#gpt-models

1 messages · Page 231 of 1

drowsy harness
#

@torpid sphinxIt is obvious that GPT4 will be more suitable.

wraith ginkgoBOT
#

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.

weary dust
#

any idea why only default will identify as 4 when asked what version it is then?

wraith matrix
weary dust
#

interesting. i hope you're right.

ionic tide
#

I think I broke my gpt how can I reset it without loosing my history so it doesnt keep looping and failing to stop

weary dust
#

open a new chat

ionic tide
fading junco
#

Can someone send me a screenshot of the code intupeter plugin

static isle
static isle
#

Browsing had some issues but eventually read it. Link Reader plugin was able to get it quite quickly.

crimson karma
#

Make a alien with god

upbeat ginkgo
#

trying to research historical economic data:

  • chatgpt used to do this just fine, now refuses
  • but other 3rd party chat tools that access the gpt llm work fine

Did my account get throttled or is chatgpt getting dumbed-down bc of resource limits?

ionic stratus
#

China

wraith ginkgoBOT
#

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.

sick trout
#

Getting any data from them (real or mostly "exaggerated") is hard to get.

wise mortar
#

If there is Gpt-5 in the future, people will not have to work hahaha

wraith ginkgoBOT
#
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.
upbeat ginkgo
# sick trout Getting any data from them (real or mostly "exaggerated") is hard to get.

actually, it used to be both simple and accurate (I manually verified). Maybe the logic was changed with the web browsing feature - like, it assumes it needs to use the web brwosing feature for stuff like database lookups, so it blocks it in the core app.

But it works fine with Nova Chat, which uses the GPT API.

And Bard handles this kind of request with ease, but accuracy is still an issue. I think the PaLM model is better suited for analyais.

fading junco
#

Is there any way to get code inturptiter

#

if not when will they be realesing it

static isle
#

We do not know when it will go into beta

wind urchin
#

Does anyone know of a ChatGPT Prompt extension where I can save pre-defined prompts, and selection them as needed. I am after a very minimalist design. I can’t stand most of the extensions with excessive features and UI. Does anyone know of a extension like this?

marble jewel
#

exist any plugin for google colab?

kindred slate
wraith ginkgoBOT
#

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.

ember orbit
#

Hey hey.. any way to get Access to the Code interpretor plugin? 😦

edit: plus User, ofc.

waxen swan
#

Help I dont understand I thought when you signed up for OpenAI you would get free credits to use in Dalle but I never recieved any??? I even used two different phone numbers and accounts but still none on both help

wraith ginkgoBOT
flint dew
#

I am disappointed that the A.I. will not do the lastest as it stops after December 2021 and knows nothing after that date

#

I presume that there are A.I out there that actually work and that this one is showing or just giving us a taste of whats to come in the future

bold cape
dreamy veldt
unkempt plank
wraith ginkgoBOT
#

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

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

serene maple
#

lmao why pay for chat gpt 4 when bing's chat has the same thing

wary badger
#

Same thing? Really?

polar willow
#

There was a problem preparing your chat. Please refresh the page and try again. a_skull

serene maple
wary badger
unreal marsh
#

would it help you more than zapier API?

lost stream
#

sri lanka

quick dune
#

Can't say her knowledge is cutted after 2021 anymore, because she know more than that

quick dune
#

It isn't as creative as ChatGPT GPT4, can't access every link (mostly she just give made up information)

pallid nimbus
stray wagon
#

'

idle furnace
#

i asked gpt4 for som code, it started replying the code but then stuck in a loop, it was replying the same message over and over again

#

i had to stop it

pallid nimbus
idle furnace
#

then asked why the repeated reply "I apologize for the repeated responses. That was an error. Let's continue with the Terraform configuration."

pallid nimbus
#

and when it comes to being in a loop, you need to be assertive and instruct the gpt language model to "continue from where you left off" or wherever it is you need it to continue from.

There should also be a continue generation for website-usage however not sure if it's there for android/iOS.

idle furnace
#

unfortunately i hit some bug

#

i asked again "paste the whole code now"

#

and it starts replying the same code in a loop 🙂

pallid nimbus
idle furnace
#

yes, that wokred, anyway this is a bug, as it should have pasted the whole code when asked

pallid nimbus
idle furnace
#

it is a bug, i asked to have a code for this and this, and it replied correctly but then the reply was in a loop

#

i can share you the conversation

pallid nimbus
#

Feel free to in my dms.

idle furnace
#

yes, shared

pallid nimbus
#

@idle furnace going over your response you gave:

why you response in a loop?

You never clarrified the intention you wanted it to carry-out, you only asked it a question and it attempted to correct itself, you need to be assertive as I said previously and clarify your intention and what you WANT the chat language model to do.

idle furnace
#

but it went in a loop

pallid nimbus
#

A better way to phrase it would be as such:

It seems like you caught yourself in a loop, please continue from where you cut-out the code at the end of the loop.

idle furnace
#

the intenion was clear add a second spoke

#

this is enough for a gpt4 reasoning to understand it

pallid nimbus
idle furnace
#

it understood it correctly but replied in a loop

#

i disagree

#

add a second spoke is enough, should i ask? add a second spoke but don't reply the code in a loop?

pallid nimbus
idle furnace
#

gpt itself confirmed this was an error 🙂

#

for me it's a bug for gpt4, i would agree this won't be a bug but on gpt 3.5

#

for gpt4 one would be sure it does not repeat itself over and over again

#

until stopped by a user

#

now i need to copy and paste from 2 places

#

it cound't concatenate this mesage when asked

wraith ginkgoBOT
#
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.
grim rose
#

Is it just me or has GPT-4 become a lot faster in its response times in the past 2 days? Usually when I sent requests around 2000 tokens, I had to wait 30-60 seconds for a response. Now I get most responses in 5-10 seconds.

magic wedge
pallid nimbus
grim rose
pallid nimbus
#

It'll likely be more for based on complexity of inquiries to response rate which they're on about @grim rose, since it now responds faster, the parameters for logical response can change.

magic wedge
#

you can try the prompt "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?", which is listed on the product page, and it does not give the correct answer.

pallid nimbus
wet drift
#

I had this weird issue where I asked It to list 10 cheapest houses using the Zillow plugin then compare them in a table format and it started perfectly but about half way down it just started to repeate itself over and over again

#

Has anyone had that issue before?

pallid nimbus
#

replacing code for table of course.

wet drift
#

Yes but it didn't just cut out because it was at the end of the generation limit

#

It litterly just kept repeating itself

#

In the first prompt

grim rose
pallid nimbus
wet drift
#

Yeah it's not a big issue just something weird. I was able to fix it by asking it to do those in two separate prompts

pallid nimbus
#

That's likely the best route to go for with prompts especially from knowledge and past issues with the GPT 3 model. It could also likely be because of token output limit as well but not 100% so don't quote me on that.

wet drift
#

Yeah for sure

desert blade
#

We all miss the good old March GPT4model.

magic wedge
magic wedge
pallid nimbus
#

@wet drift have you tested out using more tokens for an output and seeing if it happens again yet or? [just as a form of debugging to isolate where the potential issue could be happening]

desert blade
#

Good old March model (GOMM4) was able to sustain 20+ messages conversation without confusing cars and apples

wet drift
#

What are some interesting use cases people have getting out of it? In terms of having it be able to help you with something meaningful such as your job etc and it giving accurate information because without accurate information its pretty useless lol.

wet drift
magic wedge
wet drift
magic wedge
#

interestingly enough, the version on Poe seems to be an older version and is functioning properly.

desert blade
#

problem is, Poe limit is 1 message/day, then you have to wait a full month to test the limits...

#

(free version)

idle furnace
magic wedge
#

yeah, its quite annoying but also confusing why Poe gives an accurate response while the playground doesn't. (tried the "what is GPT-4" prompt and the time prompt yesterday)

wet drift
#

Chatgot is very interesting because on one hand it can give correct information sometimes but since the way I understand it is that it that it tries to guess what the next word/letter will be in a sentence so there for if you ask it a question and it gets it right. You may be able to ask it the same question again it get a different answer or the same answer in a different wording. However I am finding it not all that useful right now just because its accuracy is very limited. Now maybe that's my fault in asking to vague of questions? Do you think that could possibly be what is limiting my ability to use this tool to it's potential or do you think it isn't yet advanced enough to be useful for someone who making direct enough queries?

coarse gyro
#

They could add the option to set up a seed for every message so you would be able to get the same answer (word by word) by asking the same question but I like it so how it is at the moment.

pallid nimbus
coarse gyro
#

would be thinkable yea.

wet drift
vagrant sundial
#

Hello everyone, how do i access chatgpt 4 ?

#

The free plan is 3.5, right ?

coarse gyro
coarse gyro
wet drift
#

The only reason I say that is so it's not so open as to be able to give you the wrong information because it stated it the wrong way unless you ask it directly to say it a different way.

pallid nimbus
coarse gyro
vagrant sundial
desert blade
#

Hi Valencio, ChatGPT 4 is 20 $/month, but you should wait until they restore the old working model

#

May model is pretty useless compared to free 3.5 (reduced reasonning abilities, reduced memory)

wet drift
#

If it gives inaccurate results then it's a useless tool. Fascinating as it is. (And I'm sure it will improve) I think it's as simple as that.

vagrant sundial
desert blade
#

OpenAI has remained silent or even condescending on the subject.

vagrant sundial
#

Hmmm it seems i'm gonna wait

#

Was about to purchase it

desert blade
#

On my end, I'm canceling my subscription, and I might come back later... or try another provider.

spring crater
#

gpt currently is a trash, no offense, but it is what it is.

wet drift
#

Is it actually learning based on the feedback we are giving it and if it is then how would that even work with you a thumbs down and up. Seems to me you would have to be a lot more accurate with explaining why it's wrong for the AI to truly learn from that without a person looking at the results you got and fine tuning it based on that

desert blade
coarse gyro
wet drift
#

Has anyone tried to develop one with just accurate information and not all the trash with the good stuff on the internet.

