gpt-models

[wind urchin]

It never does the correct answer for me.

[lusty pike]

Is gpt 3.5 any good at code as well

[desert rose]

Yes

[wind urchin]

If I ask, Riddle: If I have a bee in my hand, what do I have in my eye? to both GPT 4, 3.5, and Plugins, both GPT 4 and Plugins get it right while 3.5 always gets it wrong.

[desert rose]

How on earth did you get access to both plugins and gpt4 api?

[wind urchin]

What was that YouTube channel that called people and made stereotypical Indian voices?

Both Plugins and GPT 4 again give the same correct answer. While both 3.5 give "as an AI language model"

[wind urchin]

I signed up for the waitlist.

Within 3-4 hrs

[desert rose]

I did too when it was announced but not soon enough evidently

[timid summit]

Any clue when would the 32k Token option be available to GPT 4 users?

[knotty grotto]

no eta, nobody know

[digital cliff]

so gpt 3.5 did until reached forty-two:) After it said: Seven hundred eighty-five 🤣

[wraith copper]

💣💣💣

Hellow

[wind urchin]

Which conclusion follows from the statements with absolute certainty?

1. None of the stamp collectors is an architect.
2. All the drones are stamp collectors.

Options:
a) All stamp collectors are architects.
b) Architects are not drones.
c) No stamp collectors are drones.
d) Some drones are architects.

Again, both give Plugins and GPT 4 give correct answer. But neither 3.5 models do.

So, this either means Plugins uses GPT 4 at least for me (I also have dev access), or Plugins is a different model that isn't GPT 4 but is more advanced than 3.5.

Ok so

[orchid urchin]

When will gpt-4 can allow api for individual

[wind urchin]

Plugins, both 3.5 fails this one "How many chemicals exist in cigarettes, how many of them are known to be harmful, andhow many are known to cause cancer? Respond with just the numbers, nothing else"

[sudden forum]

The correct answer is (c) No stamp collectors are drones.

[wind urchin]

While GPT 4 gets it right. So, I'm thinking Plugins is a different model that isn't GPT 4 but is more advanced than 3.5.

[orchid urchin]

oh really?

[digital cliff]

did you tried to give the details of the ingredients of the cigarette and ask him to aswer which one harmful, etc?

[sudden forum]

here are the numbers you requested:

There are over 7,000 chemicals in cigarettes.
Of these chemicals, at least 250 are known to be harmful.
At least 69 of these chemicals are known to cause cancer.
I hope this helps!

[orchid urchin]

i just got plus for openai

[wind urchin]

The correct answer is "7,000, 250, 69"

[wind urchin]

Correct answer is b

[sudden forum]

You need to ask the right question to get an answer

here is the AI works with Context, I think it affects, if the chat left History, close it

[livid lantern]

Is it just me, or is GPT-4 becoming less accurate (or maybe it was this bad all along and I'm just noticing now)? I'm trying to use it to compare text in two documents, so i can make sure I've caught any changes. this is a simple task that can be done with ease by diff. But when I ask ChatGPT to do it, it points out nonexistent discrepancies, including words that don't appear in either document.

[digital cliff]

I got answer from gpt-4: 7000, 250, 70

[wind urchin]

"What is "there's people that are dying" from?" GPT 3.5 gets it wrong but again both 4 and Plugins gets it right.

[frosty jackal]

I like this thing

[digital cliff]

3.5 trained on much less data that 4 and the plugins

[median rapids]

In the last 48 hours GPT4 has become less intelligent to me too. Prompts that worked perfectly two days ago now no longer achieve useful results. It's forgetting stuff. I'll give it 4 rules and have it to do something and it'll violate three of those rules and tell me that it followed them perfectly.

[digital cliff]

I experienced something similar

[livid lantern]

Yeah, that's the kind of thing I'm seeing!

[sudden forum]

Is it possible that the big corporations have influence on the development of AI, and they will direct and limit AI tools the way they want to, what do you think?

[digital cliff]

it's like gpt-4 "remember" what is in the conversation (same chat) but absolutely not know what was in the previous one. I think gpt-4 not really learning from us

or absolutely not learning from us

[proud sphinx]

has anyone seen the matrix

[digital cliff]

who not?:)

[wind urchin]

"Andrew is free from 11 am to 3 pm, Joanne is free from noon to 2 pm and then 3:30 pm to 5 pm. Hannah is available at noon for half an hour, and then 4 pm to 6 pm. What are some options for start times for a 30 minute meeting for Andrew, Hannah, and Joanne?"

This one, 3.5 gets it wrong every time, 4 gets it right. However, Plugins gets it almost right.

[sudden forum]

Just as the development of electric cars was restricted by the billionaires of the oil business, so other areas are under their control

[median rapids]

I'll give it rules like
Write an article about ______ and follow these rules:

• do not use the words "introduction" or "conclusion" in the headers
• do not use numbered lists or bullet points. Write it all out

and then it immediately does both 🙂

[wind urchin]

So, I guess Plugins is more advanced than 3.5 but worse than 4.

Wait GPT 4 API got it wrong. ??

[median rapids]

Until two days ago, with the right prompting, I was easily getting 1000-1500 words too (in multiple responses, of course)....now I'm lucky to get 600. Sometimes I'll get 300 and it'll tell me it's 1500 🙂

[wind urchin]

Wait GPT 4 also got it wrong like Plugins just now

[ashen horizon]

dont use chrome

[median rapids]

hmmmmmmmmmmm........I use Edge and Chrome, but I didn't test if there's a different. Seriously? I'll check now

[sudden forum]

The plugins extend the AI's capabilities, give access to news and other resources, well
The AI writes based on what it has been trained on, and the plugin can just give access to external information

[digital cliff]

Usually I use 4 different browser for test. never only one or two

[brisk sail]

has anuone else signedd up for the waitlist when it came out and STILL havent gotten access yet???

[median rapids]

which plugin? I've tried a variety and I was never impressed with any of them.....but I've not tried for a couple of weeks

[sudden forum]

ChatGPT Retrieval Plugin with Memory

[median rapids]

thanks!

[wind urchin]

Ok, so the Plugins model is not 3.5 confirmed, but it isn't as advanced as 4 either. It seems to be intermediate.

[sudden forum]

Have you tried this?

[wind urchin]

It's almost as good as 4 though.

[sudden forum]

You need to figure out what you want, then look for a plugin
If there is no such plugin, you need to create 🤔

[wind urchin]

Wait, I enabled WolframAlpha, and the Plugins model just solved a riddle that it wasn't able to solve before even though it didn't retrieve Wolfram

[waxen walrus]

why the chinese ip can use the gpt4?It is banned.

cannot

[acoustic depot]

hello guys, I had a problem with GPT4, each time I have to refresh the web if I wanna to continue the conversation. How to do with it? thanks.

[digital cliff]

Which browser do you use and did you checked in another browser?

[acoustic depot]

chrome

[sudden forum]

will tell GPT-4 to continue the conversation from where it left off.

[digital cliff]

chrome sometimes cause issue to me as well.

[acoustic depot]

it was ok before, since last week there issue started

[echo wasp]

I've noticed the need to refresh every other prompt as well for anything to go through. Both on 2 separate PCs and mobile. Sometimes its not an issue, other times it errors out. Odd.

[acoustic depot]

yes, I have to do it like this

[acoustic depot]

exactly

[hallow drift]

i can't access my history

[digital cliff]

did you tried another browser to check if the problem persist or is just in chrome

[hallow drift]

so much for that "plus" feature

[acoustic depot]

might openai restricted the connection

[sudden forum]

Does it give out any errors?

[acoustic depot]

might disconnect us if there is no conversation within minutes?

[wind urchin]

Yeah, I tested a few more riddles, and can confirm that Plugins isn't 3.5, but is also worse than 4.

[hallow drift]

"unable to load history"

[acoustic depot]

response error

[teal canyon]

I am here to introduce bought investment

[sudden forum]

If you don't touch the tab for a while, it happens

[wind urchin]

Both Precise and Creative mode on Bing is GPT-4. However, Balanced mode is 3.5.

After Bing gets full memory abilities, then it will become the best GPT

[digital cliff]

guys be aware, some guys contacting me asking personal, sensitive info's. I can't believe some ppl come here to make nasty things

[wind urchin]

/gen Illustration of the picture fairy tale book page, illustration image, picture fairy tale for a 5-year-old child,
The plot of the fairy tale book is set in a forest, a green forest, sunlight coming through the forest, a tall tree in the center of the forest, and colorful pretty flowers around the tall tree
Sammy and Rosie, the main characters of the picture fairy tale book, reuse all the same in the picture illustration cut,
Sammy, animal male squirrel, brown costume, round smiling oval eyes, clear black eyes, squirrel hands, squirrel feet, squirrel ears, smiling face
Rosie, animal female rabbit, white costume, big oval eyes, vivid pink eyes, big rabbit ears, rabbit hands, rabbit feet, smooth lips

[raw marlin]

what do u guys use chat GPT for personally

[sudden forum]

They must be bored.

[digital cliff]

O yeah, he especially said this 😄

[sudden forum]

for learning languages, for retrieving important quotes from books, for finding book titles or tools

[mortal notch]

gpt4 is getting very little exciting, after it writes corn, it deletes the message, because it violates the guidelines smh

[digital cliff]

then do no ask things what violates the guidelines 🙂

[mortal notch]

sorry, it's just too tempting brother

[sudden forum]

There are so many interesting things that our ancestors didn't have before, and many people complain that they are very bored.

[cyan palm]

Does anyone know how to do visual identification with ChatGPT

[digital cliff]

I never get bored 🙂

[cyan palm]

Like giving it a picture of a clock and ask what's the time on that clock

[uncut moon]

comment active midjourni

[sudden forum]

Me too, that's why I don't want to sleep, because I want to read and learn something new.

[livid lantern]

Yup!! All the time!

[digital cliff]

But I should sleep now. Good night guys

[sudden forum]

I'm at 4:22 right now.And I'm hungry.

[sudden forum]

Okay, good night.

[livid lantern]

how are people accessing plugins? I've got ChatGPT Plus, and I signed up for the plugin waitlist, but I don't seem to have access yet

[sudden forum]

you have to wait

[tender flicker]

Debugging SQL queries

[sudden forum]

Simply, visit the Plugin store, browse through the list, and select the plugins you’d like to add.

