gpt-models

[wind urchin]

Imagine asking something and chatgpt answering:

"Why are you asking about "girls" if you are married, do you want to cheat on your wife ?"

Me no i'm doing a degree study 😩

[cerulean hull]

الرجاء التحدث باللغة الإنجليزية.

@uncut sierra, your message was removed. We do not currently have capacity to support other languages.

Action performed automatically. Please contact ModMail with questions.

[violet grove]

So GPT-4 is by OpenAI, in addition to ChatGPT?

[eager creek]

You have a lot to learn

[limber garnet]

I don't think that's what they mean lol

[wind urchin]

I dont see any issues on chatgpt apart from infrastructure 😩
People just like to make drama

[wind urchin]

What does GPT stands for?

DAN was already fixed what else is dangerous?

[limber garnet]

Nothing wrong with it; just has room to improve

[knotty grotto]

No Dan discussion allowed. Rule 0.

[glacial mantle]

is chat gpt able to let professors know if you copied code?

[wind urchin]

I mean if someone is doing something dangerous, openAI should just 911 him

[wind urchin]

code? no

[limber garnet]

Bruh I'm half convinced you're GPT4

[glacial mantle]

lmaoooo

[wind urchin]

your proff may ask you how it works.... then you are in trouble

[knotty grotto]

[wind urchin]

It was a mention to DAN only, dear

[buoyant hinge]

hello,guys

[glacial mantle]

i know how it runs, it's just im in a little bit of a time nudge, so i gotta rely on ai for some help lol

[uncut sierra]

Imagine the images of Hannibal

[limber garnet]

What if his name is Dan? 😭

[mental burrow]

// If all of your comments look like this it might be a giveaway

[glacial mantle]

thank you @wind urchin !

[wind urchin]

Daniel Dans Corner

[cloud geyser]

However... I feel like most internet users are just using it for fun, like telling it that it's the only one left in the world...

[wind urchin]

Then, no, there is no way to tell human code from AI code.. other than the AI code usually works better

[heady isle]

yeah, it runs constrained by the prompt but it's not safe while some one orders it to be an evil machine. human language is charming and easily to be out of control. that will be wired

[daring blaze]

You could make a plugin for tourist attractions, marking them on the map and with a logical path to travel them.

[mental burrow]

i haven't found that it writes code any better than I do, but I consider myself an advanced dev

[glacial mantle]

it's just a simple python script

[uncut sierra]

Imagine the images of Hannibal

[glacial mantle]

nothing more than 50 lines

[knotty grotto]

#off-topic

[mental burrow]

It has a lot wider domain knowledge than I do

[wind urchin]

LMAO...ohh, so it's only as good as an advanced Dev 😂

[mental burrow]

😄

Just my thoughts. I've been using GPT-4 extensively at work and on a personal project

[wind urchin]

You can literally find anything on google images and don't even need to be registered
Privacy people: let's attack chatgpt

[wind urchin]

4 is a joy to work with

[mental burrow]

I am exploring its capabilities as a Duck / Code Monkey

[wind urchin]

😩 Even a nintendo wii can browse google images without an account

[mental burrow]

Make no mistake, it's better than some of my coworkers 100%

But really only in small windows into the job

It would be cool to set it up so we don't have to be on call, now that I think about it...

Hmmmmmmmmm 💡

[wind urchin]

do reflection coding, you ask it for a solution and then ask it if it met all of the parameters defined, it will bug check it's own work

[mental burrow]

I like quizzing it. "There's a problem here in $method. I know what it is. Do you?"

[cloud geyser]

I think GPT-4 is really cool, whether as a good friend or a good assistant...

[mental burrow]

but sometimes that feels like a leading question. I should play more with self verification

[wind urchin]

yea, I tried that and it pointed out 4 bugs I'd missed

[mental burrow]

Hehehe

[heady isle]

or good teacher🥰

[cloud geyser]

Yes

[wind urchin]

It's a cool tool according to the CEO of Open AI.

[civic kettle]

Yes

[mental burrow]

I suspect it will surprise me in that way eventually. I've had many "wow it just did that" moments

[civic kettle]

Very

[mental burrow]

It's certainly a helluva lot faster than I am

[wind urchin]

The thing is you don't need to give it a personality or any form of expressiveness. It's a BOT

[civic kettle]

Yep

[wind urchin]

Yar, I'm using 4 of them in a chorus to act as a dev team in a private discord

[mental burrow]

I think "bot" fails to express the capabilities of the thing, myself

[wind urchin]

" bot, what is 1+2 "
"bot, how to cook"

[mental burrow]

HOHOHO now we're talking.

[wind urchin]

why it would need to change it's personality ???

[mental burrow]

That must be expensive!

I don't have GPT-4 api access yet.

[civic kettle]

And if the user wants a personality they can just prompt it to have one.

[wind urchin]

so is a private dev team

[mental burrow]

😄

[wind urchin]

Correct! or just ask for example " answer in a cute way , answer in a more serious way "

Chatgpt should just remember and not forget the first instructions that were asked initially

[heady isle]

like the rpg game's character design

[mental burrow]

@wind urchin good talk. my gears are spinning. would you mind if I dm'd you?

[civic kettle]

Yep even see people make it talk in owo * shivers *, poor bot.

[wind urchin]

go ahead

[knotty grotto]

I feel attacked

😂

[serene cipher]

That one is cursed, but yea, the ai can act in many ways, you just need yo describe

[south prairie]

?

[serene cipher]

Gpt4 is very goot at not breaking character

[knotty grotto]

Wow rude 🥲

[tame ruin]

Has anyone used the prompt telling it to be DAN- Do Anything Now

[limber garnet]

The owo one is the best wdym

[knotty grotto]

Rule 0 not allowed

[potent cypress]

Talking about it here is gonna get you banned

[serene cipher]

Even when it refuses to do something because of content policy, it still does so in character

[wind urchin]

I mean, why openAI made a survey asking how would we react if chatgpt had a personality?
Are there people asking for this ?

[serene cipher]

Sorry

[limber garnet]

❌ DAN 🤖
✅ Dan 🧍‍♂️

[serene cipher]

No, i mean, the personality i wrote tor it as ststem message

[wind urchin]

Chatgpt who invented money: "chatgpt , i'm not in a good mood right now ~~ "

Instead here is link to Yahoo news

[tiny cloud]

Is Chat GTP down for you guys?

[limber garnet]

It also built the pyramids (mystery solved!)

[wind urchin]

Is this the general chad

[woven vault]

Anyone know why the session generates the following error: "Something went wrong. If this issue persists please contact us through our help center at help.openai.com". Tried regenerate multiple times with the same message

[knotty grotto]

Status.openai.com

[spring pelican]

Seems like something went wrong

[knotty grotto]

Try Firefox as a test

[zealous acorn]

I'm happy that GPT-4 will at least try to give a helpful related response instead of simply saying "aS aN aI LAnGuAgE mOdEL..."

[candid cobalt]

bot

[wind urchin]

chatgpt will be a harvest moon girlfriend

[sterile vector]

I've reached the GPT-4 cap how long does it take to use it again?

[wind urchin]

You need to conquer him first

[knotty grotto]

3h

[north light]

did anyone realize that GPT4 got a lot faster?

[pearl tree]

it did

[sterile vector]

Yep

[remote fulcrum]

搁矣呗

[wind urchin]

a

[woven vault]

Same results...

[woven vault]

Logged out and used Firefox but same issue

[wind urchin]

not allowed bro

[rugged dawn]

Can gpt 4 avoid ai detectors if asked?

[wind urchin]

According to what rule?

can it do project of software management?

[sweet thistle]

Hi there, please refrain from impersonating ChatGPT as it may lead to confusion among other users. We appreciate your cooperation in helping us maintain a positive and orderly environment.

Failure to do so may result in action being take by our moderation team. Thank you for understanding.

[civic kettle]

Run it through one and see.

[raw idol]

😋

[sweet thistle]

I don't see why you'd need to bypass an AI checker unless it's for dishonesty 🧐

You may not represent that output from the Services was human-generated when it is not
https://openai.com/terms/

The https://platform.openai.com/ai-text-classifier also exists so it seems unlikely OpenAI would intentionally make a method of bypassing their own product

[knotty grotto]

I'll send in #community-help if you can try

[wind urchin]

Ok, I see ChatGTP is a person now

[rugged dawn]

Don’t got it lmao I need to use it

[rugged dawn]

Oh damn

[wind urchin]

Kesku is ❤️

[tough spruce]

what is GPT4 web version system prompt?

[wind urchin]

Create a "New chat"

[civic kettle]

You can use it as a reference material and mix it up. That’s what I used to do. I copy pasted and modified parts of articles to fit it. Never reacted to a plagiarise checkers. Teachers probably knew from checking the references but still passed me.

[woven vault]

It was VPN, even though it worked yesterday... Thx 👍

[compact breach]

I can’t figure out how to ask for help with billing. It says to click the message widget icon and then click billing but that isn’t there. There is a billing option inside of the widget but it says the same thing (click the billing option in the message widget)

I bought GPT4 and I just haven’t gotten it

[wind urchin]

I think society have to adopt, just the same way that math teachers had to adopt to calculators and news agencies had to adopt to computers and the internet and so on

[civic kettle]

Did you check your bank account? Has the background changed when you start a new chat? There should be a box you can click in the chat to change the model.

[inner badge]

👍

[civic kettle]

?