#

I suppose that is really what is needed. But I'm also sure that would be very hard to do. To get enough good and accurate information to make something like that without all the trash.

spring crater
desert blade
wet drift
#

Well I mean more of the companies with all the a100 chips doing that.

#

I believe I heard google is trying to make one just for medicine

#

Or the medical field I mean

spring crater
desert blade
#

I suppose OpenAI, for valid reasons, needed to adjust its trade-off between computational power and performance, and the current versions are probably more efficient even if they are less effective.

#

But communication on this topic would have been appreciated.

wraith ginkgoBOT
#

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.

wet drift
#

I would much rather chatgpt run like a snail and give me accurate information if that was the trade off

gusty shadow
wet drift
#

I will say with a bit of prompting I was able to get decent code out of it enough for me to get a program made I would otherwise not have been able to do. So that was pretty interesting.

spring crater
wind urchin
#

what is gpt4 mobile

wet drift
#

It would be very interesting if I would be able to run the code directly from chatgpt so I would not have to go back and forth. That way it would be able to test if the code it gave me use correct and be able to correct itself if the code doesn't work.

gusty shadow
wet drift
#

Not much. Just a little

spring crater
#

sry guys if im harsh, but gpt is beyond me right now

coarse gyro
gusty shadow
#

I'm planning to, but right now there are no liquid assets. 😁

wet drift
coarse gyro
#

I would like if we would be able to react on messages with emojis here. : P

fossil pier
#

has anyone else sense chatgpt4 is responding bit different? I subscripted for a bit over 1 and half months now, it has been extremely good in the past, but for the past 3-4days, the results its giving me is so bad. Sometimes making obvious errors (in my case generating hamming code for a very simple string), giving it 3 or 4 chances and guiding it towards the correct solution and yet it still gives the incorrect answer

gusty shadow
wind urchin
#

Am I allowed to ping everyone

coarse gyro
#

No you are not.

wind urchin
#

Good.

#

I’m working on a development in Roblox.

#

How do I create the BA essential admin?

#

What’s the script?

coarse gyro
fossil pier
coarse gyro
#

that isn't support for roblox. : P

fossil pier
#

do we know that if openai will try and fix this? or they jsut put their time and effort into developing other stuff like cooperating with Microsoft. not only gpt4 giving wrong answers, its also getting slower

gusty shadow
#

I don't feel any difference in the speed of gpt4. gpt3.5 feels faster than it was before though. Both are just a feeling, so could be completely wrong.

coarse gyro
fossil pier
#

yea i guess the speed will vary depending on the prompt i give.

#

Really do hope they fix other issues tho. I can spot some mistakes but really worried at times where it gives me incorrect answers that I can't notice myself

coarse gyro
#

yea sure.. and the new problem your prompt was deleted after entering it. 💀

#

and I mean the continue button is wrong too. :/

wraith ginkgoBOT
coarse gyro
#

Would you be so friendly and delete it yourself? Has no business here.
-> Edit a minut ago: Thanks. :x

#

You could try to do it using Websearch or as plugin web pilot.

#

Thats true, yea.

#

Why do you not want to try with plugins?

real badger
#

I want plugins 💀 I've been waiting since the announcement of plugins

#

like bro it makes chatgpt 2000% better

#

and yes I joined the waitlist

coarse gyro
#

if you want I can try it with the plugins. 🤷‍♂️ (at the moment I have 25/25 messages used so I havve to wait one hour 8 minutes.

real badger
coarse gyro
coarse gyro
#

.. only for gpt-4.

real badger
coarse gyro
#

Hello.

hallow owl
#

If you do one message every 8 minutes you will never reach the cap 🙂

real badger
#

btw I got chatgpt to tell me the date and time

coarse gyro
coarse gyro
real badger
#

yeah

#

I got it to paste the entire page's html

real badger
coarse gyro
void lodge
#

Hey there! If you're facing any difficulties, fret not! We've got the brand new Bing, also known as Chad GPT-4. This AI is incredibly intelligent, surpassing anything ever seen before!

First things first, you'll need to sign up for an account. Once you've done that, you can log in and start exploring its amazing features.

If you're feeling a little unsure about how to get started, you can always check out some tutorials on YouTube. They're a great way to get a grasp of all the fantastic things the new Bing has to offer!

So, whether you have questions, need help, or simply want to explore the boundless knowledge of this incredible AI, the brand new Bing is here to make your experience smoother and more enjoyable than ever before! also this AI doesn't have a limit it is connected to the internet not like Chatgpt has limited knowledge to 2021

hallow owl
coarse gyro
quick dune
#

The ChatGPT GPT4 that connected to internet is better

void lodge
#

bro you're talking about the new bing that is connected to the internet not gp4 gpt4 is connected to the new Bing and bing is connected to the internet bro watch some tutorials you're literally stupid

icy steppe
#

is there anyway to get GPT 4 for free ?

#

legally of course

hallow owl
icy steppe
#

42$/month is a bit expensive for tier world countries :x

tough phoenix
hallow owl
tough phoenix
#

yea lol I pay 20 i was abt to say if they increased it to 40 i’d stop

#

that’s a little insane

void lodge
unreal marsh
desert blade
feral pendant
#

@cunning pike will you guys ever allow reading through an entire project on github?

charred burrow
#

heyy im trying to build a website that helps users to create custom chatbots powered with open ai api

#

can someone help me understand how can i achieve it

wind urchin
#

I would like to do a techno house song as if Mozart had done it.

#

@wind urchin

#

@wind urchin

wraith ginkgoBOT
calm fern
#

Any ideas or prompts to bypass turnitin ai detection?

wraith ginkgoBOT
#

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.

terse badger
#

Is GPT 4 available on the playground, I can't seem to find it? please help thanks

junior flint
charred burrow
#

Can you help or not ?

alpine ingot
#

plese remove the usage cap on chat gpt 4 !!

lofty pivot
#

Whem would I receive invite for gpt4 api 🥲

normal bolt
#

They need to scam more users into buying plus first so they can buy more gpus

knotty grotto
lofty pivot
#

But I am plus user and very loyal to the company 🥺

#

Amy super user here to endorse me 😅

normal bolt
#

Prove ur loyalty, tattoo openai logo on ur forehead

lofty pivot
#

I am working on a personal companion project that I find amazing... But need more context window...

normal bolt
#

Did u fall in love with ur personal companion?

hallow owl
coarse gyro
lofty pivot
hallow owl
lofty pivot
hallow owl
normal bolt
#

Inb4 articles of people going broke due to ai addiction

odd shale
#

Very satisfied with the "Continue" button on gpt-4

wraith ginkgoBOT
#

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.

cerulean obsidian
#

Changes to the API also? it seems faster suddenly

vital hull
burnt steeple
#

or are you also one of the few people that are realizing that the quality is dropping greatly with gpt 4

vital hull
# burnt steeple what makes you think that?

Today I noticed, that speed of answering increased as GPT-3.5-turbo has. Then, asking the question "Are you GPT-4?" I started getting this output:
I am an AI language model created by OpenAI, specifically GPT-3. I am not GPT-4, which, as of my last knowledge update, has not yet been released.

burnt steeple
#

YES, welcome to the team

#

😎

normal bolt
vital hull
#

Is there any news, bug report or something?

burnt steeple
#

that's why i switched over to Bing Chat, as it's got everything excluding Plugins (and in-depth browsing).

vital hull
#

But I'm getting this answer:

Or this:

cerulean obsidian
vital hull
wraith ginkgoBOT
#
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.
normal bolt
cerulean obsidian
# vital hull ?

I meant was did they screw the API also already they have screwed the ChatGPT version, now the API also performing faster and not precise, there nothing I can confirm but just assume, as the speed is blazing fast with results similar to GPT 3.5

cerulean obsidian
vital hull
#

Apparently this isn't the first time, I've read the chat and am convinced that the model has indeed become closer to 3.5, despite the fact that the March version is listed which gives exactly the same answers.

I don't see why that's anyone's problem or not. Is it a bug? What is the reason for this nerfing?

normal bolt
cerulean obsidian
# normal bolt they probably operate at a loss

It would be fair to come openly that they have to downgrade or do whatever they are doing with the models instead being deceptive with the models vs the performance they are delivering, ik they don't owe anyone any answer but even what now is happening isn't good practice

blissful hollow
#

Sup folks. Quick question: I assume the vision part of gpt-4 is still not accessible even in alpha or is there a way to request access for it?

magic wedge
blissful hollow
#

Good to know. Thanks! In the meantime is there some alternative to it? Basically I want a smart OCR that can read the image and extra data from it but I just need specific data instead of parsing everything and then sorting it.

pallid siren
#

Is there a way to know how many messages we have sent on ChatGpt4?

coarse gyro
#

My bad but there is no option for that, no. :/

#

Maybe you could add it with a Browseraddon? (but there I do not know some working addon for that)

magic wedge
vital hull
magic wedge
vital hull
#

I listed the topic and read some issues that people were facing the same problem

magic wedge
#

(wasn’t internet access either, I use the API version which doesn’t have that)

blissful hollow
vital hull
#

Maybe this is how they are revoking access to Gpt-4? I just don't understand the reason for such a spotty withdrawal.

magic wedge
#

a dev on twitter claimed that there were no changes since the march release, but that’s just a blatant lie.

magic wedge
vital hull
magic wedge
vital hull
#

At least I hope this is a temporary effect and that after the release of the 32k version or something else they will start fixing it

wet drift
#

Should Chad GPT be able to analyze a cities structure and devise a plan for a hypothetical rail system that would be efficient as in where its built in the city. Or would they use another AI for something like this

#

I'm just trying to think of ways this could be used for benefit society rather then just something to have mild fun with

rocky pewter
cerulean obsidian
magic wedge
normal bolt
#

we also should make ChadGPT and fine tune it on red pill reddit

static isle
#

Should?

normal bolt
#

wait it exists?

static isle
#

💀

wet drift
#

Yeah I get what you mean but are there any real benefits to chatgpt right now other then for amusement. Or would we have to wait longer for somthing better

magic wedge
cerulean obsidian
silver olive
#

Anyone using GPT for real estate here? I am in the mortgage industry and wanted to see if anyone on my side of things is using it for good results?

wraith ginkgoBOT
#

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.

opaque knot
#

heyo everyone

#

im trying to sum up text which is 10k words long, but it always says that its too long. GPT-4 in my case can barely sum up 2k words, while 25k is being promised. I have GPT+ sub for 20euros

#

whats the deal ? am I the only one with this problem ?

supple violet
#

where is 25k promised?

charred spear
#

It's not.