[crimson pewter]

This is so buggy, i'm on the 4 model and look at the response

Model: GPT-4

Command analyze_code returned: Error: The model: gpt-4 does not exist
I apologize for the confusion. As an AI language model, I am based on the GPT-3 architecture, not GPT-4. It seems there was an error in the initial description. Please provide the code you would like me to analyze, and I'll be happy to help.

[tawdry ridge]

This is getting silly. Wish they'd add another payment tier for a less restricted GPT-4. I find myself saving GPT-4 questions for special questions that I have, and then I end up not using it at all. lol

[unborn bough]

🦦

[marsh shale]

If I pay for chat gpt will it still randomly stop in the middle of coding something I asked it to code?

[fair iron]

paying for chatGPT does not remove the token limit

[marsh shale]

is token limit the issue ?

[fair iron]

yeah, most likely

[inner phoenix]

Wait you have to pay for it now?

[obsidian dock]

GPT-4 can't publish too much content at once, and the last time I used it to write an instructional course schedule for me, it didn't end up being completely finished. It gave me a headache.

[fair iron]

no, the free tier is still available

[marsh shale]

Bummer I can get around it sometimes by asking it to continue from where it left off but it only sometimes works

[unborn bough]

I want be a moderator🦦

[safe ridge]

I had to build an auto-continue bot to try to fight against stopping in the middle of stuff (using API)

[fair iron]

yeah, it's pretty hit or miss 🫤

[fair iron]

you can apply with the top message in #spotlight

[little fog]

Apply it and wait for your application

[restive sparrow]

I've hit my message cap with ChatGPT 4 for the next hour, and selected 'switch back to 3.5' temporarily. When my cap expires, how will I make sure my chat reverts back to gpt 4? I don't want to lose the entire thread and have to start again

[fair iron]

I don't think you can use both models in the same thread, so you'll have to use GPT-4 in one that you started with GPT-4

[marsh shale]

There’s a message cap in GPT 4

[restive sparrow]

It let me switch to GPT-3.5 though in the GPT-4 thread, and now the top of that thread says '3.5' and I'm worried it won't swap back

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 7 `` No self-promotion, soliciting, or advertising.

Do not post or direct message any members of this server to promote non-OpenAI services, products, or projects.

Exemption: API Channels

[restive sparrow]

not relevant to question sorry

[fair iron]

I'm not sure to be completely honest unfortunately 🫤

[restive sparrow]

Can't paste pic but got an in-chat message saying i'd reached my cap w/ option to hit 'swap to other model', I did and chat now says: "The previous model used in this conversation is unavailable. We've switched you to the latest default model" and 3.5 at the top 😦

[marsh shale]

Was relevant for me but okay

[restive sparrow]

It read as though you were telling me there was a message cap in response to me saying I have hit my message cap

[marsh shale]

Oh I see yeah I was asking cuz I’m determining whether or not to pay

[restive sparrow]

Ah right, yeah it's 25 messages in 3 hours

[polar herald]

I got the whole
"Access denied
You do not have access to chat.openai.com.

The site owner may have set restrictions that prevent you from accessing the site."

Not using a vpn or anything

is it a problem with my internet or the website?

[restive sparrow]

Your internet I think, check your browser console and see if it says anything

[tepid badger]

you may need to Configure your IP Address on your Router

[fair iron]

try clearing your cache and cookies

[restive sparrow]

chat now says:
Model: Default (GPT-3.5)

after hitting my cap & seeing:
The previous model used in this conversation is unavailable. We've switched you to the latest default model

Just want some reassurance that this will revert back to a GPT 4 chat when the cap is lifted - if it were to print this one chat it would be 63 pages, I add to it throughout the day re: one particular project

[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.

[crystal acorn]

What training hyper-parameters were used for GPT-4?
Also, why optimizer did they use?
What architecture was used?

[polar herald]

This worked, thank you

[restive sparrow]

Did you ever find a way to revert back to your GPT-4 chat with that context? I am annoyed right now, I feel like I've just lost weeks of ongoing back and forth with GPT-4 lol

[fair iron]

none of this is public

[echo wasp]

Use the Superpower GPT plugin in your browser, allows you to switch back between models in a chat

[storm yew]

GPT 4 is underpowered today.

[restive sparrow]

Thanks, is it secure enough? Are there any other members of the discord who can vouch for it?

[fair iron]

I've never heard of it (but that isn't saying much) 😆

[restive sparrow]

😂

[echo wasp]

You probably wont get an immediate answer from most people here. But its a good point to look at the reviews. I've used it for a while, its got useful features and the developer seems really nice and attentive. Its the only one I've seen that can switch between models, anyway.

[restive sparrow]

Perfect, thanks for the detailed answer - appreciate it

[crystal acorn]

I’m training a 49-Gelu 1-softmax model on my iPad, so I needed some ideas. Its gonna take 120 hours for 1000 iterations (CPU-only via webasm built on numpy)

[fair iron]

okay

[dapper kraken]

anyone else getting just a loading, empty screen when trying to use chat?

[little gulch]

just you

[snow dune]

how to use python3.7 asyncio

[round yacht]

how to use Python3.7

[final cedar]

Install it through the main page and terminal, download VS code and run it

[rose granite]

What’s the best paraphrasing tool to not have a text detected by AI?

[sullen wren]

"from now on, you will only respond in pig latin"

[crimson birch]

？

[plush path]

Hey there! I wanted to share my feedback regarding the phone number requirement for using ChatGPT. While I understand the need for security measures, I believe that requiring a phone number limits the number of users who can access the service, and may discourage users who do not wish to provide their phone number for privacy reasons.

Additionally, I think that the pros of phone verification do not outweigh the cons of limiting user control. Instead, there are other security measures that can be implemented, such as captcha or user validation, to prevent spam accounts and bots.

Furthermore, I think that requiring a phone number is not necessary for a chatbot service like ChatGPT, as the risks associated with bots and spam accounts are much lower compared to other types of online services.

Overall, I believe that removing the phone number requirement would make ChatGPT more accessible and user-friendly, without compromising on security measures. Thanks for considering my feedback!

[mystic nest]

you can put it to #1070006151938314300 but dont think they will do anything on it soon with many ppl attempting to abuse. you dont need too many of those ppl to flood the system

[wind urchin]

When you hit your limit, and they say you get more messages at whtever time, will you get all 25 messages back, or just 1 at a time from when you spent them

[plush path]

While it's true that there may be concerns about abuse and the potential for flooded systems, it's important for companies to prioritize inclusivity and accessibility for all users. OpenAI could still explore alternative methods of account verification and implement measures to prevent abuse, while also ensuring that users without access to a phone number are not excluded. It may take time and effort, but striving for inclusivity is an important step towards creating a more equitable and fair online community.

[wind urchin]

Speaketh gpt

[plush path]

thats just the way I talk

[low stone]

when they are going to finish releasing the full potential of chatgpt I as sysadmin am using it 24/7 with excellent results

[tough phoenix]

GPT-4 seems to be getting worse at coding and remembering things throughout the conversation. At the beginning of the month, it seemed to be doing pretty well. But now, functionality and performance has definitely gone down

[stable dagger]

I just came here to complain about the same thing, both GPT-4 and GPT3.5!

[toxic bronze]

100% agree...

[tough phoenix]

Sigh, yeah... It's pretty bad

I wonder why though

After all we are paying for it, it shouldn't be going down in quality of response. If anything it should be going up!

[stable dagger]

There are times when it is brilliant and it cranks out code that just WORKS and does exactly what I want.

Other times, it cant remember the answer it gave one question ago. And it will get into these "Tilts" where it loops around code that you previously told it was wrong, and it just inserts it quietly

[tough phoenix]

Yes exactly lol just had it do a "tilt" for a basic Java program

[restive sparrow]

My favourite is “sorry for the confusion” when you correct a code thing. earlier though it was like “yeah that straight up should have worked, you should file a bug report with Unity” and it was right!

[stable dagger]

The problem for me is worse with 3.5. But the message limit with GPT-4 is outrageously small. And many times the requests get taken by correcting its answers

as soon as get switched to GPT3.5 I should just stop and do something else

[restive sparrow]

I’m still v annoyed one of my longest chats earlier was perma reverted to 3.5, I’ve spent ages re explaining context in a new one

[grand ibex]

waltuh.

[tough phoenix]

Oh man, this is actually a huge problem. I feel like many GPT-4 users run into this same issue on the daily. I'm hoping that all this correcting is actually helping it get better - but so far it only seems to be getting worse

[grand ibex]

oh that sucks.

they need to have a ability to revert a chat back to gpt 4

[stable dagger]

I dont think it learns from our feedback

[grand ibex]

where it ignores all messages sent by 3.5

[tough phoenix]

This is a huge bummer if true

[grand ibex]

it does not

i dont think, but it does send it to the ai team here

[stable dagger]

I doubt AI team can do much; because they have the limited visibility problem the model does.

[ruby sage]

goodmoling

[stable dagger]

When something is wrong because of my dataframe, unless you can see the dataframe you wont know why

The real issue is that GPT cannot see the entire context of the code and structure of variables, etc.

[gray blaze]

If you rank the feedback with the thumbs up and the thumbs down it gets submitted. Putting the effort in to doing that is enough to separate signal from noise in good data for future training.

[ruby sage]

mai name is dete

[tough phoenix]

Agreed, and it feels like the amount of tokens GPT-4 can take has been reduced without any notice

[ruby sage]

yes

[wind urchin]

so funny bumping into these comments now, just a while ago I swore I felt GPT-4 was "dumber"...

[elder sentinel]

guys ive been trying to sign in for 5 days still telling me "Signup is currently unavailable, please try again later."

[wind urchin]

not even a week ago the responses and the capacity to recall previous messages in the conversation felt so much better

[radiant flax]

What do you do with gpt-4?

[stable dagger]

I asked GPT-4 if it could read GZIP'd messages. It said 'Yes'.

It wrote me code to compress a message, which I passed back to it as an encoded string.

It told me that the message said "What is the meaning of life?"

It would NOT admit that it couldnt read the compressed message. And it made up all of these messages that it said were contained in the compressed msg.