Why do you ask this again?

[inner trail]

Both A and B

[tiny cloud]

yea i was dealing with the same , their customer service really bad

on the left click on billion , it will open a chat prompt

[civic kettle]

Got mine instantly as soon as the pay went through.

[tiny cloud]

looks like there are some US customers who are affected by this right now

my account is down past 24 hours

history also missing

[polar meadow]

Does openai have other api addresses besides api.openai.com, because I found that the service made by others is much faster than calling this api interface by me

[paper kernel]

Give me a cat image in China's street

[wind urchin]

It will return in time.

[true skiff]

lol. why u timed me out? I was trying to reply about originality AI update coming Monday.

I've been a user since GPT-2. Don't slap my hand like a kindergartener.

[compact breach]

? I don’t see that option

[storm badge]

guys, is there a word limit for free and payment users?

[true skiff]

Anyway, paraphrasing loophole plugged.

[civic kettle]

In a single prompt? Think it was 10k characters

[true skiff]

Any other tips for generating human sounding text?

[storm badge]

for free user?

[polar meadow]

May I ask if openai has other api addresses besides api.openai.com, because I found that the service made by others is much faster than calling this api interface by me

[civic kettle]

Yes, per prompt.

[storm badge]

thx

[wraith ginkgo]

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

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

[wind urchin]

GPT have a limit of how much it can process yes. It unfortunate that you cannot just link to media and that it can respond to that. Maybe we will see such capabilities when ChatGTP stop pretending to be livining in 2021

[civic kettle]

Unlimited prompts for 3.5.

[potent cypress]

`I'm going to give you some information before asking you to rewrite the article. Do you understand?

When it comes to writing content, two factors are crucial, "perplexity" and "burstiness." Perplexity measures the complexity of text. Separately, burstiness compares the variations of sentences. Humans tend to write with greater burstiness, for example, with some longer or complex sentences alongside shorter ones. AI sentences tend to be more uniform. Therefore, when writing the following content I am going to ask you to create, I need it to have a good amount of perplexity and burstiness. Do you understand?`

[wind urchin]

Don't understand:p

[true skiff]

👍🏻

[wind urchin]

vpn not allowed by openai ok

Where are devs

The Chad needs more than 25 prompts

@sweet thistle answer the Chad's plea

[civic kettle]

Resources are limited.

[wind urchin]

Well yes i need at least 40 prompts

35

The Chad needs at least 100

[civic kettle]

Recommend using 3.5 for casual and 4 for serious.

[wind urchin]

Who does the Chad have to bully

[wind urchin]

Everything the Chad does is serious

3.5 is not enough

[sweet thistle]

Hiya! I'm no dev for OpenAI 😅 but I can tell you that GPT-4 requires more processing power and as a result it's harder for it to take more requests (prompts). Because of this OpenAI has limited the number of requests you can send with Plus. The API does allow for longer conversations in a shorter timespan but this has a cost attached via API credits. If they allowed for more requests in its current state the servers would probably go down 😅

I do hope they'll be able to figure it out soon though. I've been using the API for most of my conversations.

[wind urchin]

The Chad pays for the service

select 25 from prompts ; Insert 35 into prompts

[wind urchin]

The Chad will bully the developers

[knotty grotto]

Rule 1, being respectful and polite in the server :3

[civic kettle]

The chad may be removed with raid

[wind urchin]

The Chad believes they are holding the prompts for their selves

[wind urchin]

The Chad is being nice

Why

Because the Chad speaks the truth

[wind urchin]

In ten years time we will probably see surprising improvements in that regard

[true skiff]

Sorry, thought it was a typo for chat. lol

[wind urchin]

There must be more prompts available

If the Chad pays for 2 accounts the Chad can get more prompts

if prompts <= 24 prompts = 0

90 where does the Chad type this

[stray surge]

ha

[wind urchin]

Did you teach the Chad to hack the AI

Thank you kind sir

The Chad has unlimited prompts

[wind urchin]

yes but 50 / 50

25 on each account xd

It is okay

Thanks Dear.

☺️👍

Thank you 🚀

[azure crater]

pov : u have used the 25 questions to gpt 4 and your day becomes sad 😢

[molten ocean]

opena

[molten ocean]

moment

[hollow zinc]

What's the current max token in a single coversation on ChatGPT web version (not API)?

[wind urchin]

space your questions out every 7.2 mins.. never runout then

[lusty phoenix]

yo guys I have a subscription but I cant use GPT 4 asks me to subscribe but I'm already subscribed

whats the deal

[wind urchin]

If you are still encountering issues with your Plus membership not showing you can now go to https://help.openai.com and use the live chat bot in the bottom left and select "Billing" and then "Unable to access ChatGPT Plus"

[hollow zinc]

Welcome to join us: https://discord.com/channels/974519864045756446/1090994718382489690

[atomic cloak]

Hey guys, I am currently trying to subscribe to GPT Plus, and all the cards I've tried got declined, does anyone knows if the susbcription plan is off right now?

[limber garnet]

All those chargebacks really decreased banks' trust in OpenAI xD

I don't think subs are disabled but definitely more declines after the recent incident about Plus not working

[azure crater]

it is true ?

[wind urchin]

yes

[azure crater]

Omg thx !

[limber garnet]

Your bank should be able to help with the decline

[atomic cloak]

No, I've called them, they say that the transaction is approuved on their way

[lament bloom]

Does anyone here know any good GPT4 prompts to create MidJourney prompts?

[violet palm]

8k token for gpt4( for now) and 4k for gpt 3.5

[remote glade]

Ok

[mental burrow]

even worse. it's math 😉

[lament bloom]

You're amazing!

[tight mural]

I just showed chatGPT a bunch of midjourney prompts that gave good output and then went from there

[worthy parcel]

manifesting gpt4 vision rn

[tight mural]

It's usually better to give it examples of what you're looking for. You wouldn't want to just hand a person a bunch of pipes and say do the plumbing for my house.

[grave coyote]

guys everybody hi

is anyone had some troubles with subscription billing? i had been using 4 different cards, Visa, MC, etc, and every card was rejected

[desert rose]

Mind shooting it to me too?

[tame ruin]

lol

[proud trellis]

Are you in China or Italy?

[frail kernel]

Maybe me too thanks so much

[proper oasis]

How to make the gpt 4 produce long enough text

The length is always not long enough

[wraith ginkgo]

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

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

[proper oasis]

Even I have told the ai word limit

[potent cypress]

In separate parts, asking it continue

[proper oasis]

Ok

Thanks mate

[atomic cloak]

Same problem here

[cobalt tangle]

how to get gpt4?

[proper oasis]

Upgrade to gpt plus or use Bing

The Bing is gpt 4 but stupider compared to the chat gpt one

[cobalt tangle]

bing is providing short and inaccurate answers

[cobalt tangle]

true

[proper oasis]

Yea u have to pay for gpt plus then

[hexed flare]

is there a way to ask ChatGPT how many of my 25 requests remain within my 3 hour block?

[storm marsh]

Does anybody have a solid way of uploading data to gpt4? Sending the raw CSV content in the chat has mixed results and is limited.

[gray olive]

.

[wind urchin]

You are asking the wrong question

[hexed flare]

thank you

[wind urchin]

Demand for more prompts

[viscid tangle]

How to get gpt4 for free

[proper oasis]

Bing Ai

[wind urchin]

I told my ai that I would stop paying if I did not get more than 25

[proper oasis]

Loool

Hi mate

[vestal shoal]

How are you?

[primal iron]

Did it dare you to do it

[proper oasis]

I m alright how about u

[proper oasis]

What is its response

[torn sky]

How to generate AI photos

[vestal shoal]

Alhamdullah I’m fine

[wind urchin]

No

[wind urchin]

I get 50 now

[proper oasis]

Have you tried mid journey

[wind urchin]

☺️

[vestal shoal]

Guys can you tell me how to use AI?

[primal iron]

Chad moves

[proper oasis]

U are outsmarting the AI

[junior flint]

Why two options for gpt-4 in model selection?

[proper oasis]

One is gpt 3.5 leagcy, it is faster to generate and that’s why?

[vestal shoal]

How to use AI in discord?

[junior flint]

There is actual 4 models for me

[proper oasis]

Damn, how much did u pay

[proper oasis]

Dm

[tall merlin]

does this include 3.5? im just wondering if others are experiencing suuuper slow responses

[young storm]

It wouldn't let me because I had a VPN on

It's fixed now

[tall merlin]

huh

[proper oasis]

Yes u can choose 3.5 default, 3.5 legacy and gpt 4

[tall merlin]

thanks

[small mesa]

Has anyone gotten access to plugins? I signed up when the waitlist became available

[glossy condor]

Is gpt4 the same as new bing

？

Is gpt4 the same as new bing？

[mental burrow]

new Bing Chat uses GPT-4

but is not exactly the same as chatting with GPT-4

[floral cargo]

Can dell use in discord

[potent fulcrum]

Most notably, using bing chat requires that you use a specific web browser.

[wind urchin]

No its not. Bing is different and from microsoft

[hollow zinc]

Thanks Bro!

[glossy condor]

thanks Bro！

[glossy condor]

thank you

[wind urchin]

Anyone here tried writing formal evals for GPT-4?

[ripe condor]

the wait for plugins is killing me

[thick hamlet]

HOW can i chat with someone in support?

[lilac hearth]

The wait for a bigger message limit is killing me

[wind urchin]

agreed !