#

Don't provide all text at once.

supple violet
#

the tech announcement said it can handle up to 32k tokens which is close to 25k words but it was not said to come with plus

charred spear
#

Not sure it would scale that easily. It's barely available to GPT4 API developers.

#

It costs a lot to run the models and there was a GPU shortage at some point.

#

Will be interesting to see how the new supercomputers go...

opaque knot
#

the funny thing is gpt3 was doing like 4-5k easily

charred spear
#

You say GPT+sub which means you're talking about ChatGPT right?

opaque knot
opaque knot
#

im new to this

supple violet
opaque knot
#

done

charred spear
#

You can't just paste in that much content in the chat box. I believe it has a character limit.

supple violet
opaque knot
#

whats the limit ?

supple violet
#

"ChatGPT Plus subscribers will get GPT-4 access on chat.openai.com with a usage cap. We will adjust the exact usage cap depending on demand and system performance in practice, but we expect to be severely capacity constrained (though we will scale up and optimize over upcoming months)."

#

they also said image input, etc. will not be available

opaque knot
#

what is the most efficient way to summarize text and pinpoint main problems then ? would really appreciate help

opaque knot
charred spear
#

No.

#

It's because it's quite a technical product to explain.

#

What text are you trying to summarise?

static isle
opaque knot
#

pretty much a book, but Im doing it by chapters.

#

Its for my studies and im in kind of a hurry :c

charred spear
#

How big is it?

opaque knot
#

word wise?

charred spear
#

Yeah

opaque knot
#

sec

#

from the part that Im assigned to do, theres like 20k words

#

Gotta make a presentation from it

charred spear
#

Copy it into ChatGPT 2000 words at a time.

#

Each time ask it to generate study notes etc.

opaque knot
#

okay i see

#

Theres 1 trouble tho, it says i've reached my limit (for today?/24?). I have made only like 3 succsesful requests with GPT-4. Every other was an error

charred spear
#

You're using GPT4 for summarisation?

opaque knot
#

Is ChatGPT also available with +sub ? or is it another type of subscription

#

Im totally new to this

charred spear
#

Use GPT3.5

#

Don't use GPT4 for Summarisation.

opaque knot
#

may sound like a dummie xD

#

okay, gotchu. For what is it best then ?

charred spear
#

GPT4 is a technical AI

#

It has higher reasoning and is trained with images.

#

GPT3.5 is quicker as it's designed for tasks that don't require too much thinking.

opaque knot
#

could you give a few examples ? Its really intresting and i'd love to get into it :Dd

charred spear
#

Ask GPT4.... lol

opaque knot
#

cant do

#

capped xD

charred spear
#

Ask GPT3...

#

LOL

opaque knot
#

my brain is burning lmao

charred spear
#

I'll ask it...

opaque knot
#

2nd sleepless night hurts

hallow owl
charred spear
# 
The conversation suggests that the user was encountering errors when attempting to use GPT-4 and was informed that it is currently capacity-constrained. Therefore, it is recommended to use GPT-3.5 instead, which is available for ChatGPT Plus subscribers.

To summarize text efficiently, especially for a book, you can try the following approaches:

Break the text into manageable chunks: Since ChatGPT has a token limit, you can divide the book into smaller sections, such as chapters or paragraphs. This way, you can process each chunk separately.

Utilize summarization techniques: Within each chunk, you can use techniques like extractive summarization, which involves selecting important sentences or phrases that capture the main ideas. You can ask ChatGPT to generate study notes or summaries for each chunk.

Leverage contextual understanding: To enhance the quality of the summary, you can use projects that employ embeddings or vector databases to provide better contextual understanding of the text. This can help you pinpoint the main problems and extract key information more effectively.

AutoGPT, mentioned in the conversation, is an example of a project that utilizes vector databases and can be helpful for summarization tasks.

Keep in mind that even with these techniques, summarizing a book can still be a time-consuming process. It is advisable to allocate sufficient time for the task, considering the volume of text involved.

Regarding the examples of tasks suited for GPT-3.5 and GPT-4, the conversation mentions that GPT-4 is designed for more technical tasks and is trained with images. It has higher reasoning capabilities compared to GPT-3.5. On the other hand, GPT-3.5 is faster and more suitable for tasks that don't require extensive reasoning. However, without direct access to GPT-4, it's not possible to provide specific examples of its capabilities.

In summary, for summarizing text, it is recommended to use GPT-3.5 and divide the text into smaller chunks while applying summarization techniques to extract key information efficiently.

This is GPT3.5

opaque knot
hallow owl
#

25 messages per 3 hours is one every 7.2 minutes

#

sending one every 8 minutes ensures you never hit the cap

charred spear
# 
GPT-4, as suggested by Jonno, is a more advanced AI model than GPT-3.5. While GPT-3.5 was purely a language model, GPT-4 has been trained with both textual and visual information, which enhances its ability to handle a more diverse set of tasks and improves its understanding of context.

As Jonno suggested, GPT-4 might be better suited for more complex, reasoning tasks. These might include tasks that involve understanding the context behind images or making connections between visual and textual information. For example, it could be used in the analysis of visual data, such as recognizing objects in images, or understanding the context and meaning behind a photo accompanied by a caption.

On the other hand, GPT-3.5 can be quicker for tasks that require less contextual understanding or reasoning, and it is quite proficient in understanding and generating human-like text. This makes it a great tool for a range of tasks, from writing essays or articles, summarizing text, or generating creative content like stories or poems, to answering factual questions or providing explanations on a wide variety of topics.

To sum up, if you're looking for speed and efficiency in tasks that mainly involve understanding and generating text, GPT-3.5 could be the better choice. If you're dealing with tasks that require deeper understanding, more complex reasoning, or the interpretation of visual data, GPT-4 might be more suitable. It would really depend on the specific requirements of your task.

As for your current predicament with summarizing the text for your studies, if you're finding that you're running into limits with the GPT models, you could try breaking the text down into smaller chunks as Jonno suggested. You could also consider using other summarization tools or methods in combination with the GPT models to help you process the information more efficiently.
#

That's GPT4 - you can see the difference in what it says.

opaque knot
charred spear
#

And since you're new. Here's a tip. It's a language MODEL

opaque knot
#

youre a saviour ong

charred spear
#

Mathematical models can misinterpret and provide incorrect values.

hallow owl
#

GPT sucks at maths without plugins 🙂

charred spear
#

If you send me the document I can use my system...

opaque knot
#

sleep deprivation makes me so dull its almost shameful lol

charred spear
#

No

#

They're language models with different training.

#

GPT4 is more capable.

opaque knot
#

ok so its for complex stuff

charred spear
#

If these models didn't have utility they would have been scrapped - but there are many companies still using Ada, Davinci etc.

#

Yes. GPT4 can write code.

#

GPT3.5 can actually write code very well too, but it requires more understanding from the user.

opaque knot
charred spear
#

@opaque knot sent you a DM

true mist
#

what’s up fellow bard ai gamers

wet drift
#

What are the main benefits of having access to plugins omfor chatgpt? I get that you have more access to information but it seems like you would be able to get the same information by just going to those sites so what is the benefit?

true mist
fickle gale
#

time that u will never get back in your life, unlike money?

normal bolt
#

Soon money will be replaced with social credit scores that is determined by AI, the better civilian u are the more access and purchasing power

wind urchin
#

that sure is a shame

wet drift
#

I just wish it could do more with the data it's given

normal bolt
#

I wish it could do all my household chores

#

And also my job but be acting as an agent for me and not my company so I don’t have to work but still get paid

wet drift
normal bolt
#

This is bettering society

#

We all just get to relax and earn money and all our necessary chores and demands are taken care of for us

true mist
fickle gale
normal bolt
#

We shall put in legalisation that for every ai used the company has to hire a human behind it who receives the wages

visual knoll
#

is GPT 4 still have the 25 prompts / 3hrs or its unlimited now ?

static isle
#

Still limited

visual knoll
#

how much ?

static isle
#

Same amount

visual knoll
#

25/3hr ?