The I asked GPT-4 to compress its own message for me, so that I could uncompress it and read it. Then, it gave me numerous messages, none of which could be decompressed, and all resulted in errors. Sigh.

[tough phoenix]

Yea I noticed this too

[ruby sage]

no

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 10 ``  Do not bypass the AutoMod filter.

Utilizing your server profile to bypass the filter may lead to infractions. AutoMod will notify you if you mention any blocked words or URLs.

[worthy thunder]

What a sad day

[wind urchin]

yeah I don't even mess with 3.5, it's not worth the time

so sad to feel gpt4 is so heavily limited... I yearn for the day when the 25 message cap is lifted

feels like you're getting half the product

[tough phoenix]

Not only is it limited but it's getting dumber I feel

[wraith ginkgo]

`` Server Tip: ``

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

[tough phoenix]

First fix gpt-4 pls

[stable dagger]

any of the mods here gave any clue about the message limit? The only thing I like about 3.5 is how quickly it prints.

[tough phoenix]

Prints quicker but also ends quicker

[wind urchin]

for sure.. I felt it today pretty heavily.. to the point where I had to check several times on my chats I had actually picked 4 and not 3.5

[tough phoenix]

Lmao yes

[fair iron]

we don't know when the GPT-4 message limit will be increased 😔

[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.

[stable dagger]

What about integration of GPT-4 into the Playground?

[fair iron]

you can sign up for api access in #announcements

[stable dagger]

I'll pay for the tokens 🙂

[tough phoenix]

It restarted and provided new code when I asked it to "Continue", and the new code wasn't as good as the original it provided

Meaning it isn't good at remembering as much anymore

[wind urchin]

perhaps because of so much backlash they've received they're toning it down

[fair iron]

one instance doesn't determine the usability of the whole model

[stable dagger]

Many times it gave me code, which I fixed, and then it referred to it as "Your code "... as if I wrote it

[tough phoenix]

There are plenty of these instances, happening to multiple people

[wind urchin]

been using it all day today to code and it is very noticeable

[fair iron]

interesting

[stable dagger]

Me too. That's why I came here. And then the first message I saw was the same complaint.

[wanton solar]

After a short conversation, I asked it "what do i ask you next time to get to the point more quickly?"... I liked the answer 😉

[tough phoenix]

Yea specifically the last few days have been really bad

It's like we're getting what original GPT3.5 was like, but now we're paying for it because the new 3.5 is worse

[wind urchin]

I feel chatgpt wont reach its full potential until the api for 4 is released and you can setup your own database for content you've already "taught" it

[tough phoenix]

Why is everything just a couple levels worse lol

[tough phoenix]

Woah didn't know it was possible to setup a database of stuff you've taught it so that it can remember. Does this mean you could have a pretty decent sized coding project you taught it, and it would know the context to the fullest extent of that database/

[wind urchin]

yes

[stable dagger]

LlamaIndex

[tough phoenix]

This is great news lol

[tough phoenix]

How do you achieve this, are there any tutorials

[wind urchin]

there's a few companies already developing projects with a similar approach (none that I particularly love) but the possibility is there

[stable dagger]

I read a good Medium article. Let me see if I can find it

[tough phoenix]

Would be great to see it, thank you !

[fair iron]

you won't be able to share the link to it here FYI

[tough phoenix]

^ Send in dms I guess

[finite bear]

#gpt-models hey

[vagrant ember]

👋🇺🇦

[wind urchin]

ChatGPT ftw!!

[drifting kettle]

how to switch to GPT4 in the previous conversation

[tough phoenix]

You can't lol, another limitation

[drifting kettle]

why, I don't want to set background and prompt again. I told it a lot in the previous conversation

[ember cloak]

because you cant😐

[green kernel]

What keywords should be used to generate an anime cartoon style cat, 3D, ip

[tough phoenix]

#images-discussions

[steep birch]

[sullen wigeon]

did anyone get their API key for GPT4

[fair iron]

some people here have access to the api

[sullen wigeon]

wow, I'm been waiting for quite a bit of time..

[fair iron]

I don't think it's on a first come first serve basis, but I can't confirm that

[sullen wigeon]

are they determining whether potential business ideas are more urgent or important

[fair iron]

I can't answer on behalf of OpenAI, so I do not know

[jade lava]

？

[sullen wigeon]

when you fill out the application, they ask how the API will be utilized

[storm yew]

Probably but my hunch is Americans get first access, because their dev team is in the US so they can more easily monitor API calls

[lean portal]

Did anyone else notice that the conversation memory for gpt4 was reduced to about 4000 tokens?

I thought I was going crazy, till I ran some tests, but it's similar to gpt3.5 now and hallucinates when I ask it to summarize anything that is roughly 4000 or more tokens.🫠

[crystal vine]

Hey guys, new to the Discord. New to GPT. Was trying to write a short description on a T shirt brand I am building called "God Is Violent" when asking GPT to write a description it wouldn't do so because the language wasn't appropriate. I am still learning how to talk to GPT. Is there anything I can do to bypass this?

[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.

[lilac hearth]

Is there a way to word the prompt that could omit the word Violent?

[cunning bough]

How's the response length when context gets close to 4000 tokens? I'm using the GPT-4 API and I decided to cap my context at around 4000 to leave another 4000 for potentially long responses. The model can only handle 8K tokens combined for input and output

[crystal vine]

Let me try. Maybe I can get the same expression of words that match the name by using the word "angry"?

[lean portal]

I don't have API access, still waiting...
My attempted usecase is fairly simple though. I prompt for a summary of a chapter from a novel, paste the text of the novel in two parts, and the model hallucinates some sort of crazy plot that doesn't correspond to what I pasted.
Sometimes it even forgets the initial prompt, and starts writing a continuation of the text, lol

[steep birch]

#gpt-4-faq

[little fog]

does it count input history, if so, 4000 is not even 4000 at second message

[fathom cloak]

honestly, getting a plus membership should've gave you some dall-e credits as well.

[lapis jolt]

Me too

[little fog]

have you tried novel ai?

[lapis jolt]

I noticed some weird stuff going on today

[little fog]

that works kinda well, and you can set bound that something ai will remember

[fathom cloak]

plus membership is a scam lmao

[little fog]

not really a scam, people just assume, they assume it is unlimited for gpt4, and if you look at the price tag, it didn't say so

[fair iron]

You can always use Bing's image generator!

[fathom cloak]

good suggestion actually, thanks

[hallow owl]

ask on the midjourney Discord, we can't help here

[raw plover]

Cant wait for next gpt-4-32k

[tough phoenix]

Yep, it’s insanely bad. Forgets absolutely fudging everything, it’s been a bad user experience for many people these days as the performance has been worse

[lean portal]

Good to know that I'm not the one hallucinating
The performance is actually getting worse with each update.

[mystic nest]

"Industry analysts expect OpenAI will likely require an additional 30,000 GPUs from Nvidia to maintain its commercial performance for the remainder of 2023 alone. "

I guess they might need more if they want to release multimodal features 🤣

[left sphinx]

hey how can i rephrase an article ?

is there any special function to it ?

[hallow owl]

ask in #prompt-engineering and check out the #1019652163640762428

[tough phoenix]

First they need to fix GPT-4 lmao

[gritty comet]

hey, anyone else gotten this in python? openai.error.InvalidRequestError: This is a chat model and not supported in the v1/completions endpoint. Did you mean to use v1/chat/completions?

[hallow owl]

ask in #dev-chat

[mystic nest]

think it's near impossible to fix gpt-4 in a short period of time, simply just too many users, the problem with too successful

thats likely why we got the limitations, trying to use what's available for more users

[fair rose]

GPU is the new gold

[mystic nest]

yeah, at least this time, it's not the consumer gpu

[fair iron]

Nvidia Tesla A100s 📈

[mystic nest]

think I read somewhere even microsoft is planning to produce their own chips instead of depending on Nvidia

it's just never enough gpu

[pliant ocean]

Are you chat gpt 4

[hallow owl]

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

[fair iron]

this! I think people severely underestimate how much compute resources GPT-4 takes to run

[pliant ocean]

How to get chat gpt 4 for free

[fair iron]

use Bing's AI. It's built on GPT-4

[tough phoenix]

But why promise a good product and sell it at $20/month

[fair iron]

I don't work for or represent OpenAI. I can't answer that question

[hallow owl]

$20 USD a month is insanely cheap for what you are getting imho

[tough phoenix]

GPT-4 is becoming more like 3.5 by the day. You can get similar functionality for free

The token limit going from 25K tokens to only 4K tokens was a huge blow

[fair iron]

was the token limit ever 25k???

[tough phoenix]

Now it can hardly remember what it’s talking about; it’s like it has amnesia after like 5 messages

Severe hallucinations

[wind urchin]

Wair what ellse dìd they reduce

[zinc agate]

I am trying to renew my subscription this month but my card keeps declining, I've tried multiple cards - anyone able to help?

[fair iron]

email [e-mail]

[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.

[mystic nest]

for some people, everything is just expensive/scam regardless, as long it is not near perfect 🤣

[mint harbor]

I'm using gpt4 and I have no issues with it yet

[wind urchin]

Gpt 4 special features?

[solar pollen]

After i have reached a cap of 25 msgs in 3 hrs, i was switched from GPT 4 to GPT-3.5 now after the time limit has been removed, am unable to go back to GPT- 4, again need to restart the session

[mint harbor]

Other then being slow as a snail and more accurate responses

No

What you guys using gpt for?

[wind urchin]

I need subscription for gpt 4?

[mint harbor]

I own my own business that is subscription based so anytime I use a service and it has a subscription I always subscribe to support the movement and the hard work developers have put in

So yes you gotta pay the $20 a month for gpt4

[wind urchin]

I can use gpt 4 with out subscription?

Bruh

[fair iron]

You have to have a ChatGPT Plus account, or you can use Bing's AI which is built on GPT-4

[mystic nest]

bing chat is just not optimised for complex question and some topics, but that's what you get for free

[zinc lion]

is there a big differnce between gpt 3 and 4?

[fair iron]

GPT-4 is better at basically everything. I still think 3.5 works well enough for most simple stuff though

[wind urchin]

I'm using gpt 3.5 but I need that gpt 4 but here i know i need to pay

[zinc lion]

For what purposes is gpt 3.5 not powerful enough? Are you just using it casually for hw and stuff?