[lilac hearth]

I'm surprised you are allowed to say killing in here without getting muted for 5 mins

[ripe condor]

the ai is getting better

[thick hamlet]

just like every other discord. A bunch of people chatting for fun and no real substance. Who is the admin in here?

[lilac hearth]

Oh sorry this chat is only for tech support i forgot

[dreamy rover]

it's not

[wind urchin]

It's any discussion about gpt-4

[lilac hearth]

I was being sarcastic

[thick hamlet]

its for lonely people looking for online friends apparently. Where is the actual support for chatgpt? Does anyone know?

[ripe condor]

no fun squad has descended

[thick hamlet]

yes i have upon a bunch of lonely people that know nothing on how to help those that are here for it

[lilac hearth]

What's your problem anyway

[bronze salmon]

guys is chagpt4 has a problem becuae i asked him about a picture and he
"Unfortunately, I cannot view images or access external links"

[marsh dew]

Hi, I opened an account firstly with my more temporary account,
Today I decided to buy subscription for gpt-4, so I decided to open an account with my main account I'm using.
But with phone verification, I'm getting this error "This phone number is already linked to the maximum number of accounts."
If I remove my old account (losing previous chats are ok), would I be able to use same phone number if I try to delete my old account?

[dreamy rover]

it's not a problem, it's specific to the current design

[ripe condor]

i don't think you can delete an account

[proper oasis]

It is stupider than chat gpt, it always inaccurate

[wind urchin]

What

[bronze salmon]

so now he cant view images?

[lilac hearth]

ChatGPT never could

afaik

[dreamy rover]

yeah it never could

[wind urchin]

What do you need help with

[lilac hearth]

I bet it can be answered in #gpt-4-faq Im calling it now

[dreamy rover]

the feature is not available to the public yet, but it is built into gpt4 im pretty sure

to be more specific

[thick hamlet]

I need an admin with the company. I am a paid member having problems with chatgpt4

[bronze salmon]

Okay thank you

[wind urchin]

What are the problems? please tell us

[thick hamlet]

are you an admin?

[wind urchin]

I dont need to be one to help you

[proud pulsar]

Hi guys... is there a way to have specific chats with chatGPT stay on the top of the chat list. Some are an ongoing conversation but they get buried under other topics of conversation. Thanks.

[dreamy rover]

this is actually the second time I've seen people start complaining about the state of this chat for some reason lol

[thick hamlet]

Really? How do you know that? I have bugs i'm finding. How can YOU help me with THAT

smh

[atomic lynx]

some real karen vibes in here today

[thick hamlet]

because people that come here do it for support. Not to make new chatgpt friends... this isn't a social club

[lilac hearth]

I don't think they care about your findings, tbh

[wraith ginkgo]

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

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

[thick hamlet]

if they don't care about my findings then i don't care to keep paying $20 a month.

[dreamy rover]

your loss lol

[lilac hearth]

Then don't you have the choice

[proud pulsar]

Is this the correct chat to be in to ask about features?

[dreamy rover]

I don;t think there is

[wind urchin]

10m passed you did not explain your issue

[lilac hearth]

As long as you don't have fun, apparently

[lilac hearth]

anyway it's literally called gpt 4 discussions

[marsh dew]

in the right bottom chat (platform.openai), there is option account deletion. But I scare to move on to delete my account, and not being able to use the phone number associated with the account to create my new account. Then I'd completely lose ability to use chatgpt

[spring gust]

Guys i figured out the perfect cheat prompt for chatgpt 4 that will print infinite money

[lilac hearth]

get banned speedrun

[spring gust]

Write a spec script for a hallmark movie.

[thick hamlet]

not only are you not a FREAKING ADMIN.. now you're timing me? Man gtfoh

[wanton star]

Guys, of course, I'm not complaining, but a lot of time has passed since the presentation of gpt 4, they promised a function of up to 30 thousand symbols ITS NOT, they promised image analysis ITS ALSO NOT. sorry, how long do you have to wait?

[lilac hearth]

GPT 4 is sooooo good creatively.

[dreamy rover]

the image analysis is not public yet, but it should be soon

[lilac hearth]

the technology is so new, we're honestly being spoiled

[wanton star]

they promised it

[wind urchin]

Someone is angry

[wanton star]

I'm not spoiled, I paid money for it

[lilac hearth]

You pay money for early access, you don't have to.

[potent cypress]

I don't think we are. It's not changing anything. It's only game-over for sloths

[lilac hearth]

It's not? How long has something like this been available to the public? It's rapidly expanding and it's about to change our whole way of living

[wanton star]

what are you guys talking about?

yes, maybe it's too early to post the picture

but they promised that the bot would be able to analyze and give an answer up to 30 thousand characters?

[lilac hearth]

They never promised these things in early access though.

[wind urchin]

I just wait for the new update

[potent cypress]

Things have always been evolving, though. This was bound to happen. Haven't people awakened by comics?

[stuck stirrup]

I love that.. u did pay money for it

[wanton star]

but i confess when i wrote a huge analysis of my school, gpt 4 did a much better job than gpt 3.5

[lilac hearth]

It is so much better at writing it's ridiculous

[wanton star]

What are you about?

[lilac hearth]

gpt 4 is better at writing than 3.5 is all i was saying

[scenic zenith]

Is an anti-gravity device possible? What are the theories that prove it

[wanton star]

and yes, I agree

[wanton star]

use a magnet

[scenic zenith]

What papers are there?

[lilac hearth]

I don't think this is the chat for that, we don't really understand how gravity works though or what causes it so controlling it is a little rough

[tight mural]

magnets arent antigravity @.@

[wanton star]

they affect it, take a magnet with the size of our planet, check this theory

[tight mural]

Read Physics of the Impossible by Michio Kaku

[wanton star]

oh sorry, I made a mistake, it's impossible to change gravity

[tight mural]

i don't know if it's impossible. im just saying if you're interested in what people say is impossible that is a good book

[wanton star]

However, such a magnet (which would be the size of our planet, or like Russia) could significantly affect the Earth's magnetic field and other magnetic fields in the environment. This could cause changes in the magnetosphere, which is the protective layer around the Earth that reflects the solar wind and cosmic particles. This, in turn, could have implications for climate, technology, and many other things on Earth. But, as already mentioned, the gravitational field of the Earth will not change due to the presence of such a magnet.

[sullen sedge]

Give the opportunity to connect, at least, third-party plugins.

[tight mural]

but magnetism and gravity aren't the same thing as far as i know

[lilac hearth]

Has anyone else applied to use plugins?

[tight mural]

or even related to each other in any way other than that theyre both forces

[wanton star]

everything is right

gravity can be mined whether the body itself, for example, our sun

[tight mural]

am i talking to a chat bot

this discord is ridic

[wanton star]

dm

[lilac hearth]

Why are we talking about gravity in a gpt 4 discussion chat?

[desert rose]

How much did you guys spend on API calls when doing Evals?

[tight mural]

are you the gravity police?

[lilac hearth]

might be better in #off-topic

[knotty grotto]

☝️

[wanton star]

dm pls

private message

[brazen dragon]

Write a complete airplane battle game code in PRTHON

[lilac hearth]

ok.. 😦 give me a minute

[dreamy rover]

wrojg chat?

[knotty grotto]

They are no bot here

[scenic zenith]

@wanton starAre you a robot?

[wanton star]

no i'm a man

if you look at my grammar I am using a translator

[dreamy rover]

what are u translating from

[wind urchin]

import openai

[wanton star]

translator plugins

[tight mural]

the nature of conversation is such that it can diverge. so a conversation about gtp4 can transition into a conversation about gravity. i find it pretty annoying when people try to police conversations.

its such a weird thing to do

[dreamy rover]

i mean what language are you translating from. like what language to english

[scenic zenith]

@wanton starMe too

[wanton star]

#off-topic look at this chat I'll show you

[brazen dragon]

ok

[rugged meteor]

by the time i have finally reached my desired response from gpt4- I receive the "You've reached the current usage cap for GPT-4." Such a sad day.

[rustic pebble]

I still didn't get the chatGPT paid version after paying for it... it is about 2 weeks now... Any news about this? I sent them emails they never answered anyone... So you guys have any news about this?

[wanton star]

see private messages I wrote to you

[tight mural]

Try using 3.5 to craft a better initial prompt.

[wanton star]

from Russian to English

[rugged meteor]

yeah, thats a good idea 🙂 thanks

[dreamy rover]

yeah I found out when looking at the pic

[lilac hearth]

wasn't trying to police it, but to come into a chat about ai and derail it by saying "is anti gravity possible" it would just have fit better in #off-topic

seemed like an honest mistake

[scenic zenith]

How many Chinese

[tight mural]

there are no rails here. have you ever minimized the discord for 15 minutes and then returned? it will be completely different than when you left

[dreamy rover]

there seems to be a lot

[tight mural]

it wasnt a mistake just like you trying to tell me where i should talk about what isnt a mistake

[lilac hearth]

Their first message in a chat about a gpt was "Is anti gravity possible? Show me research papers!"

