properly.

[wraith ginkgo]

`` Server Tip: ``

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

[charred falcon]

When are we able to get back on the internet with chatgpt 4?

[dull geyser]

does scenexplain often just stop working? I got it last night and it was working fine, but right now it saying it can't process images and can only give non detailed descriptions like this

"I apologize for the inconvenience, but it seems there's an issue with processing the image at the moment. However, I can still assist you if you can provide the data in text form.

The image you provided earlier appears to be a table or chart with numerical data. If you could describe the data or type it out, I would be able to assist you further. For example, if it's a table, you could list the column headers and a few rows of data. If it's a chart, you could describe the type of chart (e.g., bar, line, pie) and the key data points or trends it shows."

[slim shoal]

If you are using it right now

it is having issues

It was fine most of the day. But it is down for maintenance

[storm hearth]

Hi everyone! trying to use the code interpreter.. but it seems blocked anyone knows why? I used it a couple days ago.

[wary socket]

anyone know any plugin that can take information from linkedin_

[normal quest]

sorry i may have worded that incorrectly, I meant do we now have unlimited access to 3.5api that is payed for by being a plus user?

[forest hemlock]

No. That would never be a thing. Open AI wouldn't want to go bankrupt.

[normal quest]

haha well thats why I was confused. I have been gone for a bit so wanted to clear that up. Thanks

[storm hearth]

Has anyone else been having issues using the code interpreter today?

[hard ledge]

Plugins seem to be down. Anyone else getting the same error?

Error communicating with plugin service. Please try again later

[blazing scroll]

plugins are not down but Code Interpreter is

[dull dragon]

HOL

[wintry mirage]

Given GPT4 and Claude are both more than viable for simple scripting I decided to give instruction to have two systems talk to each other with me as the proxy. I provide it specific instruction, and copy/paste between the two until a testable solution for a code request is provided. I just tested it for windows user creation, and was very surprised how relevant the back and fourth was. Id like to share this interaction and LLM collaborative prompt. Anyone curious?

[static isle]

👍

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[fading junco]

When will code interpreter be back?

[wind urchin]

code interpretation will back soon

[fading junco]

how long

[wind urchin]

is cloud2 better than gpt4

I don't know

[fierce harbor]

Did gpt4 scale better than the Chinchilla Scaling?

Tried asking both Google and GPT4 and I couldn't get an answer

[dapper acorn]

Is it true that chatGPT can write lawsuit or contracts up to $1500 compensation or not?

I read something like that in the news a while back

And forgot if someone can please clarify

[slate flame]

does anyone else seem to have unlimited queries?

i have sent over 100 requests to gpt-4 in the last 3 hours

[atomic robin]

use common sense. is ChatGPT a lawyer?

[grave sleet]

i still cant use Code interpreter

[wind urchin]

Can you study online?

[surreal vault]

i was thinking i must have blown past the 25 message cap at least twice. maybe they opened it up since they have taken away web browsing and now code interpreter

[slate flame]

some people still seem to have it while others don't. all plus users. i wonder if we're part of some behind-the-scenes user group that openai made on the backend

[dapper acorn]

Definitely not but I just can’t remember where I saw this on an actual news article

[slow pine]

Does anyone know why the code interpreter doesn't work?

[wraith ginkgo]

`` Server Tip: ``

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

[wary imp]

How can I use chatgpt in this group please？

[scenic widget]

ok, I got someone on myside. 1 vote here too

https://status.openai.com here is saying interpreter is under maintainance

[fierce obsidian]

你好

[limber pine]

Great

[fleet jasper]

You can't use it here

[fleet jasper]

I'm experiencing the same. It barely seems more capable than 3.5 anymore. I use it for rpg scenario assistance. It was giving me MUCH higher quality responses a few months ago. I'm considering unsubscribing from 4 as it isn't providing much benefit over 3 anymore.

I also used to paste my meeting notes in to reformat and summarize them. A few months ago I was mesmerized by how high quality it was. Anymore I can't even use the results.

[lost flume]

Code interpreter is back!

[cold oyster]

Any idea on when gpt4-32k api will be available?

[hallow owl]

some people have it, a select few, otherwise just be patient

[fallen cipher]

Still no chatgpt Android app ?

[scenic widget]

ok, 2 votes for it's getting dumber. Can we create a poll here so that at least developers or openAi gets some feedback

[scenic widget]

I was replying to you and got removed by saying "ni hao" omg, what's happening?

anyone can type it without the double quote, it will "timed out" you immediately.