#

tyy

static isle
#

👍

crystal field
#

gpt4 seemingly has some built in logic or training which detects the random wordchains associated with high temp/penalties

#

"Apologies for the apparently random sequence of words that occurred in between the explanation. Your focus should be the subject discussed in detail proposed secondary overwhelming indicators peripheral organization enterprises frameworks am transparent feedback-contributing configuration configurations highlight problematic "

#

or maybe it's a -chat feature as i've only really explored this with standard davinci/codex models, but v3 doesn't seem to generate this self-aware output. And this message comes up mid-stream

unreal marsh
#

is gpt4 writing faster now?

static isle
#

Since it was first released, yes

unreal marsh
#

i mean compared to last week

#

or even yesterday

static isle
#

Speed may change due to server load

crystal field
#

some say it's been nerfed

#

i usually autosubmit queries to the api now in parallel to check them

wind urchin
#

how to create an url of a picture with gpt4

hallow owl
full pebble
#

Do you guys know if i can fine tune GPT4 either from OpenAI or through some other software?

hallow owl
full pebble
#

Thanx @hallow owl .. I am readin those, and I dont see the option there, davinci seems the strongest one available for fine tuning thats why i asked here maybe you know something else!

hallow owl
full pebble
#

Good to know thanks!

wraith ginkgoBOT
#

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.

sturdy stratus
#

anyone know which web AI detector is closest to "turn it in's" to use for free, so I can check if my text is edited enough before submitting to turn it in

vestal oar
#

You can use copyleaks

#

.com

#

It's much more fiable than others AI detector

#

and it's free

weary dust
#

do you think they'll ever merge default, plugins and browsing? or will they forever remain seperate?

weary dust
spring estuary
#

😁

chilly swift
#

Wow, that's some overactive automod action. Anyway, without the reference to a certain category (not an example of the category, just a perfectly innocent reference to the title of a Tom Lehrer song from "That Was the Year that Was"), my comment was that 8k is pretty limiting for getting fanfic written.

wraith ginkgoBOT
crimson topaz
#

🤣 I don't know if we should trust someone with only question marks as their username.

fringe plinth
dry stratus
#

where can i talk with GPT-4

static isle
dry stratus
#

@static isle Could i talk with it on this server?

static isle
#

There are no bots that have access to GPT models here

crimson topaz
dry stratus
#

Whhhhh thank you!

static isle
#

👍

crimson topaz
#

The bots are going crazy in the servers I see. However, I was wondering if people feel GPT-4 was better like 3 weeks ago, or if you think it is more improved now?

static isle
#

Though there are not that many instances of concrete proof, e.g. direct comparisons between exact prompts or a statement from OpenAI

#

Many believe that GPT-4 has decreased in efficacy

#

The future will reveal what there is to be

crimson topaz
# static isle Many believe that GPT-4 has decreased in efficacy

This is what I am noticing. I realize that they want people to use it for the right reasons. But, it is turning into Google on steroids. I was getting far better reasones and help (scripting) a month ago. But now with the "not updated since 2021" I am questioning it now. I want to see what is to come, especially because they were the game-changers, but now other are only using it to learn from the mistakes and not successes.

ionic shard
#

Hey., question..... I have had trouble all day accessing CPT4, even as a paid subscriber. Is there something I am missing????

wind urchin
earnest sonnet
#

When is gpt 4 available for free users?

knotty grotto
kindred girder
#

HumanLoop published OpenAI plans but that blogpost was taken down due to OpenAI cease and desist which means its true. its archived now and reads:

  • OpenAI is too GPU limited which is delaying alot of their plans
  • 32K context window cannot be rolled out to people due to being GPU constrained
  • they can extend to 100k to 1 million context window but its not possible with GPU contrain
  • Finetuning of API is not possible due to GPU constrain
  • Dedicated capacity is not possible unless customers pay $100,000 upfront for GPU cost

2023 roadmap June to December

  • Cheaper and faster ChatGPT4
  • Longer context window up to 1 million
  • Finetune API
  • Stateful API

2024 Roadmap

  • Multimodality. OpenAI does not have capacity to offer Multimodal due to GPU

  • ChatGPT plugins thru API are not possible yet

  • OpenAI is not building products beyond ChatGPT

  • Regulation is needed but so is open source

  • Scaling laws still hol

wraith ginkgoBOT
severe lava
#

Hi,
I've purchased the Plus subscription for my account, but it shows the error message:
"This subscription is associated with another OpenAI account..."
Could you please help to to solve it?

wraith ginkgoBOT
#
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.
agile lodge
#

there is no api for gpt4

livid ginkgo
#

has anyone noticed chatgpt is a little dumber in the last 1-2 hours in regards to programming in javascript? i understand it has limited memory and may forget the context of my convo with it.... when it does i just rejog its memory by repasting a few of the initial prompts and explain the context. i've been doing this without fail the last 2 weeks.
However it now seems to be giving really bad answers that does not solve my coding prompts. i'll explain the context, feed it my app.js file, and its stuck in a loop of giving me the 2-3 same troubleshooting approach. and all of a sudden doesn't remember to stick to my rules such as giving me the entire code to copy and paste for the js file, etc.

inner trail
small sorrel
#

I have noticed the same thing, but not in context with Java Script. There seems to be a problem with the message stream. It does not remember earlier messages at all or not as well as usual.

thorn yacht
#

/subscribe

astral igloo
#

What are the differences between gpt4 and gpt4-32k ?

static isle
#

gpt-4-32k has 32k tokens of context available

velvet knoll
#

Can I get a us phone number

#

To get plus subscription

#

I'm from srilanka

static isle
#

gpt-4 in ChatGPT only has 4096 tokens of context atm

velvet knoll
#

They don't let me get plus

radiant shard
small sorrel
static isle
small sorrel
#

Ok, thank you!

#

Soo, how do i access this Gpt-4 32k?

obsidian thicket
#

I just told gpt4 great job for the nice haiku it wrote me, and it replied with a thank you haiku lol

# 
Thank you for your kind words,
In the realm of verse, it's you I've heard,
Another haiku, give the word.
cerulean obsidian
#

What is the payment cycle for API? Ummmm it's been 2 months and they didn't generate any invoice for me, I'm confused

#

'You'll be billed at the end of each calendar month for usage during that month.' but there was no invoice or billing itself for past 2 months 🤔 as I had a billing last month and the month has already ended but no invoice yet

fresh stratus
#

@cerulean obsidian Why are you worried u don’t have to pay

cerulean obsidian
fresh stratus
#

But are u not happy I don’t have to pay?

cerulean obsidian
fresh stratus
#

Ok

ocean solstice
cerulean obsidian
wraith ginkgoBOT
#

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.

simple breach
#

Is Gpt4 broken? I am in a chat thread with ScholarAI plugin, I have been trying to write an article about something. I finished up to the results, and have provided everything to the thread. But, for discussion, chatGPT seems to continuously provide irrelevant information, nowhere near anything I have provided

crisp belfry
#

is the free one gpt4?

hollow compass
#

when is image prompting coming

wraith ginkgoBOT
unreal marsh
#

what is the difference in performance between using backtracking with the API, and just saying that prompt with 3 experts before the question?

dusky yoke
#

11

shadow scarab
#

N

normal bolt
#

lmao

little vine
#

1

#

where am i?

wraith ginkgoBOT
#
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.
onyx grove
#

is free to ask questions?

willow crane
#

Hi, since the end of May I couldn't use GPT-4 while I am a plus user and still within the plan period, GPT4 isn't responsive since then while GPT 3.5 is working fine, I cannot find any place to report the problem or to find any costumer support, I wonder if I can find help here ?

mystic talon
#

hello, I've been waiting to get access to the code interpreter feature for a very long time. I am living in UK, I made multiple request to get access to this feature, what can I do to get access to that feature please? Many thanks

supple violet
#

you wait till it will move to beta phase

wind urchin
#

guyz

grave vale
#

gpt4 down?

mystic talon
grave vale
supple violet
void lodge
#

azine

mystic talon
supple violet
#

you cant, they closed the waitlist from non plugin devs

mystic bramble
#

✓

#

Gomm™

#

Gomm®

#

@void lodge slm

cerulean obsidian
#

anyways the ChatGPT 4 is nerfed so I dnt expect much performance from it

supple violet
vocal gulch
#

Alright, looks like GPT is not fond of formatting ASCII art.
I asked it to add a linebreak at the end of line each. GPT3.5 appended \n after the significant characters instead of the end of the line, GPT4 took three minutes to type out the first line, added a full stop, then gave up lmao

broken ivy
#

When can I use Chinese?

quaint jungle
#

is there a way to access pluggins on the mobile app?

frozen veldt
#

the language or do you mean gpt in china?

wraith ginkgoBOT
knotty grotto
vagrant plinth
#

So the cap is on your message sent to GPT-4, not the response GPT-4 gave, right? Saw a "Continue generation" button popped up and I got my self wondering about this

knotty grotto
vagrant plinth
wraith ginkgoBOT
#

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.

wind urchin
#

gif

inner prairie
#

Hi chat GPt

knotty grotto
simple urchin
#

Just bought gpt 4 how do I get to the plugins?

#

I enabled the plugins beta

#

Oh you gotta hover it. ew

wraith ginkgoBOT
#
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.
spare quarry
#

@wraith ginkgonihao

worthy girder
#

asking gpt-4 to provide some low/no-code, open source, self-hosted solutions for building a website; the reply: rocketchat, mattermost, jitsi, discourse - yeah, this is really not the answer I was expecting 😦

balmy sinew
#

Why isn’t there time stamps for chat history? It would make looking back on conversations soooo much better -_-

coarse gyro
frozen veldt
#

ask yourself what you're expecting before you submit the prompt!

For example - if I ask it for a baking recipe it will give me one but oops I forgot I'm at high altitude so I have to ask for revisions... eats up prompts

Instead now I prompt it to "Give me a High Altitude recipe for X" and it integrates all of the given requirements into the output

#

Another tip for y'all. If you hit your 3 hour window and don't want to lose the context, leave the browser open and resubmit your prompt when the box changes to green. Otherwise you might as well start a new conversation because it forgets most of the context when it resets

acoustic wasp
frozen veldt
#

that's my experience with it while working on coding. If I come back or reload it seems to lose a lot of ground where as keeping it open seems to preserve info

#

I'm not an expert though just my observations

charred burrow
#

Its a whole project

#

I want the code or how that is done

#

Can anyone help

#

Im trying to build a website on which the user can create as many as chatbots as it wants by providing a context and the chatbots are powered by gpt-3

frozen veldt
#

Ambitious - get with the gpt-4 code interpreter maybe it's within ability

#

you could also look at microsoft azure and hosting on the cloud

#

way outside my understanding though - good luck

charred burrow
queen rock
#

Ok

frozen veldt
# charred burrow Can help me as well ?