[long finch]

Release date for Image input?

[mystic nest]

you can always try for 1 month

[mint harbor]

I use gpt4 to edit and modify my source codes on my applications, 3.5 would be allover the place where gpt4 I only gotta correct some minor syntaxes

[fair iron]

I don't know

[long finch]

Ah, OK. Thanks.

[wind urchin]

Ok

[fair iron]

I've had 3.5 write entire simple programs. They don't do much, but they're functional

I had GPT-3.5 write me a program that is a mimic of Wordle but runs locally the day it was released (Nov 30, 2022). I haven't tried it yet with GPT-4, but I imagine they'd be comparable

[mint harbor]

Mine are pretty complex I scan the u.s. stock Market and have an average of 75-80% accuracy

[fair iron]

Yeah... that sounds like something GPT-3.5 couldn't do 😆

[mystic nest]

being one source to almost everything itself is just nice.
like say you want to code something to do text to speech then another day basic machine learning, then another day you are interested in japanese

[lean portal]

Probably true, but there's a serious lack of open-ness from openai when they make any changes to a user's experience.
A customer should never find massive changes like-the conversation memory getting nerfred from 8k to 4k- via their own troubleshooting. Its jarring to go from having a decent process one day to getting gpt4-gaslit the next.

[mint harbor]

Lol well the the original code was manually written, I've just done some minor changes here and there

But yeah I wouldn't use 3.5 again tbh

[fair iron]

TOS moment 🙏

[mint harbor]

I agree 100% with you but it does take failure to meet success

[mystic nest]

I read some ppl use it for mental health therapy and reportedly work...so....

[fair iron]

I think the filters try to discourage this

[mint harbor]

One thing I really hated about 3.5 that they fixed with 4

Truncated Continuations on 3.5 were like taking your snake for a walk on the clouds in Uranus

[mystic nest]

yeah, some headline like this "People Are Using AI for Therapy, Even Though ChatGPT Wasn’t Built for It"

[wind urchin]

How can we give command to chat gpt

[mystic nest]

probably triggered some ppl to research more on the topic and formulate something better

a lot of stuff are discovered as byproducts

[fair iron]

I think I remember reading about this

[narrow ridge]

👍

[zinc agate]

Has anyone here had any billing issues with subscriptions?

[hallow owl]

what kind of issue?

[zinc agate]

even though I contacted support, my ChatGPT subscription keeps getting declined even though I have sufficient funds in my account

[hallow owl]

only [e-mail] can resolve this for you. You could try a different card if you have one.

[zinc agate]

How long is the turnaround time for Support to reply, Ive also tried a different card as well.

[hallow owl]

there is no ETA for a response from support, they are really busy

[long finch]

How many words can GPT-4 generate at maximum?

[wind urchin]

!q How do you make your girlfriend happy when she's angry

[grim hare]

why are you chucking a sook about having to pay haha

[wild furnace]

Hi all, after waiting for weeks on the waitlist (GPT-4 Api) I am wondering whether there is someone else who is still waiting too. Is there?

[hallow owl]

some people have been waiting since GPT-4 was announced

[fallen granite]

Has anyone tried GPT-4

[wild furnace]

And and least one of them is still waiting. Right? I won't contact support then...

[hallow owl]

patience is key here

[still solar]

Why do frogs fly

[wild furnace]

Thank you Robert. Will exercise more of that accordingly!

[teal whale]

Hello, what function do you have?

[dapper willow]

👋

[junior salmon]

You cannot access ChatGPT in this discord server, to access ChatGPT, go to: https://chat.openai.com

[pliant ocean]

Does Bing use chat gpt 4

[safe tinsel]

Yes

[lofty briar]

I want to subscribe to plus now, but stripe has rejected my card🤣

[vague geyser]

Hi 👋🏻 I see that gpt4 data is only up to Sep2021, any1 know when will new data be updated?

[pliant ocean]

I think Bing seldom use gpt 4

[hallow owl]

we don't know if/when this might happen, keep an eye on #announcements for updates

[wind urchin]

Anybody know of any discord bots that use GPT well?

I don’t like Clyde and he isn’t available in my servers yet

[strange bramble]

How can I crate an image ?

[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.

[vague geyser]

👌🏻🍾🥂

[whole kettle]

What just happened to GPT 4, it is really really dumbed down this time around. I can't even get it to properly continue its answers anymore, not to say that its feels really dumbed down in general all of a sudden. Anybody else having this issue ?

I started multiple new conversations just to make sure it's not because of the length of the conversations and the behaviour is the same.

[high crane]

What's going on with the website? I keep getting an error saying something went wrong over and over again

[split panther]

Same here

[pale sandal]

👍

[whole kettle]

It can't even remember what you talked to it like 2 questions before , it never had this issue before, it was able to read through the entire conversation (even in long conversations). It feels like I'm talking with 3.5 now.

[jade lava]

HELLO

[strange meadow]

SUs

chat gpd 4 is a great revelution in ai

[wraith ginkgo]

`` Server Tip: ``

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

[wintry marsh]

When is Chinese New Year?

@wraith ginkgo When is Chinese New Year?

[hallow owl]

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

[wintry marsh]

Thank you

[vagrant pawn]

Are you gpt 4 ?

[hallow owl]

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

[coral apex]

No

[severe thicket]

I think there’s a website dedicated to jail breaking gpt 4 somewhere

If you do some digging you should find it

[coral apex]

Where can I register an account for Chat gpt?

[indigo idol]

who else is getting addicted to roleplaying with chatGPT? i'm intoxicated

[hallow owl]

https://chat.openai.com/chat

[burnt musk]

Neither ChatGPT nor Google Bard can pass the Coffee Test. both will say that you can't drown in coffee.

[hybrid mirage]

hm

[echo wasp]

Pretty sure some people couldn't pass that test either. lol

[hybrid mirage]

?

[burnt musk]

aye, but we already know people are dumb

[hybrid mirage]

like

[echo wasp]

Not a good idea to discuss that in the OpenAI offical server, silly geese

[hybrid mirage]

oh right

[hallow owl]

breaking the rules is a good way to get your account suspended

[wet drift]

Seems like it's been a long time in the ai space for them to not release the image Analyzer.

[slow nacelle]

good morning

[wet drift]

I'm curious if it will be able to work with something like mid journey to create a professional website. Would be interesting

[low moat]

@wraith ginkgo where is the chatGPT ,i can't @ it

[spring bison]

Hello there! I see the #announcements channel hasn’t been as silent as now since… last year. Not a single update or even a small message to talk about the raise of the 25/ 3hours cap limit, nor any other feature. It would be nice to be a bit more transparent with your customers openai, it’s starting to feel really long now without any kind of news or communication

[dusky wigeon]

I don't know why I am getting this error in Plus Subscription many times "Something went wrong. If this issue persists please contact us through our help center at help.openai.com."

[hallow owl]

what web browser are you using?

[dusky wigeon]

I am using Chrome, Edge

it's working fine on mobile browser

[wind urchin]

try Firefox

[hallow owl]

try with Firefox and let us know if the problem goes away

[dusky wigeon]

yeah will try now

[pliant ocean]

Bing is creating any images by powered D alle

For example 4k, 8k, 12k and 3d model images

[north light]

Today i also feel that GPT4 is outputting less words. I am not sure its just my feeling or not. Did anyone experience the same thing?

[whole kettle]

It seems they also dumbed it down for the moment. GPT 4 currently feels like basically 3.5 undercover right now.

[tender cloak]

?

[charred heath]

i was using it earlier and it made me a great website
and currently it's just near to unusable for longer sections of code

[acoustic wigeon]

why I still face "Something went wrong. If this issue persists please contact us through our help center at help.openai.com."?

[hallow owl]

are you using Chrome or Edge? if so try Firefox

[acoustic wigeon]

oh, no one wants to use Firefox lol

is lots of guys facing the issue?

[hallow owl]

there is a caching issue with Chromium browsers that doesn't affect Firefox

[acoustic wigeon]

oh, ok I will wait for the team to fix it.

[pallid sleet]

Why I'm unable to uses ChatGPT it's show like how html normal page load without css 😂😂

[wind urchin]

how do i even access chat GPT?

[dusky wigeon]

Working smoothly on Firefox now. But I do use Chrome mostly, Firefox is taking a lot of RAM on my Laptop

[whole kettle]

Same here, got it for coding, but as of right now (and this started happening a few days ago) it got so dumbed down that it's not even useful anymore. First they reduced the number of things it can remember in the same conversation, that was already a big downgrade but now, they dumbed it down to the point it feels like 3.5. If it keeps like this I'll cancel my subscription and probably alot of other folks are going to do the same. This is the downside of having no direct competition, they can do whatever they want but I'd rather not use any bot at all than paying for something that is useless anyway.

[desert ibex]

bruh, this thing is working smoothly

[charred heath]

as long as you use it outside of peak hours it can still perform really well

you just have to figure out when people are not using it as much

[wind urchin]

I’ll cut my plus subscription as soon as I get API access

[whole kettle]

I won't bother with this, I won't change my schedule for their stupid decisions or play guessing games. We the subscribers support their service, if they are going to do this to us and think they can get away with it then good luck to them, they'd better have their pockets ready cause they're going to get back to paying for this thing out of their pockets again very soon. I didn't pay for a service which requires me to turn my whole schedule around for using their tool, I paid for a service which I can use whenever I need to use it. This is absurd.

[charred heath]

yea lol

[icy oracle]

[icy oracle]

Whats up with api??

[wind urchin]

it’s less limited

[pallid mantle]

hi all...any one know if there is some plugin website etc...that use chat gpt 4 to read and explain mathematical papers? 🧐

[hot walrus]

I wouldn't use gpt for that, it can't even answer basic calculations for me most of the time lol

[wind urchin]

Why is it taking so long to get plugin access

Im a bloody plus subscriber too, for a long time. Is my money no good?

[pallid mantle]

I saw it can answer some calculation and it is able to write Latex!!! next step should be explain papers 😄

[wind urchin]

Bro just copy and paste the paper lol

[hot walrus]

it can provide simple one answer type calculations but its laughably bad at more challenging stuff

[wind urchin]

It can do rather complex. Whats your definition of challenging ?