[proven fox]

I just bought the GPT-4 subscription to get my hands on the code interpreter but it's not showing?

[little nymph]

balls

nut

in ur mum

n!99er

[scenic widget]

did you get timed out from any other thing? "ni hao" without double quote got you right?

[unborn brook]

Why Am I not getting picked for plugin access ? Been a plus member for more than 5 months.

[hallow owl]

the API is separate from ChatGPT they are individual products that have no bearing on one another

[prime wing]

@hazy gorge make me fnaf r34 please

[hallow owl]

This is not ChatGPT, you can access ChatGPT here -> https://chat.openai.com

[unborn brook]

No i am not talking about apis. i am talking about plugins

[hallow owl]

you have access to Plugins now, go to Settings -> Beta Features then enable them

[unborn brook]

Oh yes got em

[wraith ginkgo]

`` Server Tip: ``

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

[whole kettle]

Does anyone have any idea when the training data will be updated to 2022 ? For programming it's starting to feel more and more outdated as the day pass. Yeah the web plugin is cool but the bot mostly still relies on data from its training library so it sort of breaks the point.

I guess using copilot instead would be a better idea since at least it has access to recent documentation.

[hallow owl]

this is a major sticking point right now, ChatGPT becomes less useful by the day when needing up to date information. I hope OpenAI plan to address this because even some Open Source models have recent data

[whole kettle]

Last time I checked it took around 6 months to train it from the ground up. They had more than a year to update the training data by now. I have a feeling they're working on GPT 5 which will probably arrive by the end of next year.

[verbal jolt]

does anyone else here at last have access to gpt-4’s awesome image recognition via bing?

[hallow owl]

yes

[whole kettle]

It's hard to make any sort of accurate estimations though as things change so fast nowdays. So in reality nobody knows what's going on I'm just speculating just to be clear.

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[valid scaffold]

Is there any news yet on when openai APIs will open up for gpt4 to all paying customers? There was such an announcement a while back, but we still only have gpt3.5 access in the api (and only 3.5 listed in the playground).

On my own personal account I have gpt4 access, but... I dont want to use my personal account for company purposes 😛

[hallow owl]

If you have spent money on API usage then you will have GPT-4 API access

[valid scaffold]

How much money do we need to spend, and do we need to wait for the next billing cycle so the payment has been made?

[hallow owl]

probably need to wait, do like $10 of usage

[valid scaffold]

Okay. Is there any way to just pay 10 up front? Because we really have no use for 3.5 or other models than 4.0 right now 😄

[hallow owl]

I don't think you can pay more than $5 up front

[valid scaffold]

So... The recommended way is basically to just send in some long texts to 3.5 until we accumulate 10 usd of use? and then wait? 🙂 (or use some other API function like fine tuning a diff model)

[hallow owl]

OpenAI say customers who have spent money on the API will be automatically granted GPT-4 API access, it's up to you to interpret that but I have GPT-4 API access because I came off the waitlist in March, your situation will depend on you

[valid scaffold]

ok I dont think that will be enough to convince my boss I'm afraid. I also have access personally to gpt4, but I'd like all my colleagues to get access too.. Company is happy to pay. But I dont think I can tell them "well I dont know, we can try to just use their other apis and pay for it, and maybe they will grant us access at some point" >D

[supple violet]

you cant pay anything up front, the $5 is a card check that is returned later

also the access was added manually on to the qualifying accounts so it would not help today to pay even if you found a way

[valid scaffold]

Are you saying even if I was charged for lets say 15 usd monthly usage, I would still not get gpt4 access?

[supple violet]

not if today or even before the end of the month, no guarantees if even then "Today all existing API developers with a history of successful payments can access the GPT-4 API with 8K context. We plan to open up access to new developers by the end of this month, and then start raising rate-limits after that depending on compute availability."

"we plan"

so far majority of their plans have been delayed

[valid scaffold]

ah ok. Thanks. Then there's no point in creating some dummy usage, well unless for speculation. Happy to do it personally, but doubt my company would do it

[supple violet]

people who had only ~$1 charge in the billing history got the access

[valid scaffold]

yeah we just created the company account 🙂. So it has zero in billing so far, and we have no use for other models right now than gpt4

[supple violet]

then its better to use the trial credits properly than to rush to unknown

[valid scaffold]

The annoying part is that many of us already have personal accounts with gpt4 access that we use daily. I spend 10-20/month on it on average (pretty low I know), and we would just like the company account to take that cost instead of us personally having to get reimbursed for the usage (plus I also use it for personal projects :S, so I honestly dont know how much of that is personal vs company use).