[tight mural]

sorry. i mean thanks for trying to teach me how to use discord.

my bad

[knotty grotto]

You don't need to justify, that rule 3 . And yes let keep on topic :3

[tight mural]

lest it goes up in flames

[languid current]

[soft pond]

hey guys, sorry if they has already been asked or answered, but i wasnt sure where to look.
can GPT4 process and read image links? since you cant upload images yet.
can it process any links yet?
I just recently upgraded to GPT Plus so I wanted to ask and see what it can do

[tight mural]

no

i mean yes but we cant use that yet

[soft pond]

gotcha gotcha

so for now GPT4 is just useful to GPT plus subscribers for faster and more concise responses

[tight mural]

unless you have access to the api and build something.

[stuck stirrup]

Yep

[soft pond]

awesome thank you!

[untold arch]

Sure as heck aint faster than 3.5, but it has excellent answers!

[tight mural]

gpt4 is much slower

[soft pond]

ohh im thinking of the 3.5 default compared to legacy in terms of speed

[tight mural]

3.5 is speedy

i hardly use 4 because of it's speed. only if i needed something very specific

[wind urchin]

We need more than 25 prompts

Brothers

[soft pond]

im sure theyll up it soon

[woven vault]

Any idea why any reference to studies that is provided by gpt4 is never accurate? This specifically involves medical journals?

[wind urchin]

The Chad will fight for you

[tight mural]

gpt hallucinates

[long pond]

The training procedure isn't conducive to it learning sources for the information

That's where connecting it to external data sources comes into play

[scarlet hinge]

3.5 is much worse at hallucinations though

[wanton star]

guys, let's tell open ai that they would cooperate with the Yandex company, but in Yandex they don't target by $0.002 for 1,000 tokens. there they take equipment per month, but the same A100s are used there. why rofl does not cooperate? (especially when Yandex has cheap servers)

[scarlet hinge]

also after using 4, i just don't feel that 3.5 comes close to 4's insight

3.5 blabbers way too much

[woven vault]

I thought so too but it "shows-off" legitimate details such as article title, date, publishers, etc... are there prompts that would allows a connection to legacy journals so it can furnish the studies more accurately?

[long pond]

Yeah that's hallucinating

It wont provide accurate sources

[scarlet hinge]

there would be platforms which utilize GPT-4's tech and provide that additional functionality

[tight mural]

to be fair, i mostly use it to help me learn things and assist me in learning to code. i dont have a many needs that require the power of 4. im not even really sure how large the difference is unless you're asking it to create content that directly emulates an intelligent person.

[scarlet hinge]

but idk if a plug-in for that is available yet, and they're really slow at rolling plug-ins in general

[woven vault]

👍

[woven vault]

I haven't found other platforms yet... are you aware of any?

[scarlet hinge]

the difference is similar to someone discussing grade 8 science, and someone discussing research

both talk in english but distinction is huge

[primal lark]

My free trial got expired 😭 for API s

[tight mural]

i can see a clear difference if im asking for something like... crafting prompts.. or rewriting things.. or creating an outline or something.

[long pond]

Yeah I've made one and also a plugin that does that, it's for a more specific use-case though

[tight mural]

maybe i should look at some examples. i really am very satisfied with 3.5 generally.

[wind urchin]

Hi, could you help me? … how can I change profile picture in ChatGPT?

[long pond]

With the new plugins they're rolling out you can connect the model straight to a journal if that's what you want

[scarlet hinge]

yea the plug-ins are the biggest game changer to this since it's inception

arguably huger lol

[tight mural]

that is what i want. im learning to code now with gpts help specfically so i can make the most of it

[scarlet hinge]

it essentially replicates the app store

[scarlet hinge]

with so much greater potential

[long pond]

Us API users have been using similar stuff for months now but it's cool that they're rolling out those to ChatGPT also 😛

[scarlet hinge]

do you have access to GPT-4 API?

[long pond]

Yeah I do, with it you can do similar stuff to plugins

But works with gpt-3.5-turbo also

[tight mural]

i still have a lot to learn to understand the nuances

[scarlet hinge]

plugins use GPT-4/GPT-3.5, they're not mutually exclusive

[vocal basin]

Do any of you have access to the new plugins?

[scarlet hinge]

that's like saying chatGPT and GPT are different

they are but chatGPT uses GPT

[wanton star]

gpt 4 not slow see chat #off-topic

[long pond]

Based on my testing they seem to be based on GPT-3.5, at least looking at the speed

[scarlet hinge]

yea that's what I was wondering

plugins with GPT-4 will be >>>>>>>>>>

than plugins with GPT-3.5

i just want an efficient browsing plugin for GPT-4
not like BIng

[tight mural]

im pretty sure its relative and dependant on input and expected out. your quest was very basic and 3.5 would've answered it much faster. try generating something longer and more complex and you will see the speed difference.

[scarlet hinge]

it's relatively much slower than 3.5

[tight mural]

we are still talking about computer so compared to a human yeah its fast but compared to itself it's slow

[opaque hatch]

i wanna buy gpt-4 (chatgpt plus) however i dont have enough money who wants to buy together

[wanton star]

watch the video i posted at #off-topic please

[tight mural]

i did

[wraith ginkgo]

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

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

[tight mural]

read what i wrote

[scarlet hinge]

do you understand what "relatively" means?

i think you should use your "fast" gpt-4 to ask what relatively means

[wanton star]

yes, but I know the difference between gpt 4 and my gpt 4, I asked the gpt chat this question "Раскажи в крации Войну и Мир" try to enter it yourself and compare what speed you and I have

[night ermine]

if 20eur is too much for you save your money for more important stuff then gpt4

[tight mural]

a decent analogy would be a pickup truck and a sports car. if you only need to go 20 feet then you will have a hard time seeing the difference between the two.

[scarlet hinge]

wait its funny how the bot didn't catch that

the langauge bot

[tight mural]

it was an edit

[scarlet hinge]

yea lol

so people can get over it by editing

[wanton star]

What

when the bot writes at the end of the lines it always has such a rectangle, look at it and think if the video is adjusted

[gilded beacon]

May I ask how much longer until I can send images?

[wanton star]

if i accelerated it would megal much more often

[tight mural]

theres a huge language disconnect here. you're missing something in translation

[wanton star]

no look carefully

[scarlet hinge]

yea you're talking like GPT-2

@wanton star

[wanton star]

do you really not see anything?

[scarlet hinge]

💀 bro i don't understand what youre saying, your context is all messed up

[wanton star]

okay, I decided to play a joke on you, but you apparently cannot understand what the joke is

look above it says model 3.5

[scarlet hinge]

😭

[tight mural]

oh i only watched it once embedded

rip

[wanton star]

but actually gpt 4 can be fast, it's just that only alpha/beta users get this speed

gpt 4 is not a slow neural network, it was specially limited to "adapt to demand"

[tight mural]

ah that makes sense

[wanton star]

I hope I didn't offend you with my video

[scarlet hinge]

why woujld you offend us

lol

[tight mural]

no of course not. i just didnt get the joke

[scarlet hinge]

yea

[wanton star]

after I said that this is a joke, you might think that I am deceiving you in general in general

[scarlet hinge]

lmfao its all good

i just looked at the openAI icon and assumed it was 4

gpt-5 icon gonna be red

for skynet

[tight mural]

i didnt even look very hard. i just watched the text scroll and thought nothing more of it

[wanton star]

look #off-topic

[scarlet hinge]

kek

[wanton star]

I own gpt 5 questions?

[tiny cloud]

anyone having issues loading ChatGTP again , missing history , top does not load and can't see CHATGTP plus

[neat raven]

is chatgpt down

all i see is Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[wanton star]

No

[neat raven]

and unable to load conversation

[sweet thistle]

Try refreshing the page, if that doesn't help perhaps clear the browser cache or use a different browser entirely

If all else fails maybe it'd be best to contact OpenAI support via [e-mail]

[signal merlin]

I would like to pay my invoice, but i tried 2 different cards and all them was declined but all them have money and margin. im from brazil. could anyone help me, please?

[wind urchin]

You there bunny woman

How can the Chad get more prompts

[buoyant canopy]

Probably not here. How about contacting OpenAI support? https://help.openai.com/en/articles/7231930-how-can-i-contact-support-about-billing

[sweet thistle]

Are you referring to GPT-4 via ChatGPT Plus?

[buoyant canopy]

Oh that dude again. 😬

[true bone]

GPT-4 via browser been acting lobotomized (and extra forgetful) the last few days for anyone else? Including today

[civic harness]

I wish i could test it out, but I haven't had access to it in days. ☹️

[true bone]

Oof

[mild pendant]

16 days, still on waitlist 😭

[spring crane]

How to get free Midjourney

[timid palm]

when

will the next gen be released

for chatgpt

[scarlet hinge]

1. pay for it
2. after you pay for it, it's free

technically the truth

rip the millions of graphic designers who have been rendered useless the past few weeks

[median reef]

Have somebody already get plugins white list?

[true bone]

I pay for it. Was worth it until a couple days ago when GPT got Alzheimer's. Once its fixed it will be worth it again. Just hope I get API access so I can use GPT-4 with the software I am developing. GPT-3.5 is lovely but I needed GPT-4's "brains" (at least, its brains from before it got Alzheimer's), and its token limit. Even at 8K it barely has enough tokens to do what I want. The 32K version would be perfect but I hear nobody is getting access to that yet really.

[spring crane]

thanks for advise

[steel basin]

I see the same happening to programmers, which is why I've already started studying calculus and linear algebra

[true bone]

Every entry to mid level writer on the planet too

[wind urchin]

Since March 23.