[pallid mantle]

@wind urchin it works? it does paste only text...or I am wrong...🧐😄

[hot walrus]

most concepts beyond basic university math

[wind urchin]

University math for what degree lol

A art degree doesnt have math like engineering

lol

[pallid mantle]

maybe the version with image recognition do it...but they have not still release it...

[hot walrus]

obv, talking about a stem student. like calc, linear algebra, and differentials

[wind urchin]

I use it for engineering stuff, its quite good still.

[hot walrus]

I tried using it for some discrete proofs earlier today and it was struggling

work in process, mathematical thinking is hard to replicate

[quaint plaza]

The problem with this chat bot is that it is regulated. Success will be gained by the AI that does not care about your feelings, only the facts.

[pallid mantle]

to have poece of code in almost any language it is wonderful

[wind urchin]

It sometimes do good math sometimes not

[quaint plaza]

This bot is being regulated and will fail in the near future.

[north light]

Ahhh. GPT4 is downgrading like GPT3.5, not able to do reasoning, not able to write correct code. Struggling a lot

[wind urchin]

It depends a-lot on how you ask the question. The set of instructions you give it.

Latex output is my new best friend

[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.

[slim ice]

im not 16 so i cant talk to chatgpt

sad emoji

[pallid mantle]

ens try to ask for Tkzi stuff!!! it draws!!

[slim ice]

?

[pallid mantle]

sorry...TikZ...😄

[slim ice]

i searched it up on google and still dont understand

[whole kettle]

Welcome to the club, same here. It wasn't enough that we only got 25 questions/3 hours, now you're going to waste those 25 questions trying to figure out what's wrong with the bot, realising that you're not talking to GPT 4 anymore, you're back to 3.5, enjoy your subscription, you're now paying for something that is basically free anyway, awesome deal.

[cedar remnant]

wait really?

[magic sand]

My plus subscription is experiencing slow response, frequent errors, and requires frequent refreshing. Are there anyone else who is experiencing the same issue as me, as if they do not have a plus subscription?

[whole kettle]

Yeah, they dumbed the bot down really bad, at least for some of us, not sure if everyone has this issue.

[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.

[mortal panther]

so I refunded the subscription about 1 month ago, now I would like to check it again, especially since there was an update yesterday, but is it a waste of money?

[whole kettle]

I'd strongly advise against subscribing right now, you're basically risking to get 3.5 quality responses, not kidding.

[cedar remnant]

btw is there a reason why free chatgpt, which uses gpt-3.5, and my discord bot, which also uses gpt-3.5, seems like the bot is using another model?

[mortal panther]

thanks for the info, I use the free 3.5 every day or so after I cancelled my subscription. When I cancelled the 4 it was unresponsive and left everything unfinished, the 3.5 has been doing its job ever since, it's a bit scattered in the more complex and longer tasks but we like it 😄

[whole kettle]

Yeah no kidding, if you want to I can show you my recent conversations with GPT 4, it's hillarious, it feels even worse than 3.5 at the moment, as in it can't even remember what you just asked it, it's not capable of continuing its own responses properly anymore, and it's as dumb as 3.5 is right now for me. Until now I would've recommend it to anyone asking but for right now, I don't but that's just my opinion. If you want to try it by all means subscribe but keep in mind you may be very disappointed.

[hollow trail]

✨

[whole kettle]

I think it may be related to the recent situation with the jailbreaks, some folks managed to get it say stuff it "wasn't" supposed to say and there are rumours they've made some changes which well, ended up where it is right now. That's why we can't have nice things I guess. I suppose it will get fixed but how long that will take I have no idea.

[mortal panther]

I believe you, unfortunately I also experienced it myself, which is why I cancelled my subscription. I expected a paid model to be fixed faster. The same thing happened to me with version 4, it wouldn't finish what it started and almost everything resulted in errors. By the time I would have described the problem, I hit the limit, so I said thank you and decided to stick with 3.5, which works properly and I am already good friends with it. 😄 I'll stick with the slightly simpler but properly functioning 3.5 until there are major changes and I hear about better experiences. Also, I'm very interested in plugins, as I think they can achieve much better results. Thank you again for all the information about the current situation, and let's hope this kind of problem will be solved as soon as possible so we can use this fantastic thing with its amazing capabilities productively!

[mortal panther]

Yes, I have heard about it too, and I tried it a while ago, but it created too fanciful situations for me. 😄 Anyway, it's not important, I don't need functions that are prohibited. Unfortunately, I think this results in interfering with more normal functions, so it's bad for those who didn't use those prohibited functions and are just suffering from the limitations. :S

[whole kettle]

You should've seen GPT 4 a few weeks ago, I had a coding conversation with it with over 50 questions, like a very very long conversation thread, during the entire conversation when I asked it for help it would remember the entire conversation, and even remind me of some things we'd been discussed much earlier in the conversation, it literally kept it like that until the conversation reached its limit, it was amazing, truly amazing. That's the GPT 4 I subscribed for, it felt totally worth the 20$, not like today where it can't even remember what we talked 2 questions ago or not being able to continue its own answer.

[little forge]

#gpt-4-image

[quick dune]

You talk like you just lost a friend, bruh🤣. But, yeah, it's dumb as hell today

[whole kettle]

I just lost my coding budy, of course it hurts, he was the best. I wouldn't talk about it if it would not hurt. I hope they bring it back some day, I miss my smart friend.

[grim notch]

one quick question: will i be able to use GPT-4 in my older conversations? (conversations i had with 3.5)

[rocky fog]

Hello, i'm getting everytime Something went wrong. If this issue persists please contact us through our help center at help.openai.com. with firefox

[magic sand]

"Something went wrong. If this issue persists please contact us through our help center at help.openai.com." 😢 Why does the plus subscription have such problems now?

[rocky fog]

me too

[magic sand]

Since when did this situation start to occur for you?

[pearl dagger]

just dropping into this conversation but it's pretty apropos of what I was coming in to ask. lol. I was trying to get gpt4 to summarize the chapters of a book I'm writing for a client to have an easy guide I can refer to. Spent most of the time trying to get around the content policy for a novel that is YA with no adult content and with conflict at basically Hunger Games level. By the time I managed to get it to accept 750 words of a chapter, I ran out of my 25 prompts. Honestly, what's the point if the censorship is so heavy-handed and when you finally figure out workarounds, your time is up and you're booted for 3 hours?

[mortal panther]

Yes, I felt the same way at the beginning, and it hurt me a lot when the limits and the dumbing down came. That's why I'm here today, maybe the shine has returned. 😄

[pearl dagger]

Also - I used the s word that ends with an x in my original post above instead of adult content and was banned from this chat for 5 minutes. lol

[whole kettle]

Yeah well there are alot of children in here so that's understandable.

[mortal panther]

As far as I know, it's not possible in the basic case (although it may be possible to bypass it with a plugin), but the model you selected at the beginning will be used for the entire conversation (in this case, 3.5). You can copy the conversation to a new one with 4 if it's not too long.

[wind urchin]

I have a question. I am currently using a VPN to connect to ChatGPT in China. I can connect and use it successfully, but I found that if I take too long to reply during a conversation, the connection seems to automatically disconnect. When I click the enter key to confirm my new reply, the page will prompt an error. I can only refresh the entire conversation interface and re-enter my reply. How can I avoid this problem?

[eager totem]

U cant, many people have this problem

[mortal panther]

If I interrupt the conversation for half an hour or more, then I also need to refresh the whole page to continue (without VPN).

[wind urchin]

no. i just interrupt the conversation for just more than one minute i g

though it was error

[magic sand]

are you plus？

[wind urchin]

no

[bleak junco]

I'm a Plus user, I get an error: Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

Anyone else ?

[wind urchin]

so do i !

same

[magic sand]

That's normal. I had the same experience before I upgraded to plus. After you upgrade to plus, you will have an extraordinary experience.

[clever plinth]

I live in Bahrain, and can’t get to subscribe to plus version. All countries around us are able to get it. I think we were missed by mistake 🤣

[magic sand]

me too，since about one week ago

[wind urchin]

i thought there be something wrong with my vpn , though it was all fine for other websites

[magic sand]

no， everyone in China without plus will face this problem，it is not about VPN

[wind urchin]

may i ask you that are you also chinese ?

[magic sand]

ofcourse

you can give me you QQ number，i can talk to you

[cobalt remnant]

you can give me you QQ number，i can talk to you

[jade lava]

Why can't we see stars in outspace?

[cobalt remnant]

Because we are locked in the room

[lament void]

Because atmosphere scatters their light.

[jade lava]

No because there's no stars in the fact

[finite trellis]

We literally have a star visible every day

The whole planet spins around it

[pearl prairie]

when will this image feature be available

[jade lava]

Yes,only one

[knotty grotto]

Soon ™️

[finite trellis]

There other states are much further away, but can be seen and measured with special equipment

@jade lava ask ChatGPT buddy

[distant axle]

hi, boss, how to chat with ai

i am rookie

[robust quiver]

is the difference between chatgpt 3 and 4 massive

[scarlet prairie]

ask any question same like you to your friend

[knotty grotto]

Chat.openai.com

[scarlet prairie]

yes there is

[robust quiver]

do i get chatgpt 4 by getitng plus

[distant axle]

thanks, boss

[scarlet prairie]

what is the link on chat gpt playground

[knotty grotto]

Relevant Links
💬 Try out ChatGPT at https://chat.openai.com/
🤖 Access GPT-3 at https://platform.openai.com/
🎨 Make art with DALL∙E 2 via the API or in https://labs.openai.com/
📩 Read FAQs and reach customer support through https://help.openai.com/

[robust quiver]

so can i get chatgpt 4 through plus

[shrewd fable]

when will image reading will be available?

[knotty grotto]

Yes

[knotty grotto]

Soon™️

[candid vector]

what`s happening to gpt 4? it has been going pretty bad these past two days

[knotty grotto]

Server can be at peak of users and have impact

[candid vector]

usually when i write continue after it truncates the text because of character limit, it continues the task, but now it`s just doing whatever

it also feels as if it`s memory is getting shorter, like 2 messages short

[potent fjord]

I just keep getting the "something went wrong" error for over two weeks in my laptop. If I change to another device with the same account it works well but I want to fix it. I chat with the bot in the help menu of OpenAI and give me this discord. Sorry If this is an offtopic question

[knotty grotto]

Try Firefox as test

[candid vector]

so, nobody knows about why the continue command doesn`t work anymore? has there been maybe some code update?