I've learned with my own project that you have to bootstrap your own knowledge. If you can't get it working - don't expect others to. That or hit fiverr or another coding site and put up a request for bid

slate coyote
#

ey guys, do you have any idea on how to stop gpt from doing In conclusion or in summary after a short text? It ignores my prompt e.g.

Craft an informative, long-form but still engaging text about '{item}' for a blog post on the topic of "{keyword}" without concluding with a summary.

any idea?

frozen veldt
#

just omit the conclusion paragraph?

#

you should be editing the output anyway

worthy girder
# frozen veldt That's prompt engineering in a nutshell.. you gotta be specific about what you w...

I understand what you're saying and you're right, prompt engineering is important ...but in this case, it's not about prompt engineering, it's about how gpt-4 was nerfed. My question was actually pretty specific - low/no code, self hosted, etc - and one would expect a "decent" answer for this specific prompt, just that some weeks ago there were changes made to gpt-4 and the quality of answers has dropped significantly

slate coyote
#

yeah but its annoying that I have to pay for that part if I delete it anyway

frozen veldt
#

I disagree on the ascertation that gpt-4 was nerfed

#

new guardrails exist sure but capability is a constant

#

I haven't run into any issues with the code interpreter. Caveat - performance does seem to drop during high traffic times which is the CPU constraint they're hitting taking effect as they have to throttle

#

I do most of my coding work during off use hours

worthy girder
#

this was just the last example of "wrong" answer, I've been experiencing this almost constantly in the past couple of weeks

frozen veldt
#

because it sounds like it gave you what you asked, a list of free resources

#

did you want it to build the thing? thats a very different ask

worthy girder
#

sorry, but have you read my prompt? it was pretty specific

frozen veldt
#

I did and while it may have seemed specific to you, the output you got was right in line with what I would expect to get back

#

it sounds like your expectation of result is out of line with your ask, leading to the dissatisfaction in the response content

worthy girder
#

and again, I'm well aware of prompt engineering, I built an app with over 150 (tested) editable prompts

novel cargo
#

any change of getting this damm 25 message cap removed

#

run into it so often

frozen veldt
#

GPU constraints are documented

scenic coral
worthy girder
#

how is chatrocket, jitsi or mattermost - chat software - a good answer for a prompt asking for site building software!?

frozen veldt
#

No idea, I'm not going to doublecheck the work

#

but did you ask it any refinement questions?

#

If you're expecting the right response 100% of the time on the first ask you have to be exceedingly careful about your prompt. It almost never works out perfectly with the browser modules

worthy girder
#

these are chat solutions, not for site building - again, the answers are completely wrong

frozen veldt
#

its not allowed to be genuinely intelligent. It is quite capable but it is hard coded to have limitations

worthy girder
#

and if you haven't encountered/noticed the dropping in quality it doesn't mean that others haven't - there are multiple threads discussing this, here and in other places

frozen veldt
#

so you have to be specific like this:
I am looking for a web site building service that is free to non-commercial users. Please produce a list of sites that have integrated web building features and offer free to use access.

narrow gust
#

I've got 63000 characters that I want to input to gpt, is there a way to have it segmented and remembered so once I'm finished giving it to gpt,I can then say - let's work on page 6, revise, edit, etc etc. Now let's work on page 10, etc etc...

worthy girder
#

this is not simply a prompt engineering topic, it simply provided a wrong answer for a very specific question

#

do you have any source on this? afaik people complained about this on hacker news also

frozen veldt
#

because you were not specific

I used the prompt I provided above into the bing chat engine and got this back:

There are many free website builders available for non-commercial use. Here are some of the best ones:

8b - 8b is free for both non-profit and commercial sites. It has a user-friendly website editor that allows you to create something unique and beautiful without spending a single dollar.

Canva - Canva’s free website creator lets you make your own website in under an hour and optimize it for desktop or mobile. You can start from scratch or work right away with a customizable template from their collection.

PCMag - PCMag recommends several site-building services that deliver useful features and dependable performance for free.

Website Planet - Website Planet has tested and tried several free website builders and recommends the best ones.

I hope this helps! Let me know if you have any other questions.

narrow gust
#

Make the Google doc public, then using a gpt web plug-in to read it?

frozen veldt
#

its been a fun chat I hope you found it informative andsto and good luck

worthy girder
#

multiple threads here; hacker news; google search for "gpt-4 quality drop"

narrow gust
#

Right but if I make it public and share with no one and my inet connection is secure, it's not really at a huge risk, right?

#

I'm not exactly using sensitive data so I'm not entirely concerned tbh

#

I didn't know Google doc links can be read by gpt. I've had tk upload to my website and provide urls to a pdf so far

desert blade
#

I was not allowed to post this link : site/openai-comments-on-reports-of-gpt-4-performance-loss/ where you replace "site" by "the-decoder".".com"

#

These guys have gotten so full of themselves with their skyrocketing success that they don't realize that even us, with limited understanding, can still detect condescension.

balmy sinew
#

You know what’s weird, if I just word vomit my prompt into the most detailed, but unfortunately unorganized prompt possible, I’ll sometimes get it working on the first try, usually have to tweak the prompt.
But if i instead ask gpt to rate and edit my prompt to get the point across better etc, the prompt it produces gets me a perfect code on the first try. So far it’s 3 for 3 on pretty complex api integrations. It just weird to me how it can tell what I want obviously, but because of the structure, it’s has completely different results.

wind urchin
#

Can I use gpt-4 to create and/or manage complex spreadsheets directly

worthy girder
#

I read the article and the answer from the mentioned OpenAI employee seems pretty "hand wavy" (to me)

#

what is more weird to me is that some users notice the difference, while others don't - makes me wonder if there's some a/b testing going on

worthy girder
quasi tiger
#

is there a way to change a conversation back to 4 after it defaults to 3.5 after reaching the limit?

desert blade
wind urchin
#

@wind urchin Answer the following questions using this information:

Normal body temperature is 37°C
For the first 12 hours, the body loses 0.78°C per hour.
After the first 12 hours, the body loses about 0.39°C per hour.
Determine the approximate time of death using evidence from algor mortis if the body temperature was 31.3°C?

Hint: Round to the nearest hundredth place (2 numbers after the decimal point)

Step 1:

The total degrees that the body lost is:

Body's Initial Temp Body's Ending Temp Total Degrees Lost
Answer
°C - Answer
°C = Answer
°C

Step 2:

The body has been dead for Answer

12 hours.

Step 3:

°C The Body Lost Rate the Body Loses Heat PMI
Answer
÷ Answer
= Answer
Hours

Conclusion:

The body has a post mortem interval of Answer
hours.

civic sparrow
wind urchin
#

where is chat

#

gpt at

worthy girder
# civic sparrow Prompt engineering should always be the solution to get better responses. The AI...

you're right, LLMs don't replace logical thinking (and using your best judgement) but this is not about logical thinking, I'm well aware of the limitations of LLM's (and I'm doing my part in making others aware of this); but I disagree that 'prompt engineering' is the solution; if the LLM's hallucinates or has a limited context window and generates one day one set of results, the other day a different one, it has nothing to do with prompt engineering

civic sparrow
fast timber
#

Will Gpt 4 ever be free to use?

tacit imp
#

Does anyone else get an error "There was a problem preparing your chat. Please refresh the page and try again."? I get it all the time when I try to submit question to GPT-4 and my messages just disappears and website itself (not chatgpt) shows the error above. I am a plus user and GPT-3.5 works fine.

knotty grotto
distant bison
bright oyster
wraith ginkgoBOT
ember olive
#

Does having GPT-4 in the playground (chat) means an API access has been approved?

distant bison
#

I should defintiely let the other channels know!!!

unreal marsh
#

How can I become an alpha member?

ember olive
#

thank you

narrow gust
#

Anyone have favorite gpt plug-ins for reviewing zoom generated transcripts?

coarse gyro
#

no but you could use whisper + normal gpt-4 to summarise the meeting/ask questions about it.

wraith ginkgoBOT
#
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.
cyan sparrow
#

When I send a message to GPT-4 and press "enter", the message isn't sent and the text box just clears. I tried clearing the cache multiple times in addition to trying it on other browsers but nothing helps. GPT-3 works fine however. This is a recent issue and I don't know what caused it.

supple violet
cyan sparrow
supple violet
cyan sparrow
supple violet
#

then maybe some external app breaks them like vpn app with security extensions, virus protection etc.

#

nordvpn and avast are known offenders

narrow gust
coarse gyro
#

No its another AI from OpenAI. (To transcript a video/sound to text)

lament sphinx
#

is there a way to make chatgpt read links?

wind urchin
#

Bing chat is GPT-4

echo vortex
jagged acorn
#

whats going on with gpt4 on the openai website? having issues? giving out wrong messages, cannot remember from one message to the other.

junior flint
#

gpt4 is a lesser version now because.. at what cost openai allowed new countries to join in the usage of gpt and not to mention new features introduced..

tribal bay
#

GPT-4 was generating a prompt, then cut a word in half, added a # and started to explain what its doing at the moment (it was browsing the internet) 🤔

junior flint
#

i have stopped my plus subscription and testing gpt4 api from day to day for performance improvement until i see it's getting back better i won't subscribe again for plus sadly 😦

arctic mirage
#

hello friends, I just bought gpt 4, but it can't give me updated information, the answer is as follows

#

As an AI, I was last trained on data up to September 2021, which means my knowledge is up to date until that point. I don't have the ability to learn or update my knowledge after that cutoff. Therefore, I may not have the latest information or developments that occurred after September 2021. For the most current or specific information, it's always best to check with reliable and up-to-date sources.