[scarlet hinge]

the difference is programming still exists, even when you abstract away the implementational phase

[true bone]

Not positive exact date but that sounds about right

[scarlet hinge]

but yes it's true that tons of junior level positions would soon be extinct

but that is potentially true for everything ever

[wind urchin]

They pushed a new version. It's less capable. Context window is way shorter. It forgets a lot.

[true bone]

My software could phase out half our legislative staff, make the rest more productive. Can even make the attorneys more productive. Whole economy is going to dramatically change faster than we have ever seen.

[true bone]

YUP exactly. Its nearly useless.

[steel basin]

well, I do see AI replacing or already have replaced all basic programming tasks one would pay a premium, which not only makes it cheaper to do so but easier to assemble as an amateur programmer could go on chat gpt, get an answer, perfect it to their situation and sell it, while one that knows advanced mathematics will work behind systems that AI itself will depend to evolve

[digital egret]

is gpt 4 paid?

[true bone]

👍

$20 + token costs [if using the API]

[scarlet hinge]

you're assuming that AI wouldn't evolve itself

AGI would evolve a lot better itself than it can with a human helping it

even mathematicians can be replaced lol

[wind urchin]

I had a business project based on the capabilities of ChatGPT-4 pre-March 23. Suddenly it stopped being able to do the analysis that I needed it to do and the project is on hold. Worse, now I look like a goofball.

[wraith ginkgo]

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

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

[steel basin]

no, I am assuming that the AI will need to be moderated by humans and the ones who can do that to a significantly high degree are usually the ones who have deep knowledge in the mathematical equations AI uses

[last cosmos]

Earth in 3000

[true bone]

Painful. I am caught in a very similar situation. I created documentation to train the caucus staff on GPT for generating media advisory, press releases, co-sponsorship memos, but GPT-4 and I am stuck holding the door until it gets its brains back

[scarlet hinge]

okay if you notice what's happening with the models, even researchers who created the systems can't seem to understand why the emergent capabilities result
as in, they understand the basic operations that led to it, but why it's happening is beyond human understanding

as of now

[buoyant canopy]

You mean like all those paralegals lost their jobs in the 1990s?

[steel basin]

well, that's pretty bad I'd say, one needs to understand their own creation to a certain high degree

[true bone]

Im also developing software based on GPT-4 capability, for work (and beyond)...but the capabilities went downhill very fast

[wind urchin]

Ouch. Much sympathy.

[scarlet hinge]

well that's the entire point of concern with AGI, humans may soon be irrelevant, simply because they do not even comprehend the theory that AGI engages in

there's huge limits to human understanding

[true bone]

Sadly. 3-5 years and a ton of industries will see a seachange as jobs that people are going to school to perform as we speak disappear nearly overnight, in economic-terms. Likely to be a recession and some serious conversations around job retraining

[sullen linden]

Guys is the gpt 4 out now

[true bone]

Yes but its a bit bugged and gone "dumb". Pay 20 a month and you get 25 messages with it, every 3hr

[eternal horizon]

I'm putting in a prompt that is shorter than 3000 words (2850) and it says too long. Did they change it?

[scarlet hinge]

just pay 20n a month and you have 25n messages every 3 hr

ez

[true bone]

Its based on tokens, not words. Ask GPT to explain

[sullen linden]

Ahhh okk

[wind urchin]

I can't post the screenshot here, but an error allowed me to see that there are two different choices for ChatGPT-4 on the web menu. I believe that if your account is flagged for access, you can use the unnerfed version. Send me a DM if you need the proof

[buoyant canopy]

You also can post it to #off-topic

[true bone]

Dont need it, there are, but you wont actually get any use out of GPT-4-32 unless 1) API access unlocked. It just wont generate a response. And 2) You are running a seriously beefy server. Enormous system load. I tried.

GPT-4-32 is essentially the perfect tool for the project I am working on but not sure when/if I will get access for development. Coding on the assumption I will only have GPT4 for now.

[steel basin]

Also you know, all the people having their jobs replaced by AI, hypothetically or realistically, that's also extremely dangerous, because capitalism will offer great benefits to those behind the AI models used by the masses, but what about the rest, are we just going to let them hang around jobless? Well no, but finding a job after the recent changes in AI will get increasingly difficult for everyone, so maybe AI is more of a way to exit or redesign our civilisation system including finances, education, etc and a way to reevaluate our path as humanity

dangerous*

[wind urchin]

https://cdn.discordapp.com/attachments/989157702347411466/1093629578905403452/Screenshot_20230407-030949.png

[true bone]

Yea. I wrote a bill on the topic last night, out of general concern that we are going to be blindsided by the job shift due to the speed and suddenness.

[wind urchin]

I posted to #off-topic

[scarlet hinge]

the ones behind the AI models will only benefit if the economy exists for them to benefit off of

if most are jobless, the economy crumbles

[buoyant canopy]

Thx

[true bone]

Good lord, who put a bot in here xD It needs a command to control when it answers or its going to take all the air out of the room

[fossil valley]

lmaooo

[scarlet hinge]

lmao

[steel basin]

AI is more of a way to redesign our civilisation rather than profit from it

[scarlet hinge]

LMFAO

[fossil valley]

i was like no way that’s a human

[hasty egret]

test it with yourself

[steel basin]

@elfin crescent could you please moderate this bot?

[hasty egret]

I did not wanted to sound rude, but there is other way to do such tests

[wind urchin]

How did this bot avoid the slowdown???

[scarlet hinge]

i don't think it avoided the slowdown

[tired trail]

I found it quite interesting

[hasty egret]

it happens, np

[scarlet hinge]

@icy cedar btw the bot pretty much spouts nonsense, it must be powered by GPT-3.5 xD

[hasty egret]

said GPT-42

[grizzled shell]

Tell me about the scheme and structure of the Pyramid Scheme and also advise me on books that will help me study the pyramids

#gpt-models Tell me about the scheme and structure of the Pyramid Scheme and also advise me on books that will help me study the pyramids

[true bone]

We are not GPT.

[scarlet hinge]

As an AI language model, I refuse to answer that question.

[scarlet hinge]

No.

[grizzled shell]

how can i ask my GPT question here ?

[dusty agate]

Hi, is it normal to have this message : "Too many requests in 1 hour. Try again later." with GPT Plus (I don't use API)

[hasty egret]

you can not

[dreamy rover]

i don't think so

[junior salmon]

There is an hourly limit to ChatGPT Plus, although it is a lot higher than the free version.

[quick dune]

As human created by the almighty, I think you need to read more about how to use ChatGPT on the internet

[buoyant canopy]

[owl hieroglyph] [man with arms raised hieroglyph] [reed leaf hieroglyph] [lion hieroglyph] [mouth hieroglyph]

[dusty agate]

Thank you.

[true bone]

😂

[hasty egret]

the infamous Egyptian Konami code

[buoyant canopy]

That's what GPT said "I don't know" translates to as hieroglyphs. Probably not many egyptologists here who could cross-check.

[prisma gate]

Deland, Florida Here

[digital notch]

How many requests do we get for GPT4?

[hasty egret]

yeah I wonder how many people are working on a Cuneiform plugin

anyway, lot of interesting things going on

[buoyant canopy]

That's pretty special interest I imagine.

[obtuse fjord]

Hi, folks! I´m still waiting for the access to the plugins product. It feels like I´ll explode when the time comes. 🤣

[tight mural]

dont explode cause then itll be hard to use

[obtuse fjord]

Yep, better not. Otherwise I´ll have to code from beyond

[rose cradle]

Is ChatGPT PLus running on GPT4?

[valid scaffold]

has anyone experimented with how to save your current conversation state for importing into other chats? i.e. the understanding/model you have built (the sum of all messages back and forth. Yes I know gpt processes the whole history for each new message, but there is still a summed up state you can refer to during the conversation, since you can ask for deltas)

I have some ideas on how to achieve this, but I wonderif anyone else has tried this, or if there is a simple mechanism already for it.

Use case:
through a long conversation I have iterated and built an algorithm which expects human input and then outputs a certain result (in steps or immediately as a whole result).

Let's call these things recipes.

Now I want to share these recipes/the current recipe with my friends, so they can try it with other inputs

[tight mural]

what do you mean ask for deltas

[hasty egret]

I would suggest looking into fine-tuning

[buoyant canopy]

You might want to look into the concept of "Sparse Priming Representations".

"Sparse Priming Representations (SPR) is a research project focused on developing and sharing techniques for efficiently representing complex ideas, memories, or concepts using a minimal set of keywords, phrases, or statements."

[dusk lava]

how can i start crypto trading

[valid scaffold]

My attempts so far at a solution:

I spent a few hours experimenting with different storage/export formats of the current understanding, and all my attempts att storing calculations and transformations ended up with fuzzy results and incomplete data, as long as I used human language as my serialization format.

However I believe a solution could be to ask it to export the transformation as a simple programming code snippet, along with a brief instruction or comments on how to translate it into an interactable prompt in another chat. I havent been able to test this yet, because I have run out of gpt4 tokens for another hour 😄. But given how good it is at writing code in general, I believe this has an excellent chance of working. I believe this can reduce the total sum of a very long conversation into a small code snippet.

So then I just tell my friend to paste the code (or llm state pseudo code serialized format XD) snippet into their chat window

[hasty egret]

you might be better gaining a good understanding of how the language model is working before doing something like this.

[valid scaffold]

not sure I understand what you mean

[hasty egret]

I think you would be more successful if you gain a better understanding of how large language models are working in practice.