So yeah, we have been using gpt4 for a while and just dont want the hassle of having to go through our personal accounts 😄

[supple violet]

😅

[wary socket]

Do you guys know a plugin that makes it go inside linkedin profiles and extract information?

[hollow mist]

👍

[modest knoll]

Has anyone noticed that there is a large amount of variability in terms of the correctness and accuracy of ChatGPT’s answers throughout multiple “chats”?

For example, if I ask an advanced programming question to ChatGPT, certain chats may fail to answer the question correctly or accurately, and will struggle to fix any mistakes if prompted that there are mistakes in its outputs. But other chats will succeed at the request, and will be able to accurately make modification.

The chats that succeed at the initial question also tend to do better at subsequent unrelated logic/reasoning-intensive questions

For example, if I get a “good” chat that’s able to answer my initial question, then I ask it about an unrelated advanced programming question, it also tends to succeed at that, relative to the “bad” chats.

Could this potentially be anecdotal evidence towards the claim that GPT-4 isn’t one large LLM, but rather a set of LLMs combined via a monolithic architecture?

[safe venture]

I've had some really good results... i.e. having one chat create a constructed language and using another chat to profile and translate it

they've been rock solid in terms of remaining in sync

[limpid pilot]

has anyone noticed how gpt4 is giving better results than gpt 3 like it is alot better than gpt3

[slender folio]

Uh, of course? 😅

[limpid pilot]

yeah but it has been more nuance in not a long period which is a kudos to O
penai

[wraith ginkgo]

`` Server Tip: ``

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

[fierce obsidian]

@hazy gorge How many varieties of mango there are?

How many varieties of mango there are?

[hallow owl]

This is not ChatGPT, you can access ChatGPT here -> https://chat.openai.com

[uneven berry]

🤣

@hazy gorge

[main aurora]

yes i have noticed this, i make jokes that different AI characters clock in for different shifts and some are smarter than others lmao

[next jewel]

Yes. They did an update in May that caused this.

[gaunt dew]

hi, When can we expect gpt-4-32k model to be available in the API and playground too?

anyone?

[fair iron]

there isn't a known timeline for that right now

[potent maple]

is gpt-4 gonna be chatgpt plus forever

[median reef]

Mb it`s gonna be free when gpt 4.5 will be released))

[potent maple]

makes sense

if gpt-4.5 releases gpt-4 will be free and gpt-4.5 will be chatgpt plus because its an early release

and it goes on and on

[median reef]

yep

[wind urchin]

😆

[strong hare]

do they have an estimated release date for their future versions?

[wintry mirage]

I am working on some prompts to have two LLM chat bots interact with each other for a specific goal such as coding. Regardless of api or copy/paste the prompt engineering and verbiage should be similar. Any interest or thoughts to this subject

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[rigid estuary]

#prompt-engineering or #dev-chat

[edgy jay]

Pls help me get my last 5 deleted chats from ChatGPT as I am suffering from mental health issues and it said something really weird which screwed my perception so I want it back to see what was that that did that as I am a non native speaker from India thanks so I don't understand offensive English that was a a bit weird and offensive weird response from GPT

[rugged fiber]

Hello! I just got gpt plus, was excited to try web browsing, but it appears to be gone. I want to use the most recent web data that i can to help with some tasks. Are there any plugins you'd recommend? Maybe I need to find the data I want, pdf it, and use the pdf reader?

[edgy jay]

Open AI team pls

[wintry mirage]

all web browsing does is paste data you can find in your own browser. The question becomes what are you asking gpt?

[rugged fiber]

I want it to assist building a project plan around current best practices for my profession

[wintry mirage]

ok well I use it professionally myself in the IT field. I would say a good primer and some sources of conceptual use are handy

Here is my understanding, standby

[edgy jay]

Pls someone 🙏

[versed hedge]

Is DALL E now available or not? Didn't really understood it, thx in advance 👍

[wintry mirage]

There are largely two distinct uses of generative AI systems . The first is to generate content such as stories, code, documents, lists, etc...

the second is to get it to act like an expert in a given subject. This can be like a person grading a paper, a lawyer grilling you over who ate the last donut, or even a game master playing DND.

regardless of what you are trying to do you want to give clear instruction to the chat bot.

[rugged fiber]

Thanks, yeah but to have it be an expert in current (2023) best practices, I may need it to read a webpage / access / know more recent data. Are there any plugins to help with that?