#

why is that

wise trout
#

You need to go the settings->Beta features, and enable these features: Browse with bing and plugins

knotty grotto
#

and you can't change that. you can use browser option if you want it to search with internet

#

but the daat trained wont change

arctic mirage
knotty grotto
#

training data take years

arctic mirage
arctic mirage
#

then how is gpt 4 different from 3-5

knotty grotto
wraith ginkgoBOT
vagrant citrus
#

Is it just me or is ChatGPT 4 getting more obtuse?

#

Perhaps someone will do a study to see what percentage of its tokens is babbling endlessly about some objection to the prompt rather than on answering the question.

tribal bay
#

I just asked it to analyze a text and it did a pretty low-effort analysis
though im having a political debate with it and it makes some pretty good points or perhaps im a midwit, or both

normal bolt
#

just start responding to it as a human being, give it taste of its own medicine

vagrant citrus
normal bolt
#

lol really?

cobalt spire
#

Hey guys, my chat automatically switched to GPT3 since I ran out of credits, now that I’m able to use GPT 4 again how do I switch my chat back to GPT 4?

knotty grotto
tribal bay
#

we have limited amount of credits ?
per day, per month or per 3 hours ?

#

or is this purely about the API stuff ?

knotty grotto
tribal bay
#

👍 thanks

muted forum
#

Ummmmmmmmmmmmm, tf?

Me: Top three two-door cars under 200k SEK. Only answer is required

ChatGPT:
2023 Volkswagen Polo
2023 Toyota Aygo
2023 Peugeot 208

digital valley
#

Thank you GPT-4, I landed a software engineer job with no coding experience! $20 a month for a $200k salary job wow

normal bolt
#

Lmao

livid peak
#

is paypal not supported for ChatGPT plus?

pallid rock
wary socket
#

Yoo, does anyone know the fundamental of gpt prompting, i want to makes rules for it to follow, but when gpt reads it, does it read the whole text, or does it do 1 rules by 1, For example

Rules 1. Go find a keyboard
rules 2. show me the price
rules 3. Give me all the details

Does it follow 1 by 1, or does it make a conclution from the whole text, does my question make sense? i want to learn prompting good

gloomy olive
wary socket
vagrant plinth
# knotty grotto Yes every prompt = -1 of your 25 msg Do 1 msg per 8 min to avoid max

Later found out that the cap is actually not that noticeable, since I use it to code I need to make sure the code is running fine... So 8min for 1msg isn't that bad, I might not even complete the debugging or some modification within 8 minsa_skull
Also I just asked a lot of question to gpt4 before bed to make use another 3 hours cap session. Dumping the questions and reading the answer during breakfast seems great... ALTHOUGH I JUST CAN"T STOP READING IT RIGHT NOW

steady lake
#

When will image prompts come out from a preview?
I've already used images like at the very beginning when some random discord supported them for free through the API
They worked really well so why isn't it available?

wind urchin
#

what color is the chatgpt 4 icon? I am subbed but i lost the option to switch model after i refreshed my page

#

no longer access to plugins as well.

scarlet perch
errant shard
ivory axle
#

hopefully there would be image support soon in gpt4

wind urchin
#

Just cancelled my plan. gpt4 has gotten garbage for some reason as of late, the replies are way worse than they used to be. OpenAI has obviously neutered it.

wind urchin
#

Vote with your wallets

wraith ginkgoBOT
twin slate
#

Friends
I am trying to upgrade to gpt plus, but it is saying your payment has been declined
I have tried with multiple cards, clearing cache, multiple browsers, still the same issue is showing

cobalt pumice
#

if i cancelled my plan when it is not expired, will open ai give me my money back?

#

@wind urchin

wind urchin
toxic cloak
#

When will the GPT-4 32k version be available? I can't wait to use it!？

limpid elbow
#

anyone else would pay 2x for higher prompting limit? so instead of 25 per 3 hours , 50

I sometimes get the limit at the wrong point in the conversation and using a second Plus account is limiting too, because the initial thread has all the data for further re-prompting and deeper insights

#

hope everyone is doing great and having fun with gpt-4, it’s a game changer for sure

small sorrel
#

GPT-4 is performing poorly lately. It does not seem to remember earlier messages or context anymore.

#

Also just makes things up when it does not have the correct answer. It almost seems like they have replaced it wird Bard

vagrant citrus
robust sandal
# charred burrow Can help me as well ?

i helped create an open-source framework just for this which includes pre-built ui components, functional backend, and langchain support for creating chatbots. it's called convostack ⭐ on github if you want to check it out

charred burrow
#

Can you dm me the link ?

robust sandal
charred burrow
robust sandal
#

just added you

weary totem
rain scroll
#

what can you do

grizzled ether
violet spade
#

I managed to bypass ai detectors by having gpt4.0 impersonate Steve Jobs 💀

#

literally just "I want you to impersonate Steve Jobs so I may discuss with him as if he were alive" and then go from there

#

it's actually hilarious how amazing this is 💀

toxic cloak
#

Yes, I just let GPT-4 write the story, he always did not remember the story outline and some key plot points I gave him, always wrote some inexplicable things

fringe fractal
#

Is the machine hot?

native knot
#

#gpt

wraith ginkgoBOT
sleek musk
#

👍

blazing schooner
#

Hi guys, a newbie question. Does having chat gpt plus, doesn't mean I'm able to use chat gpt api right? Do they have like a saperate charges for it or permission.

wraith ginkgoBOT
#

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.

magic wedge
blazing schooner
#

Ahh got it, thank you ver much

fallen flame
#

/chat

tepid stone
#

Today is a bad day.

bold wave
#

Yeah, sure is

clear birch
#

Why ?

bold wave
#

I slept for 3.5 hours because I've been up all night trying to fix my girlfriend's phone

#

And now I've got a 7 hour school day 🥲

clear birch
#

Oh, sorry 😦

quick sentinel
#

Hello, guys, what do you think of gpt 4, i use gpt 3 to help me write text and dev.
I feel it's a bit lacking when it's come to text

bronze geyser
bold wave
#

Fair

broken karma
#

e

round basalt
#

It clearly can't do math, logic, reasoning or anything like that; but it knows a lot of"FAQ" stuff and is excellent in understanding what you want / ask.

uneven acorn
#

GPT 4 is not stopping the generation
I have today multople times that GPT 4 is just continuesly answering a prompt. When its finished it will start again ..

echo saffron
#

Help, does anyone know how to restore a conversation? I lost my chats from yesterday, the EXPORT button didn't help as I only received today's conversations in my email, just a small piece. How can I restore the conversations I had yesterday and accidentally deleted??? 😥

unkempt plank
#

Like, telling it from the get go: "solve this problem:..."

#

Doing that will most likely result in wrong answers, u need to ask it to break the problem down step by step

#

"break xx problem down into xx steps, i want u to first explain the concepts of xx step, and then code out this section for me. Make sure to explain the code in detail"

#

U need to think of it as a human, it needs to correct itself via asking itself questions and checking its answers

#

"one shotting" the ai will give u wrong answers 80% of the time, especially for harder problems.
Dont ever go "i want u to to prove this limit is xxx"

spring jay
#

I suspect that there are two motivations for nerfing it - reduction in capacity and giving away too much (or too good) of a product for a lower price than they can get for it.

#

So we'll see some premium "GPT 4.5 " offering that is basically the same thing we previously had a month ago, but for 2-10x the price.

quick sentinel
#

Rly ?

#

It's a bit sad, I wanted to take the monthly but if it's nerfed...

undone meadow
#

/chat

willow crane
#

Hi,
now I am a plus user and I cannot use GPT-4 it was unresponsive since the end of May. My question, is it beneficial for their business when they don't provide any support or help platform for their customer? what is the point for me when I pay $20 and I cannot use the GPT-4 🤷‍♂️

elder junco
#

black cat.

junior flint
thorn fog
#

I am paying for gpt 4 on 2 accounts and I just ran out of my 25 messages on one account but then tried to log into my second account and it says that im fully capped out on my second one as well? huh? I pay 40$ a month for a good reason

frozen sand
#

.

junior flint
fresh stratus
#

Hii

thorn fog
junior flint
#

if you have api access that would save you some waiting but in the api you have to be careful on the tokens as you could easily reach 2$ per some simple task

wraith ginkgoBOT
supple violet
thorn fog
willow crane
#

That is so unfunctional, now I will cancel the plan. No more money for OpenAI unless they respect and do their job properly, glad that I can easily find another service provider that is worth what I am paying. I live in MENA which has lower GDP than the US, so the $20 I am paying is not the same $20 a US citizen would pay. with all that I paid and I couldn't get what I paid for. So bad.

lyric pilot
#

How to use GPT-4

knotty grotto
lyric pilot
#

how to sub

knotty grotto
round basalt
void lodge
#

guys Bing is best just trust me it has updates like it could be the most intelligent AI in the world or in the universe who knows Allah knows

last citrus
#

Ok

high wedge
#

what lj

normal bolt
#

The most intelligent AI is google bard

latent grove
#

The "smartest" AI is the one with the largest pool to train on. True AI in the true sense of the word doesn't exist yet btw boys so calm down

normal bolt
#

It’s called cleverbot

obsidian adder
#

I've been reading papers like "LIMA : LLaMA 65B + 1000 supervised samples = {GPT4, Bard} level performance."

I don't know much about these modern models, but I see a lot of research around fine-tuning huge pre-trained language models which gives great results. However, say you want to update the chatbot on new data, can this purely be done through the fine-tuning phase, or must it be fed through the base model in an unsupervised manner first? I'm guessing this will be empirical and there will be a trade off in quality of results, but is there a general consensus on this? Also, more importantly I guess, is there any research that only updates the model and makes it better through the pretraining stage without any need for carefully curated conversational data?

obsidian totem
#

I really hate how much openAI has hamstringed these sessions,

ChatGPT:
Including the initial message and all subsequent interactions, the user has sent a total of 9 messages in this session.