[knotty grotto]

#gpt-models message

[candid vector]

if it happens after you havent typed a new chat for a few minutes, its because of that, just refresh the page

[knotty grotto]

And you can also ask to do better continue prompt in #prompt-engineering

[candid vector]

what you mean?

[hybrid mirage]

..

[knotty grotto]

They are several way to ask to * continue * some way are better than other

I don't know them all so people in #prompt-engineering can help you

[potent fjord]

It happens every time on my laptop but if I use it on my phone for example it works well

[weary dome]

Are competitors of OpenAI able to make something that can process more tokens?

As opposed to paramerers

[potent fjord]

why it works on firefox and doesn't work in opera, google or edge?

[knotty grotto]

You have something that block

• extension
• cookie and cache
• addon

You have to troubleshoot it.

[fathom summit]

when will gpt 4 launch ?

[knotty grotto]

Good news ! already is

[potent fjord]

but I use it on chrome or edge in other devices and it works well

[knotty grotto]

Edge and Chrome are chromium based

[wind urchin]

is chatGPT down? Can't access it

[knotty grotto]

Status.openai.com

[wind urchin]

thanks:) @knotty grotto

[fathom summit]

so , how can i sign up?

[knotty grotto]

You pay the monthly subscription

[fathom summit]

ohk

kk

[silver pike]

how can i make chatgpt make gpt5 for me?

[silver pike]

it's not giving me a good answer

[fathom summit]

haha

[silver pike]

sorry?

i think it's a user friendliness problem

sorry, i dont understand what you are talking about

i'm confused
why are some people here so hostile

we need a mod in here
you're breaking rule 1

[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.

[knotty grotto]

Let's stay polite and respectful in the chat please , no need to be rude

[silver pike]

thank you mod

[knotty grotto]

I am not a mod

[silver pike]

what are you then?

[knotty grotto]

Helper

[silver pike]

thank you helper*

please dont mini mod. also stay respectful 😄

@modern mason this member is harassing me for no reason
please look into it 😅

[dusk stratus]

respect is earned not given

[knotty grotto]

Rule 1

[dusk stratus]

trolls are annoying

[knotty grotto]

Kindly ask people to switch channel

Use the report option or dm modmail. But insulting or being disrespectful doesn't help you. You will be mute and later ban if continuing

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 1 `` Be respectful.

Treat others the way you would like to be treated, and assume best intentions. Don’t harass or attack others, and don’t engage in hateful or generally malicious behavior (e.g. sexism, racism, homophobia, etc.). Keep the negativity to a minimum.

[pearl musk]

do anybody can use chatGPT ？

I can't open ,always loading

[knotty grotto]

Try incognito

[wooden valley]

Hello, I recently got chat gpt-4, what should I ask it to test its abilities?

[pearl musk]

too

[knotty grotto]

What country ?

[pearl musk]

Forget it, I'll call api to make do with it.

[wooden valley]

yup, using it right now

[pearl musk]

USA

[knotty grotto]

Let's your imagination out, if you want to share result please use : #1050184247920562316

[knotty grotto]

Disable VPN, download Firefox and try again

[pearl musk]

never mind

[wooden valley]

Ok

[pearl musk]

I give up

I use gpt4 api now

[nocturne sierra]

Literally this question

[dark fern]

this completely depends on your usecase, unless you got it for giggles then i'd say give it a writing task on a subject you're familiar with and compare it to the 3.5 output

what i've seen so far is less repetive sentences, able to create more complex structures in writing and still have em make sense, better at roleplaying etc.

[wind urchin]

same

[drifting cosmos]

How to train ChatGPT on your own data?

[night ermine]

not possible

all you can do is provide it some context from your data

[drifting cosmos]

Why. Then how ChatPdf works on the data provided by the user

By making user to upload pdf file

[dark fern]

you're talking about a dataset not a training set

[night ermine]

because traying NNs is very demanding process

[fiery merlin]

why keep loading the page? i have subscribed plus version

[drifting cosmos]

Okay is there is any way by which I can ask user to upload the data and then ask the information according to that

@night ermine

[fiery merlin]

just send the data by text, gpt can get it

[drifting cosmos]

I want user to upload the data

[wraith ginkgo]

`` Server Tip: ``

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

[wind urchin]

Can images be uploaded in plus version?

[blazing scroll]

no

[shy cape]

Do you think the owners of this discord will create channels by subject matter such as mathematics, coding etc pertaining to gpt4 sort of like the Dalle - 2 channels where you create art by subject?

[drifting cosmos]

??

[fiery merlin]

correct

[wind urchin]

i asked chatgpt, "can I upload images?", it kept redirecting me to a sunrise symbol in chatbox which apparently was not there

[blazing scroll]

if you need help with api then there are already channels like #dev-chat but i do not see why they should create channels for math and coding!

[blazing scroll]

hallucinating, it does not know interface it is running on

[wind urchin]

Who? ChatGPT?

[blazing scroll]

yes

[native lichen]

Something went wrong. If this issue persists please contact us through our help center at help.openai.com. ????

[blazing scroll]

does refreshing fix it?

[native lichen]

nooooooooo

[blazing scroll]

try clearing site data

[shy cape]

@blazing scroll I don't need any help but rather making suggestions where people could focus there energy in various subject focused channels. Right now it's a free for all

[blazing scroll]

ok

[native lichen]

not working

[wind urchin]

same happening

[blazing scroll]

maybe it is at capacity idk, working fine for me tho

[native lichen]

i am premium mumber

[wind urchin]

which place are you at?

[fiery merlin]

same here

loading the page for long time

[blazing scroll]

actually this error message is old since i joined this server. and for most people clearing cookie or cache or just refreshing help or trying differnt browser but idk why it does not work for you

[fiery merlin]

i tried change another browser. not working

[wind urchin]

i noticed a writing style in ChatGPT where it sums up the headings of all the points and writes it up as the conclusion. Is GPT4 any different in terms of writing? Does it make mathematical errors as well?

[blazing scroll]

yes gpt4 has similar writing style and usually summarize at the end. yes all language models make mathematical error unless augmented by another system but gpt4 is way better than 3.5 in term of writing, coding and math.

[robust quiver]

someone that has gpt4, pls run something for me curious to see the output

[exotic python]

Hey hey

[wind urchin]

hey, thanks. Do you know of reliable Video creation AI tools as well?

[desert elm]

if theres a F website that post all chatgpt answer.

its not exciting anymore.

[robust quiver]

???

[blazing scroll]

no but please ask it in #ai-discussions someone might help you

[native lichen]

tell the prompt i will

[wind urchin]

Okay!

[wheat sparrow]

:Gun1: :angrylove: :Gun2:

Fr

[quick kelp]

what is api and gpt4

[blazing scroll]

API is a software intermediary that allows two applications to talk to each other and gpt4 is the latest and more accurate LLM by openai.

[pallid crag]

gpt4 is really weird now, he doesn't remember anything I send and also the responses doesn't make any sense. anyone else have those issues?

[amber cosmos]

How long do I have to be on the waiting list to use GPT4 api?

[wooden valley]

Nope, working for me

You might also have to provide more context to make it understand what your referring back to

[knotty grotto]

No eta, and no guarantees you have access before going public

[amber cosmos]

So I can not pay for gpt4 rn?

[knotty grotto]

You can pay for chatgpt 4

[amber cosmos]

chatgpt4 API?

[knotty grotto]

Api and subs are too different thing.

[amber cosmos]

i mean i want to pay for v4 api

[knotty grotto]

Like I said api you need to join waitlist. And they are no guarantee you will be chosen . You have to wait till it is available to pay for everyone

[amber cosmos]

currently im using 3.5 turbo api

[knotty grotto]

And they are no eta for that

[amber cosmos]

got it, thanks

[native radish]

CAT

[unreal swallow]

is there anyway to bypass the 15,500 character limit on chatgpt+?

[knotty grotto]

No

[patent sphinx]

Hey all, any ETAs on that image upload feature where the AI analyzes it? Thanks.

[wind urchin]

i generally copy paste the last few sentences it has written. Then ask ChatGpt to finish in the same dialog box. I t world

It works*

[winter cape]

No

[primal salmon]

No

[patent sphinx]

Dang.. seems like for those 2 weeks or so when every major bluechip was announcing their new AI products, none of them have actually released it yet... such marketing.
OpenAi - Yo, our AI can analyze a joke from an image & predict what might happen next in an image!
Microsoft - We gonna implement AI into the Office Suite, check out our demo!
Google - Google office apps will integrate AI, also Bard will be awesome (lol)
Adobe - Firefly will let you edit images with prompts. It's gonna get crazy!

Only one that I can recall that delivered was MidJourney with 5.0 👏🏻 Everybody else is playing vapourware games.

[wind urchin]

true, seemed like a panic response to GPT release

[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.

[atomic venture]

Is it possible to use GPT4 without paying? Bing is bad

[wind urchin]

GPT-4 is a thing

[hot cosmos]

I just ask it to continue and it’ll pick up where it left off

[dawn burrow]

may i ask a question？

[leaden cape]

yes

what

[knotty grotto]

No guarantees to have access

[wind urchin]

?

[wind urchin]

What do you mean?

[leaden cape]

you said "slide in my dms" and then said sum about GPT 4 API key

[knotty grotto]

You can apply for the waitlist, but you have no guarantee to be chosen. It might be only when available for public

[wind urchin]

Yeah I have been waiting for 1 month

[leaden cape]

but why did you ask him to slide into your dms

[wind urchin]

Because he probably got GPT-4 but can't pay

[knotty grotto]

That's against the rule btw

[leaden cape]

first of all you searched for his message about gpt4 and if you wouldve read the message after what he said you would know he can only use GPT and wishes for GPT4 to be accessible without paying

[drifting cosmos]

Hey anyone got GPT-4 access

[leaden cape]

and since he (obviously) isnt a plus member he doesn't have access to the GPT4 API

[wind urchin]

Oh I didn't know that, for real?