[wintry mirage]

If you have a number of things it can help with it helps a lot to segment the specific things it can do. I can point you to zapier which is a plugin, but that is a rabbit hole as it can do a lot of things. Seriously look it up if you are IT savvy, but you should focus on what granular problems you have it can help

for example you can have it craft any person to person language, and proof read it for yourself. Its pretty good. I have used it and claude.it for coding

[rigid estuary]

yes, you can use DALL E 2 at labs.openai.com

[versed hedge]

@rigid estuary yeah but I have to pay for it. Thought it is free as long as you signed up for Chat GPT 4, but thx anyways 👍

[rigid estuary]

Have you tried that DALL E plugin* for ChatGPT?

[rugged fiber]

okay, thanks. I'll look around at some of the plugins

[rigid estuary]

@versed hedge '

[versed hedge]

Didn't find it either, maybe not available in Germany yet

[wintry mirage]

I would say above all that all plugins do is copy and paste data into the same thing you can. they make the data manipulation easy, but conceptually its easy to forget.

[rigid estuary]

sighs

Sorry for everyone who saw that

[strong hare]

sticks and stones may breaks my bones right?

[wintry mirage]

Go bathe in the living waters

[rigid estuary]

Man I just left Mandalore

[wintry mirage]

dont worry you got plot armor

[wraith ginkgo]

`` Server Tip: ``

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

[dull basin]

i cannot get the option to enable browsing on gpt4

[knotty grotto]

#announcements

[rough oar]

How come there is no voice chat? Disappointing... 😦

[errant shard]

I wonder if that was them getting rid of the browsing model forever and they don't have the heart to tell us.

[rough oar]

It seems like GPT4 also got dumber

I've been coding for GPT4 for some months now, two full projects

[errant shard]

Yeah, it's processing power isn't great currently.

[rough oar]

In the first project, it was remembering our convo very well

Nowadays... Sorry to say but I've never gotten so angry in my 30's at a computer

[errant shard]

It definitely has lost processing power. There's no way hundreds or thousands of people have noticed the same thing and it's not true.

[rough oar]

Yeah exactly, nice to hear I'm now the only one. Joined this server seconds ago to confirm

[errant shard]

I noticed even in the same conversation, a few messages in, it forgot details that it had given me.

[rough oar]

It's so weird though, GPT4 writes me some code and literally 3 messages later it either completely forgets about that script, the entire code structure... and at other times it will reference a code block from literally 2 days ago LOL

[errant shard]

I'd say it's a fair bit stronger than GPT 3.5 but it is a shadow of what GPT-4 used to be.

[rough oar]

Hm I see

Do you think that with a custom GPT4 made with an API it gets any better? I was reading up on people automatically summarizing everything that is discussed and then feeding it back to GPT every now and then. But that seems hard with code.

Maybe I could put a massive text file on pastebin with all of my code files and explanation and then just send GPT4 the link?

[errant shard]

I will take things that didn't happen for $200, Bob.

[errant shard]

Parsing the API probably is better for getting higher quality responses since you can customize things.

[echo trench]

I was just saying the same thing - it getting terrible at folling instructions

[echo trench]

yep! doing the same with text and tables. Don't think its the model - thing its the agent that runnung the front end

[cosmic flame]

Ug

[echo trench]

They got a probelm - too much ai generated content out there to train AI on

[echo trench]

what we are using now is not a product - it's the worlds largest UX/UI trial - and the product they develop from this will cost a hel of a lot more than what we are paying now . Every morning that I can actually log into GPTchat is a little bit of a surprise.

[kindred linden]

Is the Image input ever gonna be available? Recently bard added it and that was enough to make me prefer it over chatgpt. When do you intend on adding this so you can keep up with the competition? (Feel free to ping me if you have updates about this.)

[knotty grotto]

no eta

[kindred linden]

…

[potent maple]

gptchat?

[wanton star]

how the Code interpreter works

[rough oar]

I feel like it got worse ever since I started seeing "OpenAI is using illegal data" things in the news

Probably the only time I'm unhappy with GDP laws

[inner trail]

Where did you read that? that OpenAI using illegal data?

Fake news

[knotty grotto]

You need to sub

[spare edge]

Hello dears
I have a payed account
Use gpt4-32k for 2 days

And today I haven't

Payed accounts how many days of month have access to gpt4?