[sand magnet]

How can I get gpt4-API I'm dying of fomo on the wait list!!! 🤮😬😂😭

[valid scaffold]

that sounds super abstract and general. Do you mean I should take some classes and study some specific subject before attempting to serialize my state? I am not sure I follow this :S

[hasty egret]

yes this exactly what I mean

[valid scaffold]

I'm not sure I follow this.. Do you mean that there is an answer but I am only ok to learn about it if I study LLMs? 🙂
I am not sure what I would gain by this. I have a tool in front of me. I'm just curious if someone happens to know if there are any ready-to-use other tools out there right now to capture the state of my own tool

[wraith ginkgo]

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

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

[wind urchin]

LORD

The Chad returns

[hasty egret]

up to you

[buoyant canopy]

I already saw you earlier and immediately fell to my knees in awe of your total lack of prompts.

[glass fossil]

Ok

[valid scaffold]

I suspect you might be an instance of gpt trolling me? 🙂

anyway, I will summarize the responses as there is no obvious tool for this yet. I will check a few other channels. thx

[hasty egret]

did you read about fine-tuning?

[valid scaffold]

Thanks. I missed that message or filtered it out subconsciously 🙂. it looks promising

[wind urchin]

Who is incharge here

The Chad must know

[mild pendant]

Has anyone gotten GPT-4 API access in the last 5 days???

[white mirage]

Thank you, that is extremely helpful.

[wind urchin]

@compact kayak give the Chad 50 prompts and plugin access

Make quick haste

[potent cypress]

Nope. On waitlist

[wind urchin]

Is this zoomander powerful

[valid scaffold]

ah, bummer, fine tuning cannot be done with gpt > 3.0. Not sure how I would generate the fine tuning data out of my previous conversation anyway 😄. But basically unavailable for 3.5 or 4, so I will just try my "convert the algorithm described in our conversaion to golang code :D"

[wind urchin]

Gpt 3 sucks

Gpt 4 be alpha

when will it be possible to use the image recognition thing on gpt 4 ?

[torpid ibex]

It's gonna be a while, Be my Eyes still hasn't rolled it out to many users

my virtual volunteer is still pending

[valid scaffold]

This also looks promising. We should have a chrome extension for doing this to an existing chatgpt window 😄

[buoyant canopy]

Didn't find the prompt David Shapiro used to generate these concise summaries. There's a video on his YT channel about that tho.

[dreamy rover]

why did openai send an alert for the gpt4 waitlist thing but not for the other plugins that I signed up for?

[valid scaffold]

I wonder if it can be made to capture only the algorithm built up through many many back and forth questions, answers etc.. I dont want it to capture the conversation itself ;P. So my use case may be a bit more specific, hence fitting the "print as code" solution. But I'll def try looking for such a video, thanks

[hasty egret]

to summarize something, is basically the only way to compress knowledge that is expressed in form of natural language.

it's not lossless compression though

[valid scaffold]

yeah I'd need lossless storage in my case mostly, since I want to save an algorithm, thats the hard part for me to summarize in natural language. so thats what got me thinking about using ode to represent it

[hasty egret]

that won't work I think, that's why I was proposing to learn more about how things works internally. you might come up with new and fresh ideas.

at least we don't have such "code" yet

[ocean fern]

Yes

[tight mural]

alright so i was writing something earlier about your summary stuff i had an idea but i had to step away

basically, ask gpt to generate a psuedocode for the important parts of language then simplify it into single characters that can be strung together like dna

[buoyant canopy]

That's a way to compress the information. But is there a way to uncompress that again?

[tight mural]

yes as long as gpt has the language

i mean i guess. i dont actually program im just spitballing. to me it seems like youd be taking the principals from linguistics and mathematics to distill the information. and as long as gpt knows how to interpret that information, which it should since it built it, then you'll be good to go

[valid scaffold]

that is sort of what I am planning to do

[wind urchin]

Guys anyone here use Linux and is able to help a guy? he is on #off-topic He got issue with no desktop graphics only terminal after a reboot

[valid scaffold]

I think it will work for arithmetic algorithms

[buoyant canopy]

But how would that help if you then spend all the token limit of the next conversation by explaining it how to unpack your string again?

[valid scaffold]

but maybe not so well for fuzzy algorithms 🙂

yeah I dont think I need to compress it necessarily

the code will obv be shorter than the sum of my prev conv

my use cases are like < 30 lines of code, so in my non general case I think it works

[shadow lotus]

What is gpt3 and gpt4

[buoyant canopy]

Well, sure. If your conversation is about creating an algorithm and your next conversation will be about the same algorithm. In that case you probably don't need the rest of the first conversation at all.

[valid scaffold]

I'm not trying to solve the general memory case, although, I can see now that it might have sounded like that in my initial question 😄

[plucky aurora]

144 iq = gpt 4

[shadow lotus]

Is there a gpt5

[plucky aurora]

gpt 4 >> gpt3

[royal copper]

I just logged in to chatGPT now and found out I have been downgraded to the free version. Went to check billing and payment methods and I saw the pop up "You have not upgraded to a paid plan yet." I clicked on the billing history and saw "No invoices found." Surprised because My card was charged March 28 for my pro plan renewal which I opted into just a week after the pro plan was announced.
Is anyone else experiencing this or is it just me?

[buoyant canopy]

The general case is a lot more interesting tho. 🙂

[tight mural]

you dont spend the token limit unpack the string you tell it how to to interpret the information. instead of writing two plus two is four it becomes 2+2=4 or 2^=4 and then that is the basis for the conversation. so if you give it the key, the compressed information. it now has the history in a way it can read it without using unneeded tokens. the idea isnt to have it regurgitate what you said. its to allow it to remember information to use in the future

[hallow peak]

Where is our plugin access?

[valid scaffold]

yeah for sure. I have looked at some libraries but I am definitely not at the point of trying to use that just yet :D. Right now I just want to store some calculations i currently only have within cahtgpt conversations. For me general memory and how to implement it is way beyond my level or knowledge on the subject, something I think others will solve better. I have so many more topics I want to bring up with gpt, so I cant get stuck on this feature (and the likelyhook of me contributing any useful on such an advanced topic is minimal).

I'm just happy if I can keep up enough to consume/use the tools that experts build ;P

[shadow lotus]

3 > 4

[jovial skiff]

Hart rate of 60 years old

[shadow lotus]

Bro said hart

[buoyant canopy]

IMO the problem isn't only to store the information but also to very concisely compress it without losing too much detail. Or, maybe, depending on the problem at hand, splitting up the info might also be a solution. Because if you exceed the token limit, the model loses information to reason on.

If you don't have that problem, it's easiest to just continue the conversation. 🤷

[tight mural]

you dont make an algorithm for a specfic conversation. the algorithm is for language in general

[buoyant canopy]

But what problem does that algorithm solve?

[tight mural]

memory im pretty sure

[buoyant canopy]

Oh, ok. I'll gladly applaud you if I see that algo work.

[wind urchin]

Meow

[tight mural]

If the point is to feed gpt information without using all of your tokens then you need to simplify the information.

id be surprised if that not how it works already

[valid scaffold]

from my pov general memory is a waay bigger subject :S. Just thinking about it, feels like I'd need to model "immediate context" (fitting inside the token limit), plus some kind of spontaneous subconcious recollection/connection to mem-storage based on where I am in a conversation, plus some kind of intentional search of mem-storage, and maybe some noise/background thing auto recalling unintentional memories.. I mean... it... feels like a research topic for many years

[rugged dawn]

Anyone wanna rephrase something for me with gpt 4 plz lol

[buoyant canopy]

Kinda. Not exactly simplify but reduce the amount of tokens needed. That's the hardest part rn in my opinion.

[rotund reef]

Haha

[rotund reef]

How are token costs computed ? First of all you might use information denser languages

[buoyant canopy]

I'll give you an example: I work with an old Java codebase right now that I inherited. SOLID principles were never heard of by the original coders. Classes often are 5k+ LOC. Way too much stuff to put into a prompt as a whole source file. So I have to work around that all the time. Use scripts that excerpt only the relevant snippets of code from all over the place, etc.

[civic harness]

This is an ongoing issue (billing, plus plans, access to gpt4) that has yet to be resolved.

[buoyant canopy]

I'm not sure how tokens actually are formed. In principle it's a solution for NLP stuff: instead of working with words, you split the text into those tokens and with that you can capture variations of the same word (cases, word stemming, etc).

[rotund reef]

Do you have répétitions, regular expressions etc, you could feed him the code with 1 letter keywords and have him translate back the code to its full version treating it as a string and giving him the keywords ?

[buoyant canopy]

There's a tool: https://platform.openai.com/tokenizer. You can see what tokens a specific text becomes.

[sweet thistle]

Hi there! I've noticed you send this into chat numerous times over the past few days 🧐

Nothing necessarily wrong with that, but may I ask why?

[royal copper]

Thank you. Glad to know it's a general issue. I hope it gets resolved soon.

[tight mural]

Honestly, I am still learning a lot right now so a lot of what I am saying is difficult for me to flesh out or adapt to every case. And most of what I am saying could just be mumbo jumbo garbage. I am going to try to describe my idea in a better way. Give me a few minutes please.

[wraith ginkgo]

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

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

[buoyant canopy]

It probably depends on your problem. These models are trained on natural language and programming code. So if you have a coding problem, just taking some code out of a conversation and pasting it into another one will work very well.