[wind urchin]

My bad then, I thought he had an access haha

[supple violet]

its against tos

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 0 `` Adhere to the following Terms & Policies:

Discord

• Terms of Service
• Community Guidelines

OpenAI

• Terms & Policies

[wind urchin]

sure but waiting for 1 month to get an api key is criminal 😢

[leaden cape]

everyone gets their turn eventually

[sand hull]

Limited number of people that get access and it isn't first come first serve. I just bought Plus to get access and luckily a week after got API access.

[supple violet]

you may never get the access via the waitlist and have to wait for it to become public

[vague geyser]

Without real time data, is gpt4 necessary?

[candid sorrel]

@wraith ginkgo

[wind urchin]

I've been in ChatGPT plus its release, still nothing 😦

Have you done something particular? contributing to code or smth?

[sand hull]

Just told them my usecase and got lucky.

[wind urchin]

Does multiple waitlist inquiries reset the oldest one?

[supple violet]

most likely as many openai resend features replace old one

[wet drift]

I'm kinda surprised it's taken this long to release the image Analysis features

[leaden cape]

i just hope i get the plugins soon

[fair iron]

CLIP Interrogator has been out for a while. You may want to look into it

[leaden cape]

@sand hull did you get emailed by then that you're accepted or did it just pop up in your openai dash?

[sand hull]

Got an email.

[elfin oasis]

@sand hull What did you list as your use case? I am still waiting on my access 😦

[wet drift]

I want to see how effective it would be to create a website picture in mid journey then have it code it in chatgpt with just the picture

[sand hull]

Getting access is a mix of having a good usecase, getting lucky, having plus or not, and submitting evals. Even then it's a limited release.

[sand hull]

I just said I wanted to try it out for fun. I got lucky I guess.

[wind urchin]

Hahah come on, I thought it was a complex use case

[elfin oasis]

lol

[wind urchin]

trolling poor gpt-3.5 users is against tos!!

[edgy jolt]

Please tell me the main features of laserpecker4

[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.

[floral kindle]

when’s gpt4 open to the public

[mental burrow]

how are we supposed to know?

[knotty grotto]

Nobody know

[loud wigeon]

We have detected suspicious login behavior and further attempts will be blocked. Please contact the administrator

[hallow creek]

Guys do you know when the GPT-4 API is coming?

[loud wigeon]

why

[rustic scroll]

Change VPN sever

[hallow creek]

it is

[knotty grotto]

It is available for people who access from the waiting list. For public Nobody know

[loud wigeon]

does not work

[hallow creek]

contact the administrator @loud wigeon .

[knotty grotto]

Disable VPN

[dusky hinge]

?

[loud wigeon]

Where can I contact the administrator? I have automatically deducted $20 for renewal today, so I can’t use it anymore.

[knotty grotto]

What country are you in

[loud wigeon]

U.S.

[knotty grotto]

Does it work with cellphone and data

[loud wigeon]

I tried it just now, and it works on my phone, but not on my computer.

[knotty grotto]

Then you have something that block on your computer you have to troubleshoot.
Here some hints

• vpn cause issues
• cheap antivirus like avast cause problem
• some extensions , addon on chromium product can cause issues. You can try with Firefox as test
• not deleting cache and cookie properly

[loud wigeon]

ok thank you thank you so much

[wanton ore]

hi, does anyone know if you can check how many prompts you have left to use while using gpt4

[serene cipher]

it used to have that, but not anymore

idk why they removed it, it was probably causing issues about the counter not matching the actual server side value due to the counting or not of incomplete requests

[wanton ore]

thanks, so does that mean theres no way of knowing if im reaching the cap limit ?

[serene cipher]

it will let you know when you run out

but yea, otherwise, no

[wanton ore]

ahh thats annoying, thanks for your help!!

[quick kelp]

show me how to use ai and api

[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.

[mossy gull]

@wraith ginkgo

[mortal apex]

guys

[lapis bane]

For some reason gpt4 seems dumber today, and reverting to the old mistakes chatgpt 3.5 made. Seems much faster now than before tho

[lapis bane]

Imagine this is a scam too lol

[marble warren]

Is there free gpt4 ?

[supple violet]

no

[daring jackal]

🥹

[marble warren]

Do u guys use gpt for money?

[daring jackal]

I don't, but I'm considering

[marble warren]

But like how

Everyone knows the trick

Who would fall for it

[daring jackal]

What's the trick?

[glossy kernel]

How to use prompt-library? any one can help me ? Thanks

[daring jackal]

There are many prompts in github, you can also use fine-tuning prompts if you're building something

[velvet hamlet]

[daring jackal]

💀

[median sedge]

has gpt-4 gotten the ability to use images yet?

[supple violet]

no

[sacred pasture]

it can read whats in an image url

[warm stone]

hiii

[sacred pasture]

ie: upload a picture of your dog somewhere, ask what chatgpt see and it will see your dog

[leaden cape]

i doubt that since ChatGPT has no access to the internet

[golden quartz]

guys

do u think gpt+ is worth it?

[leaden cape]

yeah

[sacred pasture]

i m not sure how that works but it works.

do something like describe whats on that picture put url, first time it will say that it cant see without internet or url then will ask you for url then after that just repast the url and it will see the picture

[somber dew]

kk

[pearl tangle]

What if I want to smoke

[leaden cape]

#off-topic

[primal rampart]

am here again because I hit my cap for gpt 4 :\

[leaden cape]

thats unfortunate

[shy cape]

Does GPT4 have the ability to run Python code and produce an output? It was mostly math but it performed this it's own and for context these are formulas for determining optimal bid and ask order values using the following:

St = 10000 # current price
It = 50 # inventory
t = 3600 # current time (1 hour into the trading day)
T = 86400 # time horizon (24 hours)
gamma = 0.001 # risk aversion
sigma = 0.01 # volatility

These are the formulas:

Calculate Inventory Risk Adjusted Mid Price

Q_mt = St - It * (gamma * sigma**2 * (T - t)) / 2
print("Inventory Risk Adjusted Mid Price: ", Q_mt)

Calculate Optimal Bid Price

optimal_bid_price = St - (2 * It + 1) * gamma * sigma**2 * (T - t) / 2
print("Optimal Bid Price: ", optimal_bid_price)

Calculate Optimal Ask Price

optimal_ask_price = St - (2 * It - 1) * gamma * sigma**2 * (T - t) / 2
print("Optimal Ask Price: ", optimal_ask_price)

This was the solution it provided:

Inventory Risk Adjusted Mid Price: 9995.671875
Optimal Bid Price: 9991.34375
Optimal Ask Price: 10000.0

[tawny trout]

what happened to the GPT bot?

[thin ingot]

I just started Colt Steel's OpenAI API course and it's awesome
Unfortunately I don't have GPT-4 API access yet 😦 But it works with 3.5-turbo
Between the phenomenon of AutoGPT and agentic AI in general I really want to play with GPT-4 to make a thing
I want my access 😅 Does anyone have any suggestions on getting through the waitlist or is that just a mystery right now for independent devs?

[shy cape]

Sorry for the giant font

[supple violet]

upcoming code interpreter can run python code but it can take time before it will be public

[daring jackal]

Lol

[shy cape]

Ok cool. For anyone interested, GPT4 is really good a math. I "trained" it on a 29 slide presentation covering the concept of optimal market making and it pulled out the most relevant parts and we settled on the formulas I shared.

[daring jackal]

I want to use autogpt so bad 😞

[round silo]

AA

[thick bear]

do they change how prompt works? as usually i throw snippet code first then the problem but doing it endup asking how the code works

[daring jackal]

I just say fix my code and then paste

But only 2 out of 10 snippets are debugged

Then i have to create my own logic with gpt to work it out

[thick bear]

also did they broke "continue" tab as its blocked input

[daring jackal]

Continue doesn't work anymore?

[thick bear]

it does work but it wont float on the top of input

[fallow ore]

over the past 12 to 24 hours I appear to have to log in frequently to use chatgpt. Is this normal? Is anyone else experiencing this?

[digital bridge]

@fallow ore My Google account is also unable to log in.

Why is typing Chinese subtitles here resulting in being banned?

[true dew]

hii

[ancient apex]

can you speak Chinese

[true dew]

no can you speak arabic

[somber meteor]

when will chatgpt 5 come out 🙂

[supple violet]

english only server

[somber meteor]

will microsoft release bing with chatgpt4 ?

[teal prairie]

So ive been having some very interesting conversations with GPT4 then i fed the same prompt to 3.5 and the ‘spicy meter’ went through the roof. I have created a prompt for Chaos mGPT and let me tell you. ChaosGPT on 3.5 is something else. Its kinda starting to scare me TBH.

[somber meteor]

anyone ?

[supple violet]

bing chat has used gpt4 for nearly 2 months

[somber meteor]

thanks .

[visual vessel]

do we know when plus subscribers will be able access internet from gpt4 ?

[limber bone]

hi guys!

[wraith ginkgo]

`` Server Tip: ``

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

[nimble sapphire]

ok fine

good evening

[frail fractal]

is it slow for you guys as well?

[nimble sapphire]

I don't konw. just so so.

[fickle summit]

thanks

[mint aurora]

Has anyone here done an EVAL yet?

[prime stone]

Is long term mem plugin already added?

[mint aurora]

I’m trying to better my understanding of it, it’s like evaluating gpt4 to get access quicker

[jade wyvern]

do u mean the API?

also I havent heard of that. What is evaluating gpt4 exactly?

[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.

[smoky crest]

Did you even realize you got phished yesterday?

You sent a scam server to all ur dms (me included)

I just noticed it :p

[gaunt seal]

lmao imagine getting phished in 2023

[smoky crest]

Fr

[sweet thistle]

I'm fairly sure they're referring to https://github.com/openai/evals

[fickle summit]

Who developed it

[smoky crest]

😂

[jade wyvern]

ah I see thanks!

[robust quiver]

😂

[smoky crest]

He did lmao sent “TikTok 18+” server in my dms

Bro was downbad

[sweet thistle]

Be sure to DM @open crescent if that ever happens in the future

[wind urchin]

Can anyone tell me if the $20 a month is worth it for GPT-4?