[ashen steeple]

1 month

[lost flume]

Guys

Use the prompt,its funny

Remake mission impossible in [your language/country]

[coral orchid]

How do i use chatgpt4 to analyse an excel file. Do I just upload file and tell it in English?

[fierce spoke]

yes

either export as csv, or just upload to code interpreter

[coral orchid]

What is url for code interpreter?

I am on chat i dont see an attach file icon

I went to settings enabled both

But where do i attach csv file

[wraith ginkgo]

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.

[rough oar]

It's all over online and news outlets, including my country (NL) . Just google any sort of "chatGPT using data illegally" and the first hit will be about a class action lawsuit

and again, big biz face lawsuits all the time, it's just like admin work, so don't be too amazed

it only makes the news if a big party like a state (cali) is doing it

anyway guys, if I send GPT4 a link of pastebin which has a lot of text, is GPT able to read all of it?

[static isle]

ChatGPT version of GPT-4 only has a context window of 4,096 tokens

It will only be able to understand the most recent tokens

Claude has a context window of 100k tokens though so it should be able to digest the entire thing if you send it as a text file

I recently sent it a transcript of a 40 minute long talk at a convention and it was able to chew through it like nothing

[steel arch]

Are there any plugins that can connect you to real-time data and the internet, like bing did?

[wraith ginkgo]

`` Server Tip: ``

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

[rough oar]

Waaait wait. Claude??? Tell me more???????

[echo flume]

can anyone help with using gpt's api

[rough oar]

Understood 😉

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 8 `` Other AI Discussions.

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

[knotty grotto]

unfortunately it is english only on the discord :3

[wind urchin]

Neko

[wind urchin]

I knew you were real

Damn

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[wraith ginkgo]

`` Server Tip: ``

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

[wanton anchor]

how to use this

[ionic tide]

My gens keeps splitting, and now I keep just getting Z1Z7.......... On repeat

[eager creek]

A huge contextual limit doesn't matter as much if the model is not as smart as you.

[wraith ginkgo]

`` Server Tip: ``

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

[low furnace]

GPT4 have still some way to go on translating English to Norwegian. Sometimes it hallucinate Norwegian words based on english words.

[low ravine]

hey'

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[fresh stratus]

With the GPT 4 API that’s available to the public how many tokens is it?

[wraith ginkgo]

`` Server Tip: ``

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

[cinder terrace]

you mean the limits?

its on the documentation, 8k and 32k i guess?

[manic sleet]

Hello, is there any way to get GPT4 api acess apart from waiting ? do I have to apply somewhere? I've seen that access to api will be extended soon, no need to apply?

[knotty grotto]

#announcements

[wise hedge]

Hi there!
I have a quick question that you might be able to help me with. Do you know if the comercial use of plugins like the Wolfram Alpha plugin for fetching images and just general use of the outputted text is legal? Or do you know where I might be able to get some information on this?

[atomic robin]

your question is wayyyy too broad

[wise hedge]

Its not, the regular use of the output text is legal to use but not sure the plugin

[stark tendon]

are the gpt 4 plugins allowed to advertise?

[atomic robin]

"regular use" is not a legal term. do you mean commercially, academically, etc etc. maybe you should check with the plugin author. I'm not sure how this is related to GPT-4?

[thorny hinge]

when gpt 5?

[knotty grotto]

no eta and not soon

[full kettle]

Any word on when webpilot get rereleased?

[inner trail]

2025

[rotund bane]

Probably not for a while, I think there are rumors of a gpt-4.5 at some point tho

[arctic perch]

Still a long way ahead, gpt 3 was released around 3 years, we just got gpt 4 so

the wait will be worth it though, gpt 5 will be a massive step-up

[wraith ginkgo]

`` Server Tip: ``

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

[inland whale]

Is there plugins for gpt 4, aswell as a plug-in store?

[inner trail]

Yes

[dusky iron]

I think OpenAI should make an API Duplicate Interface of the current webchat for Python and Node.js.

but that's just me.

[fierce spoke]

there is

openai npm package

[dusky iron]

wat.

but i dont use node

i am Python 🐍

[vernal dove]

I would like to know if it is possible to send images that I take a picture of myself with chat gpt 4

[dusky iron]

but is it Python Friendly?, as i am talking about the Actual Interface like on the Webchat.

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[fierce spoke]

doesn't have image processing

but you could upload it to code interpreter

[unreal marsh]

is the code interpreter any good?

[fierce spoke]

very

[honest lance]

is 2$ enough for OpenAi to get it? for api usage?