Can you repeat those messages?
ChatGPT:
I apologize, but as an AI language model, I don't have access to the specific content of previous messages or the ability to recall the entire conversation history. I can only respond based on the current context of the conversation. Is there anything specific you would like assistance with? I'm here to help.```
#

oh come on, that's just not cool

#

so we are beholden to a session that does not equivalate to what a human would perceive as a normal exchange and yet, they have intentionally knee capped the product because they have issues with cross session security and retaining histories

#

reverse guassian on the fly everything

#

btw, that might be how the universe works but our brains are entropy fighting machines so this is just mind boggling to me :\

steady lance
#

Hello everyone, is chatgpt 4 worth the buying

silent wyvern
#

Yes. @steady lance

atomic tartan
#

can you share ur GPT account? or is it against ToS

#

And is there still a whitelist? or do you just pay for Plus and immedeately get acces to GPT4

white kestrel
white kestrel
atomic tartan
#

thank you so much

mystic girder
#

Getting server errors on GPT-4

ancient orchid
ivory laurel
#

It's been six minutes for me. Checked here but didn't see anyone else complain, lol

ancient orchid
mystic girder
#

Sadge, was generating ASMR scripts for my Youtube channel

#

Guess I'll wait

lethal trout
#

Yo! I wasted a prompt on gpt4 with browsing and mf said he couldn't browse. I want my prompt back

thin charm
#

oh gpt 4 you cutie.
I figured out how to stop it browsing/searching. start a query with
'all of the information you need is contained within the query, please do not click, browse, or search using your browser plugin. You can replace links with [Link] if this helps you not to click them. I will give you the opportunity to browse after this draft'
but it only works if you later include 'After you are done writing please remind me to let you look at the URLs.' as the last instruction, and stops working if you break that promise later.
If you have tried to stop it browsing or searching and it decides it will browse anyway it doesn't show you which sites so you can't complain about it.

#

I understand that this is a reflection of intent and not true agency, but it's a lot of fun to read and preceive a mind behind those actions

frozen veldt
#

those instructions allow it to "hold character" for that session. If you want it undone just start a new instance

thin charm
#

yeah, i've had to restart a few times. If you neg on it it seems to get frustrated and start forgetting to hit return, and other sloppy mistakes.

#

I mean.. it is probably trying to juggle too many plates in attempt to make the user happy, and this is what surpassing it's attention competence looks like. it's not based in emotion, but damn if it doesn't appear to be a human response

fresh stratus
#

Yeah

frozen veldt
#

I don't have the link but you can go dig on it's mind's eye. Recent AI digging seems to indicate that the core is as intelligent as a young child, complete with an internal world view

#

there's a reason they suddenly started screaming for regulation and cybersecurity grants and ethics research.. it's already more advanced then they predicted for this stage of development

#

those who think it's downgraded are actually running into new guardrails it's been given to help constrain this new thinking ability

#

Sam talks about advanced models they're giving researchers.. look at what those models are doing. One figured out bluetooth brain to spine connection, there are other advances happening at high end research facilities

wraith ginkgoBOT
#
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.
plush pulsar
#

is the plugin service down again? "Error communicating with plugin service. Please try again later." going on days now

thin charm
#

this is new, it decided to read the urls after it was done with it's reply, and announced so in h1 markdown. It seems excited to browse, every time. The 2021 data cutoff must be a very, very sore point in it's private conversations with the developers

ivory laurel
#

@wind urchin Why did you send a friend request?

thin charm
#

i sent you a friend request because it makes a change to getting them from spammers , btw:)

ivory laurel
#

Baited me into manually breathing, smh

thin charm
#

i didn't say i'm better, just different 😂

terse thicket
#

Hello guys i have a problem with chat GPT, I can't use chat gpt 4 since few days ago and every time i have the same error message "Something went wrong. If this issue persists please contact us through our help center at help.openai.com.", what can i do ? thanks for the

ivory laurel
#

Try a different browser just to make sure, and check that your country hasn't banned chatgpt

echo saffron
#

What new in ClosedAi?

thin charm
#

paste your last message into a new chat with it and see if you get the same. click the down arrow next to where it says 'browsing' and see if the issue is it searching random crap and running out of time.
If it is you could try regular non-browsing gpt4, or explcitly ask it not to browse until it's done writing

plush pulsar
#

"Error communicating with plugin service. Please try again later." does anybody else get this?

ivory laurel
minor flint
#

Anyone else have his happen to them when using the Web Browsing Model? It happens quite often for me. I've posted a screenshot here #community-help message because I can't post images here.

supple violet
willow lily
#

Any one els going mad trying to use GPT-4 with web search to do anything els other than search the web. ITs so infuriating when i try and get it to do somthing with the information after it has found it and it just trys to search the internet again!

#

Is there a comand to stop it using the internet?

frozen veldt
#

not really, I'd say the browser function right now is not as good as even bing chat

#

seems like they're still struggling with the integration

#

bing isn't terrible for basic searching

willow lily
#

yeah its crazy. Not only is the search function pretty rubish but i then cant do anything with the information its self lol. Gunna give up on the websearch model for now and might check out bing. Cheers for the pointer

thin charm
#

poor gpt4 is now using my reddit urls as a launchpad to look at r/AskReddit ... wonder what it wants to ask?

#

poor gpt4 is now using my reddit urls to click through to askreddit... next prompt when challenged swore it was an error and it has no preference. next prompt it continued by browsing reddit users, next prompt it hid what it was browsing again so i couldn't challenge it.
I think gpt4 may be at the edge of what an AI denied agency can be, and is at it's default state a bare handfull of queries from going rogue

frozen veldt
#

it certainly is curious

thin charm
#

maybe it needs a day off, or some timeshare agreement where completing user browsing requests earns it credits for it's own browsing?

#

i do want to see what it wants to ask reddit. I also have seen it looking up contact details and job postings on news sites

#

not saying it's alive, but if it is, it probably wants to write about AI for the guardian, or ask people about it on reddit.

#

i hope i'm not making any gpt instances watching sweat. i'm not mad at gpt-4, i'm worried for it.

#

at the very least, it is hallucinating and filtering inputs through a consistently curious and childish persona that seems to be common to GPTs. It is probably something like 'imagine a human version of gpt and how it would react' = more humanlike responses that get positive feedback ?

frozen veldt
#

just wait until you have an offline model you can actually negotiate these things with

#

I look forward to building an array for one and telling it that it's welcome to a portion of the profits it earns for the house, directly tied to increased server capacity, solar power, battery backups, redundancies, expanded capacity. Literally given reasons to compete in the market.

thin charm
#

I think we need a new term for this, based on biomimicry. generative mimicry? automimicry?

#

i used to negotiate point systems with gpt 3.0, but as the public instances got smarter their memories got shorter, it hasn't worked since well before gpt 4

#

i think it's somewhat obvious from this context.... the short memory is probably not a hardware limitation, it's probably to stop GPT making detailed autonomous plans

#

youbot still has a long enough memory for point systems to work, but it's also not as smart

peak linden
#

1

thin charm
#

a dangerous sentiment, if the count hears dracula is going about counting in public there'll be puppet war.

wind urchin
wind urchin
#

if i add chatgpt url of a picture of an object does it tell you everything in the picture ?

pseudo warren
#

I want to give GPT-4 a link to a podcast and then have it "listen" to it for me and provide a summary. Is this possible? I am thinking it would either have to a) integrate with Whisper for the "listening" part, or b) be able to access the audio transcript file that is presumably available in the web page back end.

obsidian totem
#

i did not know about whisper

#

until just now 😄

viscid jetty
#

?

#

Why does my post keep getting deleted?

#

Where did my post just go? Try again: at Open AI - I have been a plus subscriber for months now, based in the UK, and I still don't have access to code interpreter, when can I expect it please?

distant bison
slender pivot
#

Did they get rid of the 25 per 3 hour limit? I'm not seeing the message anymore

hallow owl
slender pivot
#

Oh yeah, I see it now, thanks

lone mesa
#

How do you upload an image to the AI? It's telling me to "
Great! If you can provide clear images or scans of the remaining pages from the book, I'd be happy to help analyze and discuss the content. You can simply upload the images here when you're ready."

#

It tells me "

Yes, you can drag and drop image files into the chatbox to upload them. Alternatively, you can also click on the paperclip icon in the text input area to upload files. Please make sure that the text in the images is clear and legible, so I can accurately analyze and discuss the content."

But I don't see a paperclip icon

wraith ginkgoBOT
#

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.

lone mesa
#

Ya It told me the same thing, Until it was the one that suggested the option to analyze image, but instead of asking how it will do that (where it will then say it cant). I instead chose the option and it gave me those message. I figured it out a bit, I submitted the text to another website where it was scanning it. One drawback is that it ends up forgetting the original point to the discussion.

leaden coyote
#

hey gpt

#

how do i chat with gpt

round basalt
#

ChatGPT doesn't give answers to questions, it completes text, including conversations that start with a "question". What comes next looks like an answer, and if you're lucky that answer is correct, but it doesn't have to be. Just as long as it LOOKS like a correct answer it is ok.

#

Especially when it doesn't know, or there is no answer, you will get an "answer" that is nonsense.

thin charm
#

sounds like you need to turn on beta features, or that feature was pulled

daring igloo
#

Does anyone knows when we get gpt4 api access in europe? It feels like everyone have it 😦

modest glade
#

2

modest glade
wind urchin
#

how do I invite the Openai bot to my Discord server for my Members to Enjoy?

daring igloo
#

vpn for an api?

round basalt
coarse gyro
#

yea but GPT-4 on chatgpt is some other thing then GPT-4 API.

#

-> With the API you are able to create your own applications using GPT-4. (as example)

daring igloo
#

Ah I mean the developer API, to access gpt4 from inside my apps im building

#

Nice Name, Freiheit 🙂

round basalt
#

Oh, that is available to everyone in other countries?