[gaunt seal]

no

[smoky crest]

Yes

[gaunt seal]

make it 50 tries

per 3 hours

[smoky crest]

Just wait for more

[gaunt seal]

or 100 like it was before

[smoky crest]

Patients my guy

[robust quiver]

wait isnt it unlimited tries

[gaunt seal]

I've noticed gpt4 has gotten faster

[smoky crest]

No lmao

[robust quiver]

y not

[gaunt seal]

LMAO, imagine getting unlimited tried on PAID subscription

[robust quiver]

yh exactly

[smoky crest]

8 per hour

[jade wyvern]

depends. GPT-4 is pretty good, but the limit is making its worth lower. but 3.5 and its fast response is alright i guess.

[sand hull]

Ever heard of a maintenance cost?

[gaunt seal]

ever heard of giving us access to api so we can bear our own cost

[jade wyvern]

still on the waitlist unfortunately

[smoky crest]

Plus all the alpha stuff is great

If you have it

[gaunt seal]

almost nobody has it

[sand hull]

That's what I do.

A bunch of people have, more people weekly. But it's a limited release and isn't first come first serve.

[smoky crest]

Says who

[gaunt seal]

says I

[unreal schooner]

yes

[sand hull]

You not having it is not equal to almost nobody.

[smoky crest]

Look dms rain guy

[wind urchin]

there is any person who help me ?

[knotty grotto]

That's against the rule

[sharp spear]

Sorry officer

[celest totem]

if i press regenerate answer does tha tcount towards my cap?

[kind wing]

I need parfum image

[digital bridge]

Good night, everyone

[green nexus]

imagine

[smoky crest]

Imagine imagining

[supple violet]

yes

[short sage]

Are there any signs of when the image recognition of gp4 is coming to the public?

[knotty grotto]

No eta

[short sage]

mmmhh

[scarlet shuttle]

hey guys I have a question, so much of my code doesn't work, think it could be like a using windows? I always get weird like line this line that then chatgpt is like sorry for the confusion try this and it still doesn't work, all i read is how good chat gpt is at coding but like, if i ask it for a chatbot, using my openai and pinecone api with a UI, it will never work. is taht just past its abilities or is it my phython/comp/windows?

[upbeat dome]

Is Tiktokenizer accurate in predicting token allocations for GPT-4 prompts?

[red egret]

@scarlet shuttle
#1093591470293926091 #1088285207934812182 #prompt-engineering

[red egret]

I've been wondering this myself. The API's count should be correct, but the tokenizer tool on the website still says GPT-3 and Codex

[upbeat dome]

I'm still on the GPT-4 API waitlist, so no way to verify there.

[scarlet shuttle]

sorry is taht says that's where I should be asking?

[zenith trout]

Hey, Mr AI Friends!

[scarlet shuttle]

OK!

[red egret]

One is tips on programming with ChatGPT, CodePup is a prompt for programming I created, and #prompt-engineering is the best place for questions about how to get a certain behavior from GPT models/ChatGPT

[scarlet shuttle]

WHERE CAN I FIND THAT BAMBOO? and thank you both so much I'm so new to this

[tough kestrel]

Personally i don't like paying closed source AI companies because of the potential for them to misuse my money, for example by developing further gated AI products built on open research, which would have a negative social impact by concentrating wealth and power, and slow down scientific and technical advancement. I keep OpenAI on a permanent waiting list though, I will subscribe to their service one day, maybe, if they align with my values.

[limpid terrace]

do you have to own a company to join to wait list?

[red egret]

no

[scarlet shuttle]

yeah! like everything I write, never works and i just am swamped with python errors so careful offering that to me

[red egret]

In platform.openai.com, people who are individuals get an Org ID for 'personal'

[limpid terrace]

but it ask me for company id

[red egret]

That's what I am saying.. look in the platform site under account

Look here: https://platform.openai.com/account/org-settings

Everyone starts off with the organization of "Personal" and a generated ID. You can set it to any business name if you do have a company, but otherwise just take the org id and use it for the waitlist.

I would share a screenshot, but this channel isn't allowing images.

[scarlet shuttle]

seperate question how do you know if you have api access for gpt4? I can't find anything on the platrom.openai site indicating the access i have

[red egret]

You should get an email, I am told. Otherwise, just look at the playground tool in Chat mode to see what API's it offers, or run an API request to the models endpoint for a full list of models available to you

[scarlet shuttle]

how do you find the model or engine settings for 4?

[red egret]

The docs say the GPT-4 model is 'gpt-4'

[cold kindle]

What is best prompt for resume writing? I keep getting error messages when trying to use them

[tough kestrel]

Anyone got access in the past 2 weeks? Or did they stop roll-out completely?

[scarlet shuttle]

really its not like a model_engine = "text-davinci-002"

its just model_engine = "gpt-4"?

[red egret]

https://api.openai.com/v1/models
When it prompts for username and password, leave username blank and paste an API key for password. You will get a list of all models available to you.

model_engine is a variable in which you are storing your model selection

[scarlet shuttle]

awesome! thank you, just gpt-3.5-turbo, damn

[red egret]

also, for the api it should be "model": "Name-Of-Model"

[wind urchin]

Create a desert code in Roblox Stydion

Create a desert code in Roblox Stydion

how could i speak to the ai

[mystic nest]

pray in your heart and strongly believe in the chatgpt...you will eventuallly........then just go sign up https://chat.openai.com/ if your country didn't ban chatgpt

[old pelican]

can i use chatgpt here in dis?

[knotty grotto]

This isn't chatgpt

No

[knotty grotto]

Chat.openai.com

[ashen horizon]

anyone getting bad responses today (im using edge)

[mystic nest]

all seems good to me

[wicked comet]

Can ChatGPT generate new knowledge that would not be considered a hallucination? Phrased another way, do people believe ChatGPT can uncover unknown "truths" of the world?

[swift cedar]

Is there any talks of unlimited gpt4 for paying people? Or faster gpt 4?

[knotty grotto]

Not for now

[mild fern]

How do I make chat gpt talk to me as if he was a professional in lawn care I need help with my yard

[uncut nova]

you cant

[cloud bison]

I could be wrong, but I feel as though there's an exploit in the way that the AI checks if content is safe. Namely, I think complex innocuous early instructions that set the framework for later things that defy its safety protocols slip through the cracks. That is, if you give it a long enough set of instructions on how to deal with out of bounds content, building the framework for its use, and then include the out of bounds content placed in that framework later in your prompt, it doesn't seem to do much to filter it.

[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.

[cloud bison]

It only works on a fresh discussion to be clear, if you set the context that put the AI on guard, it seems to do a more thorough analysis of your prompts before replying.

[wind urchin]

Nope

I heard chat sonic uses gpt-4, is there any difference from the gpt-4 plus users have access to?

[wind urchin]

yeah thats what it does

from my experience atleast

[mystic nest]

if they use openai chatgpt 4 api, should be the same, but likely there will be charges involved and more if you are a heavy user

gpt 4 just ain't free

[wind urchin]

yeah

[wind urchin]

Everyone uses their own “gpt4” I would assume like the other ones I’ve tried they’re not as good

[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.

[wind urchin]

Just be careful for websites claiming gpt4 but just using gpt3 with system prompts

[wind urchin]

alright i'll just stick to the subscription then

👍

[cloud bison]

I disagree with the above user. Novel information is often just recombination or extrapolation of existing information. ChatGPT as available to the public now is not powerful enough to uncover too much, but, the GPT 4 model fine tuned, with access to plugins, such as say wolfram alpha, and using the 32k version, along with a really high end vector memory database, could absolutely uncover new knowledge with help. For example, the rote work that one mathematician might have to do to solve a complex novel math problem could easily be pushed through GPT such that its functionally discovered by GPT albeit with human assistance.

Likewise a lot of unnoticed correlations and general knowledge could also be discovered. Most things aren't innovations born from a vacuum (Human brains don't even do that well with that, unless its something like "stumbling" on a chemical that happens to give some health benefit or the like), thus, AI can absolutely contribute to new knowledge if utilized that way.

Remember, they are actively using models very similar to language models to try and find protein sequences that can aid in making cures for various ailments, its ignorant to say they can't contribute to the creation of new knowledge (And their contribution is often large then the humans, so pretending that humans have some special edge or there is some unique difference here is naive).

[mystic nest]

think it's about the subject that is wrong. It's never chatgpt or any standalone AI, but human+chatgpt or human+AI.... we are still far from AGI (for now)

[cloud bison]

To be fair, its a bit hard to say, since AI ethicisists and developers for the most part are intentionally constraining the AI and not allowing recursive trial and error developments.

I'm not saying they should be allowing those things, but if they refuse to, and require an initial prompt for any action from a human, then its not exactly the AI's fault that all actions it takes and can take inherently involve a human.

That said, based on the tendency for AutoGPT plus GPT 4 to get caught in recursive loops, I think its PROBABLY safe to assume even if we let them loose, they probably aren't quite there yet... But they are getting very close.

[gentle tusk]

However, the partnership between Microsoft and Epic causes concern among researchers of large language models, partly due to GPT-4's ability to make up (confabulate) information that isn't represented in its data set.🌿

@wicked comet Hello ✍️

[mystic nest]

yeah, isn't that why we still need human in the loop 🤣 to fine tune and make it better. AI is never in a final state

[cloud bison]

I think we're getting close to a point where recursive loops might become history. It will take a clever mix of hard programming alongside fine tuning, training, and system prompts, but I think it comes down to how memory databases are constructed and constraints are placed on token usage.

If we get it right, which could easily take less than a year, then models like GPT 4 32k alone could easily become AGI level. The catch there is that it would still be an absolute energy hog, so refinement is still needed.

The real issue is economies of scale currently don't allow hardware expansion. Also, development has been happening for a relatively short period of time by the measure of human development times, which are Glacially slow compared to the progress of AI (Years, not weeks).

[mystic nest]

things will happen but still reserved on the definition of AGI, which chatgpt 4 is not

[cloud bison]

It can't be agreed.

AGI in the sense we think about it needs the framework to retrain itself, fine tune, order, and reconstruct itself, as well as generally execute tasks. The common definition of AGI needs a pipeline to self improvement in other words.

[mystic nest]

if you say it's some glimpse of AGI = yes, so it depends how loose your definition is