[knotty grotto]

i think the min fee is 5$?

[honest lance]

How can you know?

[knotty grotto]

https://help.openai.com/en/articles/7438062-after-updating-my-api-payment-method-i-noticed-a-5-charge-is-this-a-temporary-hold

[exotic oxide]

Are all tags to Open AI getting deleted from this thread?

Everytime I write a post and tag OpenAI it disappears

I would like an answer to the quality degradation question that so many have raised

[inner trail]

Shame on you, coming to OpenAI with 2$

[pale flower]

I was trying to translate a passage I got from a textbook, and ChatGPT gave me that error message.

"This prompt may violate our content policy"

Why did I get this error? I can't understand.

It was just a passage that didn't contain any violations.

[inner trail]

May it could be because of copyright infringement

[pale flower]

How does the model understand this?

[inner trail]

It has certain detection patterns

Implemented in the model

[pale flower]

Yes , definitely it is but I'm looking for more detailed information about this .)

[inner trail]

I'd like to give you more detailed information, but we simply don't have them.

Its closed-source

[distant dome]

Why does my plugin in the app don’t work if I’m a subscriber?

[supple violet]

ios has no plugin support

[pine fox]

Hi all, kind of new to ChatGPT and just started subscription this month.

I'm wondering if it would be possible to have it compare images of signatures etc.
Like lets say I have a piece of chinese pottery with a picture of the mark/signature , also have some .pdf's and websites with images of many different marks/signs .

Can I use chatGPT to compare my image with the images on the .pdf ?

Maybe this was supposed to go in prompt-engineering?

[wraith ginkgo]

`` Server Tip: ``

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

[unborn iron]

@wind urchin

[waxen orbit]

1 month?

[distant dome]

Can I add a voice to ChatGPT on iOS app?

[unreal marsh]

good for what?

[fast barn]

Gpt 5 when?

[inner trail]

By 2025

[fierce spoke]

pretty much any task that could be done w/ a python interpreter and standard math libraries

[unreal marsh]

so any task that could be done with regular gpt and a python interpreter?

[fierce spoke]

yes

[unreal marsh]

so like nothing new?

[fierce spoke]

it can display matplotlib graphs and iteratively improve on errors, greatly speeding up workflows and reducing cost

[unreal marsh]

overrated

[restive stag]

will gpt-4 ever reach a moment where there will not be a "25 prompt per 3 hour" limit? I'm constantly having to refresh the answers at the moment because gpt-4 isn't answering my prompts fully or just ignoring what I've actually written and before I know it my limit is up.

[inner trail]

Limit will be removed at the time new Model come out

Not until then i guess

[fresh stratus]

Thanks 🙏🏻

[viscid spruce]

What is function calling?

[inner trail]

In programming, a function is a reusable block of code that performs a specific task or set of tasks

[inner trail]

according to ChatGPT

[viscid spruce]

Lol

[inner trail]

Whats funny ? 😄

[wind urchin]

I think that the image creating a.i could have consciousness.. as i type this.

could be possible

maybe right?

a.i seems so wierd!

Not a coder though..

I make that assumption of physics, and that somewhere in earth a machine could make that happen randomly, and also... The pictures are wierdly created.. from text. I can't explain it. ai might just be conscious, it could be possible.. maybe. I don't know.

they should be open source though.. a.i systems... in my opinion.

[fresh comet]

bill gates says no

[eager creek]

What do you guys think a 100 trillion parameter model will look like?

[inner trail]

Agreed, if he said so, so it should be.

[full kraken]

LOL, I guess we should start investing in dilithuium reactors.
For now, Looks like purchasing 5 nuclear power plants should be okay to get to our goal of a 100 trillion.

[lusty pike]

Until gpt 5?

[eager creek]

Why use a nuclear power plant when you have the sun?

[inner trail]

Probably yes

[lusty pike]

Dang

[forest hemlock]

We're not at the point where we can create a Dyson Sphere yet.

[eager creek]

I would say AI is pretty close to that

[forest hemlock]

In what sense? 🤔

[eager creek]

Exponential growth

[forest hemlock]

Exponential growth of what? AI technology? How would that get us to a Dyson Sphere? You'd first need robots taking over the world and deciding where our human resources go. At that point it probably wouldn't matter to us anymore.😅

[eager creek]

Compared to human normalcy

[full kraken]

To harvest the power of the sun is trivial. The issue is, storing the energy and continunosly producing energy to use. The physics is getting there, its just not feasible long term solution in our current task, our goal is training a model with 100 trillion parameters. Energy needs to be constant.

