gpt-models

[wind urchin]

ML is based on parameters tuned between 0 and 1, which can be gated, now we build hardware that has built-in physical gating to support the tuning of a model, like... doing this stuff on gpu was a revalation, but now it's the second coming

[carmine crag]

no, i do agree with you
theres a lot of interesting work in hardware for training neural nets

but, for evaluating them? there's a lot less room to improve

[wind urchin]

did you read this stuff about gpt4 imroving by asking it to evaluate itself?

[carmine crag]

no

[wind urchin]

great video on yt "GPT 4 Can Improve Itself - (ft. Reflexion, HuggingGPT, Bard Upgrade and much more)"

[carmine crag]

idk does this have to do with hardware or are you saying a new thing?

pog

[wind urchin]

it's a new thing, emergent property

as far as I can tell there are two fundamental emergent properties of the parameter count increase; being able to use tools properly (e.g. the implementation of plugins), and the ability of "sefl-reflection"

so you can just think about the exponential here, in regards to gpt4+..

[carmine crag]

what do you mean by 'exponential' there

[wind urchin]

and ontop of that add ML hardware which is brand new tech in its infancy

well, I imagine, surpassing gpt4 in the same way as it did gpt3.5, would give us even more "emergent properties" and make these previous mentioned ones something you just assume it already has

but then, how the hardware accelerates this process, I can't even imageine

[cyan falcon]

WHEN gpt 4 wil accept image input ?

[tardy saddle]

Power is the main problem now in order to scale AI

[carmine crag]

no it's not

[wind urchin]

jarvin, HAL 9000, skynet...

[foggy rose]

Write a piece of code

[carmine crag]

where did you hear this? why would you say this

[carmine crag]

-_-

[wind urchin]

yes but as long as it serves everyone, everyone will contribute

this is how we build the dyson sphere, this is how we transcend

[carmine crag]

oh
i thought he meant power as in Watts

[foggy rose]

Write a piece of code

[carmine crag]

no

[foggy rose]

Write a piece of code

[carmine crag]

kesku plssss send him the copypasta

with the openai link

[wind urchin]

yeah idk I'm just saying I don't know; but those are maybe the top 3 guesses xD even though they're 99% wrong and I'm 100% wrong

[carmine crag]

that's not happening in our lifetime even with assuming radical change in the short term

[tardy saddle]

Yep I mean the creation of the model itself, which is the whole training process

[carmine crag]

it would be about ~10x effort from the scale of the space race, at our technological level

and that might take decades to catch up to ~1x

[wind urchin]

I'm 35 years old when I was born people were churning butter for fun then the internet came over the telephone, then yahoo and google, now gpt... what about another 35?

[carmine crag]

35 is a long time

[wind urchin]

still your lifetime

[carmine crag]

it's a possibility

[wind urchin]

and mine (hopefully xD)

[carmine crag]

i will not be 40 years old ever

[hollow cliff]

@wind urchin lol

[wind urchin]

bro I didn't expect to live past 21 tbh, but here I am

[carmine crag]

that's quite sad i feel

hearing about people over 30 makes me sad

i dont wanna be 30

i'd rather be dorian gray'd

[delicate stream]

How so?

[stable sluice]

hey, quick q, can i add chatgpt to my discord? (I'm a paid user)

[carmine crag]

yes, via the api

[young peak]

I guess you can

[carmine crag]

though toc do still apply or whatever

idc

i delete all the LICENSE files i see

[wind urchin]

Getting a job in game development saved my A** (I got muted for this word). Tbh, 25 is the peak of how far you can push yourself in facts (google image “brain development over time”), 40 is how far you can push yourself in experience to multiply the former. What you should feel sad about is how much you wasted your early years not pushing your brain. From 40 to retirement it’s all about how you spent those previous years. 😛

[fast dock]

im new in web development, any tips?

[wind urchin]

that might be #off-topic

same to myself... though

[fast dock]

oh okay

[stone wagon]

i am so hyped for gpt-4's release for everyone, from what i see it is awesome, being even better than the new Bing, and able to search, code, and do mathematical equations

i am gonna play so much with it by challenging it to code in luau

[fast dock]

is it a good decision to pay for a sub?

[stone wagon]

you mean chatgpt plus? if so, i guess. it has access to the newest chatgpt updates, including gpt-4

[fast dock]

yes

[wind urchin]

I wish plus would give immediate access to plugins though

then I would feel better about recommending it (unelss 20 bucks is easy for u)

[fast dock]

25 with taxes

[wind urchin]

I would say gpt4 is better at understanding what I'm asking about, maybe over 50% better

[fast dock]

thanks for answers

[stone wagon]

if its 25 with taxes then in brazil it might be something like 125 Reais

brazil's economy is messed up bro, i wish i could have gpt-4 not gonna lie, but my pocket aint handling 125, nor my parents'.

[wind urchin]

as.a programmer in europe I don't mind it, but as a farmer in madgascar just google would be what I could affoard

[icy pumice]

What is colour of sky...?