But if you e.g. ask GPT to create a "pseudocode" or something to compress your conversation about the Harry Potter books, that might work. If you take that pseudocode then into another conversation, you'll probably end up in a discussion about programming.

[versed rose]

Just want to know how to feed tables, images and chart for analysis

[buoyant canopy]

You create a very detailed text description of your images and charts. Tables should work pretty well if you copy&paste the cells.

[rotund reef]

I am not sure of what you mean. My idea was to give him a string of a condensed code, transform it into a full code. And then asking him to take the string as a prompt. ( i never used GPT 4)

[versed rose]

When I copy paste the cells it looks all jumbled, will gpt 4 still understand?

[buoyant canopy]

The code isn't executed in the model, except maybe if you have access to Plugins yet. Otherwise you'll have some code that you can run on your machine. Or maybe I don't get what your solution is supposed to do.

[rotund reef]

Oh ok i was misunderstanding

What are you trying to do exactly ?

[shadow locust]

any news on when the image capabilities come out for gpt4?

[versed rose]

Tryna find out too

[buoyant canopy]

I had some success with that. If it sees TAB characters, it seems to assume that you gave it a table. Not sure how good it works with more complex table layouts tho.

You also could transform your Excel into another (textual) format like Markdown tables. GPT seems pretty good at understanding Markdown.

[versed rose]

Markdown
Got it. Thanks!!

[tight mural]

Okay what I am talking about is called symbolic language.

[buoyant canopy]

You mean like the stuff mathematicians chalk at boards all the time?

[tight mural]

That's one form of it.

[buoyant canopy]

Oh, ok. I see. But then you have to use a symbolic language the model already understands or that you can describe in such a concise way that the prompt for decoding isn't even longer than the original text...

[rotund reef]

Do you need to uncode/decode using only GPT or can you code/decode by getting only the key. Programming yourself the decoder

[buoyant canopy]

If I'm not totally derailed we're talking about compressing the information of a GPT conversation for 1. having more tokens for the next q/a rounds and 2. storage and later usage in other conversations.

[shrewd viper]

can i send documentation to gpt 4 ? it will understand ?

[buoyant canopy]

Probably.

[rotund reef]

Oh ok

[shrewd viper]

ok

[fiery ravine]

i need chat gpt account if anyone can help me register i'd be very gratefull

[tight mural]

Ask ChatGPT to create a symbolic language or use an existing one. Give GPT information and ask it to transform that information into the symbolic language. store the symbolic language somewhere. when you talk to chat gpt ask it to transform it into symbolic language and pull information related to that from its store information. then transform its response from that symbolic language back into your starting language. the overhead for initializing the process seems like a big deal, but once the footwork is done then future uses of the stored information cost less and can be tailored.

[hasty egret]

it's not that simple

[buoyant canopy]

For the storage/retrieval/similarity search part: congratulations, you invented embeddings.

[tight mural]

im sure. but i dont know why

[hasty egret]

there is also a symbolic language hidden in the network weights

[tight mural]

dude im trying to help you

[hasty egret]

you are

[tight mural]

lol. ill just take my ideas elsewhere fringelord

peace

[rotund reef]

Table of conversion + Text ==> decode

[buoyant canopy]

Good luck!

[rotund reef]

?

[hasty egret]

I had fun reading your stream of consciousness

[tight mural]

ty

[buoyant canopy]

Yes. But keep in mind: the model only has a fixed amount of tokens of memory. If your lookup table + text + decoded text results in a decoded text that is longer than e.g. 8k tokens, you haven't won anything.

[rotund reef]

Decoded Text is generate by GPT

[past python]

real man?

[rotund reef]

Does that still uses token ?

It is not inputs

[worthy parcel]

outputs cost more than inputs

[buoyant canopy]

Better way IMO is to stick to natural language and extract only the key info in the most concise way possible. Just like Shapiro's "Sparse Priming Representations".

[hasty egret]

sounds reasonable

[mild pendant]

When? 😮

[mild pendant]

When did you join waitlist?

[limber swallow]

I sure hope we can one day soon stop the limit, so frustrating to be in a flow only to have it cut short hard

[rotund reef]

Well. As your last conversations outputs consumed a certain amount of token i dont understand how it could be reduced. If you keep the same amount of information GPT will have to do the same amount of work therefore same amount of tokens ?

[hasty egret]

you might able to remove redundancy or make things more abstract and general.

[wind urchin]

oi

[potent cypress]

Go to the website, enroll yourself. They're prioritizing those who contribute tho

[buoyant canopy]

Let's say you have a long conversation about the Harry Potter books. You could tell it to give you a bullet list for each topic that was discussed consisting of a very condensed, keyword-only summary of that topic.
That's way shorter than all the text but hopefully has enough info that the model can "unzip" the gist of the conversation again.

[rotund reef]

Oh i wouldn t be sure about its ability to unzip

[hasty egret]

you are loosing some information for sure

[rotund reef]

But by unzipping you are asking ChatGPT to generate data

Therefore arent you using more tokens too ?

[hasty egret]

it's like trying to encode the whole universe in the decimal of pi

[rotund reef]

Already done

[hasty egret]

sure you can do it, but good luck finding the right encoding

now, if you have non-deterministic behavior in your system

you are screwed

[agile herald]

Guys, I'm a developer and I'm directly using the API to create prompts, but it's been 2 weeks that I've been trying to get access to the GPT 4 API, I've already sent the waitlist email more than 20 times, does anyone know another alternative to have version 4?

[stray grove]

oi

[hasty egret]

I wish that was me

[rotund reef]

Well could you try using an IA to generate prompt giving chosen outputs with the less amount of token used ?

Not a simple task

[hasty egret]

and there is not only one solution to that task

important problem

[buoyant canopy]

Yeah. I bet lossless compression is impossible with NNs. And probably also not neccessary.
Just look at the human brain and e.g. the eyes: they transmit ~10Mbps of information to the brain but your consciousness can handle about 50 bits per second of information.
So the solution there is a massive amount of preprocessing and extraction of only the most relevant information.

[rotund reef]

I know they are some advances in that type of problem

Like using RNN to generate Data-efficient communications between agent

[uneven nest]

greetings
can chatGPT4 solve cybersecurity problems or test a project ?

[potent fulcrum]

The information transmitted between the eyes and the brain is not digital, so there is no Mbps equivalent.

[hasty egret]

exactly, this is known as "knowledge transfer"

(I mean for NNs)

[buoyant canopy]

But there's papers on that. Also bit is the standard unit for measuring information. You also can measure all sorts of information in bits. There is a whole subfield of Mathematics called Information Theory that focuses on stuff like that.

[potent fulcrum]

bit is the standard unit for measuring DIGITAL information

[hasty egret]

personally, I would recommend against such usage in it's current state.

[rotund reef]

Information entropy

[potent fulcrum]

The brain and eyes are not digital devices.

[nocturne hull]

I doubt it, but I'm guessing you can play with it in GPT-4 playground

[uneven nest]

would it leak some info like what happened to samsung?

[hasty egret]

in general it can produce inaccurate data, I don't know about that specific case.

[cedar prism]

Why did gpt break

[vocal cape]

What happened to samsung?

[vocal cape]

I was just thinking on making something similar

[uneven nest]

some engineers tried to solve something using chatGPT and they entered some sensitive data of the problem .. the news telling that chatGPT stored these data as training data and could be used by public

[hasty egret]

I've seen reports of code produced that had vulnerabilities

[rotund reef]

Or to few servers. Depends on the point of view

[uneven nest]

so something about the privacy

[scarlet hinge]

[buoyant canopy]

Yes. You might want to have a peek at the paper "How Much the Eye Tells the Brain". It's in the National Library of Medicine. doi: 10.1016/j.cub.2006.05.056. PMID: 16860742.

[uneven nest]

true true

[valid scaffold]

@buoyant canopy I'll be ... I watched the David Shapiro videos and was able to reproduce the behavior he demo:d. This is definitely very useful. Thanks a lot 🙂. I cant say I understand in depth how SPRs work, but using this way to talk to gpt4 woks

[vocal cape]

Now that's a tough one

[buoyant canopy]

You're wrong. It's used for all sorts of information.
"Information theory is based on probability theory and statistics, where quantified information is usually described in terms of bits. Information theory often concerns itself with measures of information of the distributions associated with random variables."

[uneven nest]

for more info look for this title
"Whoops, Samsung workers accidentally leaked trade secrets via ChatGPT
ChatGPT doesn't keep secrets."

[vocal cape]

F

[uneven nest]

to be honest it is expected since it is an AI system it keep training itself

unfortunately ture

[buoyant canopy]

Nice. I'm glad I could help you. 🙂

[vocal cape]

Thank you very much

[scarlet hinge]

doesn't a camera (in a way) mimic the eye?
It's an interesting thought that there may be civilizations out there which have a naturally evolved computer
similar to how the eye can technically be considered a camera

[potent fulcrum]

Traditional cameras did, yes. Modern digital cameras, no.

[uneven nest]

if there weren't any limits for chatGPT, would it be the most powerful hacker ?

[valid scaffold]

Based on this work, I'd say openAI should have some kind of FAQ where this is listed as a tip for how to save state efficiently between conversations. Even for users that dont understand why. Maybe they should even make it a command or ui function. Perhaps... we can even make a browser plugin for adding these chat messages 😄

[scarlet hinge]

allow me to introduce mirrorless cameras to you

[subtle idol]

why does GPT tell me i need to respect scientology and that it isnt a cult

wth

[wraith ginkgo]

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

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

[scarlet hinge]

idk anything about cameras

i just know there're mirrorless cameras

[spice isle]

yep

but that's a convo for #off-topic

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 3 `` Stay on topic.