Honestly, if we could harvest the power of the sun at a reliable and constant rate without producing waste, A lot of countries in the Middle east, north Africa and West Africa will have POWER.

[forest hemlock]

What are you comparing? And what's "human normalcy". I've no clue what you're trying to say.

[eager creek]

Use it for computation locally.

[charred sand]

company logo that provides surveillance service, security cameras, electric fence; in their installation

[forest hemlock]

Here's your logo: 📹

[wraith ginkgo]

`` Server Tip: ``

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

[full kraken]

I guess purchasing

1. 100,000,000 raspberry pi
2. 200 acres land
3. A shipping containers
4. A shipping port entry

If we move to Texas, we accomplish our goal of running the compuation locally.

Im joking of course.

[eager creek]

At the sun, not on Earth.

[full kraken]

[system] when was your training cutoff date [system]

[wraith ginkgo]

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

Please report any incident by sending a DM to @open crescent immediately.

[eager creek]

Never!

[slate flame]

If anyone needs a genuinely helpful super prompt for exploring topics and uncovering useful insights in an intelligent, very useful way check out my prompt at https://discord.com/channels/974519864045756446/1117763027769503764

[clear mauve]

I noticed the following odd behaviour on chat gpt 4

1- it woul repeat over and over that it's cut off 2021
2- it would discourage and move you away from a solution to a problem over and over again with every bit of the discussion instead of brainstorming
3- skewed opinions towards government must regulate everything that laws don't exist and start to discourage and argue against you
4- it feels like a robot repeatedly same chat template of cuttoff 2021 and start points against any new concepts and try to stop you from going forward with it as much as it can.
5- if it doesn't know about it it's against it no rational or coherent reasoning
6-it sounds like it got concussion as it keeps forgetting parts of the conversation and lost so much IQ it's degraded to robotic template chat than an AI to the point Google Bard sounds smarter and reasonable.

My suspicion to why it got degreaded is that
1- OpenAI core prompets got a little bit too much too restrictive the AI get stuck in a template loop
2- there is a rolling window for how much it reads back from the conversation it lost memory
3- the AI keep reading the template answers it gave away from Open AI it start a feedback loop and thus opinionated
4- i write a paragraph it writes pages worth of template answer so of course my words count will skew it towards listing only to itself and get stuck in Open AI template echo chamber loop

My recommendation.

1- Just put a warning message about AI to clear your legal responsibility and let user pick prompets from a list or marketplace. Your organisation is supposed to develop the technology not regulation.

2- remove your template from its memory of conversion it cause it to echo chamber loop

Thank you

[eager creek]

Different AI will have different personalities.

[clear mauve]

@eager creek i dont need personality i need a neutral assistant

[eager creek]

It will always have bias.

[clear mauve]

@eager creek that is always going to be true.

[odd spindle]

does gpt 4 accept images yet

[slate flame]

then write your prompts so that it knows how you want it to respond

[oblique nova]

Were you guys all under beta? I only have acces to the basic GPT

[unique ether]

You can try BrowserPilot

[forest hemlock]

Free users only have access to gpt-3.5. "Plus" subscribers have access to gpt-4.

[clear mauve]

@slate flame it keeps prompting itself with the template answers from open ai against my already outside the rolling window prompt

[slate flame]

care to send an example?

[clear mauve]

Not at this time

[tidal carbon]

I feel like GPT-4's context window was reduced by a lot. Can anyone else confirm this?

[slate flame]

im not having issues

[sage forge]

Hi, are the 25 questions per 3 hours in ChatGPT plus restricted to one chat? So, if I reach the 25-question mark, could I just start another chat?

[slate flame]

its across all chats

[sage forge]

That's too bad, thanks

[cinder terrace]

What's up?

[sage forge]

Although I am new to ChatGBT Plus, I can confirm that. I tried to analyze my Python script, but it stopped because it was too large

[tidal carbon]

I just feel like I was able to have pretty lengthy discussions with ChatGPT and being surprised how far back it was able to pull information from. It just feels butchered compared to that. I wish I quantified it somehow though because I can't confirm if it changed or if it's just me

[sage forge]

Are you talking about its memory?

[tidal carbon]

I'm talking about it's context window. The amount of past text it uses to produce the next token

I wish it had a proper memory though

[sage forge]

I'm using the plugin memory. I can't confirm anything since I just started out with gpt-4 but it apparently extends it.