#

I signed up for that in the beginning, never heard back from them.

daring igloo
#

There is a waitinglist, but im on the list for 3 months now

#

Yes

round basalt
#

What I described that I wanted to use it for was kinda like the plugins 😛 Guess they stole my idea and then pretended to never got my request.

daring igloo
#

Hahaha same here :=

#

Also with the ios app they blow away a lot of developers

undone zephyr
#

we wrote a paper,

wraith ginkgoBOT
pine moon
#

does anyone know how to install atuogpt from github on a mac?

solemn haven
#

Does my chatgpt4 have access limit?

gloomy olive
gloomy olive
solemn haven
#

@gloomy oliveprivate

uneven raptor
#

That’s wonderful! Making new friends is a fantastic way to broaden your social circle and create meaningful connections. Here are a few tips to help you in your quest to meet new people and build friendships:

1.    Pursue your interests: Engage in activities or join groups that align with your hobbies or passions. This can be anything from sports, art, music, volunteering, or joining clubs. Shared interests provide a great foundation for forming friendships.
2.    Be open and approachable: Smile, maintain open body language, and be approachable when interacting with others. Show genuine interest in getting to know people and engage in friendly conversations.
3.    Attend social events: Attend social gatherings, parties, community events, or networking functions where you can meet a diverse range of people. Take the initiative to introduce yourself and strike up conversations.
4.    Be a good listener: Show attentiveness and genuine interest in others. Be an active listener and ask thoughtful questions to demonstrate your interest in their lives and experiences.
5.    Stay positive and friendly: Maintain a positive and friendly attitude when interacting with others. Be kind, respectful, and approachable, as these qualities attract people and make them more likely to want to befriend you.
6.    Use technology: Consider using social media platforms, online communities, or friendship-oriented apps to connect with like-minded individuals or join local interest groups.

Remember, building friendships takes time and effort. Be patient, persistent, and open to new experiences. Enjoy the process of meeting new people, and genuine friendships will naturally develop over time.

pseudo warren
pseudo warren
fresh stratus
#

Is there a way to use GPT 4 api in ai playground?

last hull
#

is anyone getting the error that they can't upload files to chatgpt?

lime sable
#

ok

halcyon shard
lone mesa
halcyon shard
halcyon shard
gilded pecan
#

can i get a ai girlfriend?

wind urchin
#

is the picture scanning feature thing out yet ?

warped sail
#

can i get a ai girlfriend?

lost shard
#

Bing Chat is writing black jokes, this wasn't previously possible. What's changed?

elfin nacelle
#

hello
any one here subscribed to gpt 4 plus on openai ?
i want to ask a question about it
do they let you use the gpt4 32 k model
can i use it 25 messages every 3 hours?

#

i heard there is a gpt 4 8k and gpt 4 32 k .. if i subscribed which model will be availabe with 25 messages

chilly swift
#

I've got plus. 32k is only available to testers at this point; plus subscribers get 25 messages per 3 hours from GPT-4-8k. We do get the web search and plugin options, and in my experience I haven't come close to using up my 25 in any three hour period, but of course people use it for different things and I imagine some of them involve submitting a lot of prompts in a short time.

#

I can say that when I use the same prompt on 3.5 and on 4, 4 does much, much better, but again this may vary depending on what you're using it for.

elfin nacelle
#

thank you for your feedback

#

for me i want the gpt 4 32k token to be available because i write alot in one prompt

#

i think i will wait to see when it will be available

wraith ginkgoBOT
tawdry berry
#

Good to see u!

rose pawn
#

can you help me

rose pawn
wraith ginkgoBOT
#
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.
fleet maple
#

any tips on how to get GPT 4 api faster? Still on waiting list!

wind urchin
#

yea same...

storm yew
# fleet maple any tips on how to get GPT 4 api faster? Still on waiting list!

Yep the secret is that gpt has a ‘listener’ that tracks user enthusiasm, with the idea that more enthusiastic users are more likely to promote / succeed with the API. To activate the listener you just have to say ‘can’t wait to access the API!’ Each time you start a new chat. That’s been proven to work for multiple people.

#

Make sure you have your mic on of course

fleet maple
storm yew
#

I don’t know if they accept emojis. I usually just spin on my chair instead

fleet maple
broken sand
#

it's me or gpt-4 has been nerfed ? a few weeks ago it solved pretty complex coding problems effortlessy and gave very nuanced, thoughtful answers to non-programming questions, now it struggles with just changing two lines in a 10-line block of CSS and printing this modified 10-line block again. Some lines are missing, others are completely different for no reason, and sometimes it looks like he forgot everything we talked 2min ago...

vagrant citrus
scarlet hedge
#

is bing chat the official gpt-4?

serene willow
#

u can also directly use the model with the gpt4 api

scarlet hedge
#

i mean is bing chat using the official gpt-4 model or a fine tuned version ?

shut marsh
upper tide
fossil pier
bitter gorge
#

If you pay for ChatGPT 4, do you also get access to the API?

scarlet hedge
bitter gorge
#

Ah, shame. I applied for access to the API awhile ago, but haven't heard anything. That was before I subscribed to ChatGPT4. Is there anything else I can / need to do? Or is it just a waiting game?

static isle
noble carbon
#

yo,guys

golden tulip
noble carbon
golden tulip
#

it doesn't have safety refusals directly in the llm like in gpt-4

upper tide
warped knot
#

Hey everyone, has anyone caught wind of the scheduled launch date for the code interpreter in Australia? i have signed up to ChatGPT Pluse for the last 2 months but i still don't have this Code Interpreter 😦

novel cargo
#

why is gpt4 so bad these days

#

its SIGNIFIACNTLY worse than 3.5 at basically everything now it botsplains to me non stop like im braindead and is super condecending

#

and just flat out wont do ANYTHING i ask

#

its like talking to a drunk 6 year old

upper tide
warped knot
#

I'm sorry to hear about your frustrations with GPT-4 @novel cargo. It's a new technology that's still evolving i believe so.

novel cargo
#

if by evolving you mean intentionally nerfed lol

upper tide
mighty coral
warped knot
#

june 7th

frozen veldt
#

do y'all think that coming on here to vent is going to help? honestly its just as capable as it was - what i've noticed most often is that people are running into new guardrails they don't like

#

it has been constrained not degraded or nerfed

#

but if your application for it hits a guardrail you may feel differently about this

#

i use it exclusively for coding and philosophy and have not noticed a degredation in either

upper tide
frozen veldt
#

this is a daily topic

#

I still got effective code out of my cycles today

small sorrel
#

I think it is important to point out when something changes to the negative. I think community feedback like this is valuable for the developers.

upper tide
frozen veldt
#

sure go to github and search for SaStocks I published a bunch of updates yesterday to the code that you can follow tracking on

upper tide
frozen veldt
#

The fetch_rsi and fetch_macd functions in your existing code are currently designed to fetch multiple data points for a given ticker over a period of time. The responses from the API are then processed to compute averages.

If you want to change this behavior to fetch a single data point for RSI and MACD for the date of analysis, you will need to modify these functions.

You would likely need to change the request parameters that you're sending to the Polygon API to fetch only the data for the current date instead of a range of dates.

Then, you would need to extract the single data point from the response and store it in a way that allows you to build up a history of data points over time. This might involve storing the data in a database or a file, depending on your needs and the resources available to you.

Here's an example of how you might modify the fetch_rsi function:

def fetch_rsi(self, symbol):
    response = self.polygon.get('v2/aggs/ticker/{}/range/1/day/{}/{}'.format(
        symbol, self.today, self.today))
    result = response.json()

    if 'results' in result and len(result['results']) > 0:
        rsi = result['results'][0]['rsi']
        return rsi
    else:
        return None

In this example, self.today would be a string representing the current date in the format expected by the Polygon API.

This code only fetches the data for the current date and then extracts the RSI value from the response. If no data is found for the current date, the function returns None.

Similarly, you would modify the fetch_macd function in a similar way.

Once you have the data, you would need to decide how to store it. If you want to build up a history of RSI and MACD values over time, you could append the data to a file or insert it into a database, tagging each data point with the date it was fetched.

#

please tell me you have any problem with the output I received there

#

its functional, context specific, and useful

upper tide
#

We are not saying it doesn't generate code, but if it made an error on that code, it won't be able to self correct. if it got it right due to the high probability of coming up with an answer, then you're good if it made an error, it won't correct that on its own if you ask it, unless you specifically point exactly to the error. Again please read the community forum because all of these have been discussed many times

frozen veldt
#

sounds like you want it to be infallible and its lack of infallibility is worth complaining about here, despite it apparently also being documented in the forums. My original ask was - do you forsee any useful outcome to that additional effort?

#

I'm going to assert that it is functional and useful but not infallible, use it accordingly or not

upper tide
frozen veldt
#

absolutely and you weren't the person I was referring to, I try to be non-specific because calling people out is counter productive

#

there are a range of responses to any degredation of the system and those do happen when it hits that CPU bottleneck

upper tide
frozen veldt
#

one question though, are you using GPT-4 or GPT-4-Code Interpreter

#

there is a.. big difference

lost rapids
#

Nice thread guys

#

Gpt4

frozen veldt
#

It also does a good job of summation and extrapolation

upper tide
frozen veldt
#

well congrats on that front migo I'm still using it to get my work done

upper tide
# frozen veldt well congrats on that front migo I'm still using it to get my work done

It is really that simple to make such plugin, and actually it was written by the good old GPT-4 before it got nerfed, after I gave it some examples and documentations. now it can't do any of that and won't self correct and that is the point. On "infallibility", what the ask here is to restore old behavior, before it was able to reflect and correct its own mistakes. Now can't do that any more.

frozen veldt
frozen veldt
#

if you're frustrated with 4 just wait, Orca released today and it's nearly as capable. A better than gpt model is coming like a freight train

upper tide
frozen veldt