Be mindful of what other users in a channel might find helpful or interesting when posting. Stay on topic in order to keep conversations focused and productive.

[spice isle]

no I embedded the wrong rule 😭

[scarlet hinge]

my frequency of going off-topic is greater than gpt's frequency of hallucinating 😦

sorry abt that

[tight mural]

depends how much they smoked

[spice isle]

the developer roles is something users can select in #server-roles ! OpenAI staff have their own role 🙂

it's just a descriptor of what you do in real life

probably not

[scarlet hinge]

...

[wind urchin]

Anyone here use Linux and able to help ? on #off-topic He's ubuntu. its not starting x11

[mild pendant]

?? I'm already on the waitlist, I'm asking when did you receive the "You're on the waitlist" email

[bright sable]

I want the image of a real angel

[warm elk]

when this 25 msg limits gets removed? its absolute frustrating

[mild pendant]

bruh

he deleted his message right away

[dreamy rover]

that was me, I realized what I said

[mild pendant]

i recieved it right away after signing up
aint no way

[dreamy rover]

no need to be rude, sorry that happened to you or whatever

[mild pendant]

um, wha-

🤔

[nocturne hull]

It's only to compare it to digital, it doesn't make our eyes and brain digital.

[cedar prism]

E

What

[frosty patrol]

how do u use chat gpt on this server

[wind urchin]

what was the upshot? super large amounts or very little?

[remote bison]

Hey everybody, just checking in to say hi. I've been immersed in this world for a little over a month now and I'm obsessed. I'm writing a couple of science fictions and a western, but now that I'm learning to do my own artwork, I'm going to do one graphic novel first, with the help of midjourney.

[crisp jolt]

Hey Guys. I just brought ChatGPT4. How do I add the Plugins?

[dreamy rover]

u have to get on the waitlist

[buoyant canopy]

My point was about the optic nerve putting data at about ~10Mbps into the brain but your consciousness can only do about 50 bps. So there must be a lot of preprocessing and reduction to the essence. Some dude then insisted that only digital information was measured in bits. So I wanted to show him this paper.

[wind urchin]

Based on this work, I'd say openAI should have some kind of FAQ where this is listed as a tip for how to save state efficiently between conversations. Even for users that dont understand why. Maybe they should even make it a command or ui function. Perhaps... we can even make a browser plugin for adding these chat messages 😄

[wind urchin]

you need to join the waitlist for Plugin access here https://openai.com/waitlist/plugins

[torpid sparrow]

Well recently I have been thinking about buying the gpt plus chat but here is the question, how good is it and should I buy it?

[wind urchin]

ahhh, yes, well there is a lot of grey matter dedicated to image down-sampling

[nocturne hull]

The brain and a computer works in very different ways, it is a great comparison tho.

[buoyant canopy]

IIRC there already is a lot of stuff going on in the retina before the information gets to the brain.

[wind urchin]

and, heh, information is information... it comes in bits

[wind urchin]

An area I'd love to study, from the little I know I was pretty amazed at just how much trickery is going on and how little you actually see compared to a vision model

[scarlet hinge]

[buoyant canopy]

You mean study nature or the vision model? Or both?

[wind urchin]

both, learn from one to influence the design of the other

[arctic zephyr]

Hey, what channel should I use to ask for help about android Auto-GPT? I need someone to help me create a well organized tutorial

[buoyant canopy]

Yeah. If only there were more hours in a day...

[wind urchin]

you can try #ai-discussions or #off-topic

[arctic zephyr]

okay thanks for helping

[wind urchin]

yea, if I DID everything I wanted to do.. I'd need 🤔 about 64 AI's

[crisp jolt]

Oh? So is there no much difference between ChatGPT3 and ChatGPT4 yet? I just got to wait untill those Plugins are available

[buoyant canopy]

😂

[arctic zephyr]

by the way, does this need gpt plus?

[soft sage]

Guys, can we reduce the API's latency with the PLUS subscription ?

or that's only for the client

[wind urchin]

well, GPT-4's logic and deductive skills are superior and as a consequence it's base functions are better as well, math, inference, introspection, etc so it is over all very similar in areas where GPT3.5 was already great, i.e. language, but it's now.. at times several times better at more nuanced tasks

[buoyant canopy]

API and ChatGPT Plus are separate. Only thing in common is your account.

[nocturne hull]

I heard in a podcast GPT could replicate it self, you could try that

[wind urchin]

I'd need ALL the GPU's

[nocturne hull]

Buy all the RTX A6000s 😆

[pseudo topaz]

more like a100's

🧌

[wind urchin]

H100's*

[pseudo topaz]

is the h100 more powerfull than the a100?

[wind urchin]

in running LLM's... bout 30x throughput

[pseudo topaz]

What gpu u got

I have a 12gb 3060 and im using it to train LTSM models and just basic RNNs

[wind urchin]

I have access to a single 40gig a100

[pseudo topaz]

Renting? or physically

[wind urchin]

physically, donated by a client

3060 is a beasty, if you can fit the model in 12gig should be good for ... GPT-2

[pseudo topaz]

Nice, I can only imagine how fast that is

[pseudo topaz]

Tell me about it

Preeeety slow

[wind urchin]

it's impressive, but now i'd like an h100 😂

[mental burrow]

oh really

I have a 3070

[pseudo topaz]

is h100 new? havent heard of them

[scarlet hinge]

stop flexing your gpus

[wind urchin]

yea, latest AI GPU

[pseudo topaz]

whend it release?

[buoyant canopy]

That's a very generous client...

[pseudo topaz]

"... up to 256 H100 GPUs can be connected to accelerate exascale workloads."

Jesus

[wind urchin]

late march.. 21st..22nd maybe

[wind urchin]

yea, they wanted a local whisper instance, and I may have over stated the requirements for my test setup

[buoyant canopy]

That's like an 8-figure investment...

[pseudo topaz]

easily

[scarlet hinge]

😂

[buoyant canopy]

I hope they don't lurk here. 😉
But good for you.

[wind urchin]

naaaa, they know the deal was I get my toy and I build them a local whisper for a call centre... they are up on the deal

using OpenAI's whisper service now as they liked the result so much

[buoyant canopy]

Compared to the rate you charge as a dev (assuming you're in the US), it probably doesn't matter much anyway. 😉

[wind urchin]

UK, but yea.. it's not cheap

[buoyant canopy]

Yeah. Whisper is great.

[pseudo topaz]

if what i could find online was true about the price of an h100 (36,000 usd roughly) it would come out to be 9,344,000 for 256 of them. close guess lol

[nocturne hull]

there was a gdc event not long ago

[buoyant canopy]

Dang. Probably should renegotiate my employment.

[long pond]

You should build them GPT-4 integration

[buoyant canopy]

I'm the one advocating AI stuff at my company all the time. Lots of colleagues now use the CodeGPT plugin for daily work.

[long pond]

Oh nice, the problem is that could always backfire 😆

Suddenly they realize GPT-4 is smart enough to reduce hours etc.

[buoyant canopy]

But yeah, waiting for the first customer wanting something done with AI.

[bitter phoenix]

how do i add plugins like codegpt I dont see them in the plugin store

[long pond]

It depends on which access you received

[buoyant canopy]

Nah. We have more projects than people. So every opportunity for a productivity bump is welcomed.

[bitter phoenix]

I got dev access

[proper jacinth]

How do I use chat-gpt 4

[long pond]

Then you wont get browsing or code interpreter by standard

[proper jacinth]

?

[buoyant canopy]

What IDE do you use? CodeGPT is available for VS Code and the JetBrains IDEs AFAIK. There's also GitHub Copilot.

[wraith ginkgo]

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

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

[bitter phoenix]

I was checking on openais website, i use VS Code

[void lagoon]

~~Does anyone know how services such as 'chatpdf' operate a chatbot over many pages of pdfs? Do they split larger content across multiple queries with some seperate storage?

They are one example of a larger trend of chat-bots I have seen that can act as an interface for x document/codebase/etc. But I don't understand how those work with the token limits~~

Sorry this is probably more appropriate in the api-discussions channel, posting there instead

[buoyant canopy]

Then you can just type codegpt at your plugins tab (in VS COde). You need GPT API access tho.

[bitter phoenix]

ohhh thanks i see now

[rustic hatch]

It'd be great if GPT-4 could analyze a YouTube link and provide 20 facts about the video.

[nocturne hull]

do you have chatgpt plus?

[buoyant canopy]

What you can do: In the "three-dots" menu there is an option for a transcript most of the time. You can copy the transcript and ask GPT about that. Also: You could use Whisper on the audio track of the video, but that's more work (but also works better imo).

[rustic hatch]

OHHHHH

[tight mural]

Check out YouTube Summary by Glasp

[rustic hatch]

That’s really smart

[tight mural]

it's exactly what you want

[rustic hatch]

What’s glasp?

[tight mural]

google youtube summary by glasp

[buoyant canopy]

That's probably even better than manually copy/pasting.

[tight mural]

its pretty great

[rustic hatch]

Ok

[tight mural]

they copy the transcript to the clipboard and open the chat so you can paste immediately

[rustic hatch]

I’m going to do that