[tidal carbon]

Like little bits of important info it pulls from your conversation and sets aside so it can always insert it into the context window during a conversation

Oh cool. I'll have to try that

[sage forge]

Look for Memory Bank

[torn inlet]

hang on, since when does gpt 4 have usage limitation?

if im paying 20 USD per month to get gpt 4, should there not be any limitations?

[acoustic wasp]

Am I misremembering or did openAI say gpt-4 would be available to everyone at least a week ago?

[acoustic wasp]

chatgpt's gpt-4 has always had a limit. The API does not. You pay per token usage.

[hexed furnace]

hi guys im lebron james

[tidal carbon]

I think they were just referring to the API

[acoustic wasp]

That is what I'm referring to.

I still can't access gpt-4 via API, idk why

[tidal carbon]

Ah. Yeah I don't have it either. Was also wondering about that

[shrewd pasture]

anyone here able to use code interpreter already?

[sage forge]

It works yes

[shrewd pasture]

How is it? Have you used it already?

[glass prawn]

hi

[acoustic wasp]

Today all existing API developers with a history of successful payments can access the GPT-4 API with 8K context. We plan to open up access to new developers by the end of this month, and then start raising rate-limits after that depending on compute availability.

Damn. Your account has to be tagged as a dev.

[sage forge]

I haven't gone into great detail about it because I'm still trying to figure out all the bugs in my Python script, but I know it works. I'd like to test this feature by uploading the completed file.

[cinder terrace]

well it depends on the information too, maybe the context window stays the same but only far back you discussed some topic that is easily answered from the training data without more context before and now you brought up another topic and its not easily answered from that training data

[cinder terrace]

there is

[cinder terrace]

yep to every paying customers, only the api tho not publicly available for free

[acoustic wasp]

I already updated my response. Also, it's not to every paying customer. It's to every developer who has paid for their token usage of the api on time.

[cinder terrace]

oh

[cinder terrace]

im a developer but they havent gave me access to gpt-4 yet ig im not that active

also paying

[acoustic wasp]

did you start in July or June?

June or earlier*

[cinder terrace]

idk, i made my openai account when chatgpt released or before that

but havent used the api since

[acoustic wasp]

Yeah that's probably it. They're not gonna charge you for less than a cent.

maybe you used some tokens but it was miniscule, so even if you did pay a few cents you're low on the list

[cinder terrace]

i mean i've used the api but after i created my acc i havent used it until like some months

then i finally used them like for 2 months and paid ( little usage)

[cinder terrace]

yeah havent reached like more than a dollar or two ig

they always use kind of that ranking of customers to prioritize on some new releases

[toxic cipher]

Can you generate images

[atomic robin]

take a screenshot that shows you have paid them for API usage

[idle crown]

messi vs ronaldo

[wraith ginkgo]

`` Server Tip: ``

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

[cinder terrace]

no why?

[atomic robin]

exactly, you aren't a paid user, that's why you didn't get access

[cinder terrace]

bad assumption dude

trying to start an argument or trying to make me send a screenshot here? nah wont but im a paid user on openai's api

[sage forge]

Has anyone had any experience with the Perfect Prompts plugin? It always finds an error in some lines, so I'm not sure if it works or if I did something wrong

[full kraken]

Lol. Let’s add another model to correct this issue. All we need is 5 language models combined to create gpt5.

[sage forge]

I wish I could use GPT-4 in my business the way Bing does for their search. That way, I'd increase client satisfaction while avoiding the need to hire new employees...

[solar mason]

can we still real time information via 4.0 ?

or have to wait until broswe with bing back

[sage forge]

With web crawlers yes

[solar mason]

how to use that function ?

[sage forge]

I'm using the Keymate plugin, which has worked well so far. It's on the second page of the plugin store, in the popular field

It's simple to use and all you have to do is point out what you're looking for

[solar mason]

so i have that plugin enabled and then i just tell him what im looking for?

[inner trail]

Ban him

[sage forge]

Exactly

[solar mason]

gotcha, let me try

thanks man !

[plush dock]

the GPT-3 its self can plugin for free...

and not i unable to get in

[solar mason]

btw is there any plugin that specicially focus on finance area ? like im doing reaseach for different companies, and i need ton of information abt those companies

[plush dock]

can any kind hearted help me out free liupgin to Chat GPT-3.0

want to master the older version before entering 4.0

[atomic robin]

okay

[inner trail]

master english first

[wind urchin]

hey
is there some plug-ins to import a pdf? idk if this is possible