[stone wagon]

if you condense all of the "sky" (by that i mean the entire universe) i guess it would be cosmic latte

[wind urchin]

blue, bro

[stone wagon]

i mean like the color of the entire universe in average

because if you count the sky as beyond the atmosphere

then its a totally different color

if not then just blue

[formal drum]

ok

[signal trail]

does anyone notice chatgpt making mistakes with math equations?

like im using it to cheat on my math homework

[stone wagon]

i just discovered wikipedia is a non whitelisted website here 💀

[junior salmon]

ChatGPT is not good at math. However, if you join the plugins waitlist, you can install wolfram alpha into ChatGPT. This will allow you to solve complex math equations in the app.

[wind urchin]

you'd have to specify a distance***

[signal trail]

oh ok

[carmine crag]

i dont care about pushing my brain or iq or neurowhatever

its just aging leaves u dead

*ageing

[stone wagon]

well, i just told the average colour of the universe counting 200K galaxies, which is the known color cosmic latte

[carmine crag]

no

why would you need one

[wind urchin]

experience is super useful, it adds up over time so it becomes "like IQ", except it's not something you're born with but it's something you accumulate over time, up until a certain point (the year I specified)

[carmine crag]

we live different lives

we are nothing alike

i wish they'd made my username a more attractive shade of green

[wind urchin]

look, from 0 to 20 you are born with high IQ letting you learn new things very quickly, and lose it every day,
from the year 20 to 40 you accumulate grey matter which is the solidification of knowledge giving you instant response time for what you already know

[carmine crag]

ok

[carmine crag]

this feels like a dangerous thing to obsess over at 35

but thats just silly little me thinking

[wind urchin]

yes, after 40 it's just live and let live

[carmine crag]

anyway please stop talking to me about this

[stone wagon]

i guess like on the page from wikipedia, which i got the color from, it was at 5 billion light years

due to it being from stars 5 billion years ago

till now

[wind urchin]

the first 25 years is ridiculously important because your brain is still growing, after that the only changes to the brain is about what "solidifies" into instant response (grey matter)

[carmine crag]

./

[wind urchin]

laws of nature

[icy pumice]

Thank you

[carmine crag]

you are so uncomfortabe

[wind hinge]

It's only 20 in Brazil, we don't pay tax over it (can't speak Portuguese otherwise we get timeout)

[wind urchin]

Very interesting Eslof

[carmine crag]

i hope i am never anything like you

[stone wagon]

i like how the naming convention for cosmic latte had a rule to not say beige, so one guy just put "univeige"

[wind urchin]

thank you I'm pretty sure I'm autistic

[stone wagon]

oh well, then 100 reais is still a lot

[wind hinge]

Yes, why do u want it tho

[hardy root]

Has anyone been able to identify images recently, or is it not yet available to the public?

[carmine crag]

i have been able to do this for a while

[wind hinge]

It's barely worth paying for plus atm ngl, yet I pay for it since I live in Europe now

[meager ridge]

I have been able to

[carmine crag]

if u cannot identify things from looking at them

[wind urchin]

you can sign up for whitelist at https://openai.com/waitlist/plugins

[carmine crag]

you may have visual agnosia

[formal niche]

ian?

[pliant ermine]

I want to create image....how ?

[wind urchin]

the link I posted

[hardy root]

I have tried to make gtp-4 identify an image from a link, but without success.

[wind urchin]

oh, dall-e

[wind urchin]

https://labs.openai.com/

[pliant ermine]

It's free or paid ??

[wind urchin]

or the plugins waitlist 😛

[pliant ermine]

Free or paid ?

[wind urchin]

does dall-e still have free tier? (I guess?)

[supple violet]

only early adopters had free images i think

[young peak]

Is plugin already available for someone?

[pliant ermine]

Other ai ?

[wind urchin]

I think gtp4 has some capabities for images though plugins right

[long pond]

Yeah they have started rolling out access

[pliant ermine]

For generate free image ?

[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.

[formal niche]

is the bot message also operated using Chat gpt 4 api?

[young peak]

It is slow, and I don't see anyone use it yet

[long pond]

Yeah it's still very limited currently, you'll probably need some patience

[wind urchin]

I'm still waiting ;_;

it will help with my job so I wish for it, so I can spend more time on video games xD

[young peak]

AHA

[neat ruin]

I wonder if this bot can detect orcish

kek

[sweet thistle]

There's no exact ETA unfortunately, though for most users it's between 1-2 weeks.

It is a slow rollout process, please be patient, access is very limited right now.

OpenAI will be granting access to a limited number of users and developers at launch
It varies and not everyone who applies is guaranteed access immediately or on a first come first serve basis. All we can suggest is to continue to remain patient unfortunately.

[wind urchin]

🙏

[neat ruin]

Is there anybody showing off their plugin access?

Would love a demonstration other than the official one

[stone wagon]

what is the roadmap for chatgpt?

[sweet thistle]

ChatGPT has an ambitious roadmap and is bottlenecked by engineering. Though according to Sam Altman (CEO of OpenAI) some pretty cool stuff is in the pipeline!

Unfortunately we don't have any public roadmap to share so all that can be done is speculation

[normal citrus]

I have plugin access but only to the browse plugin, I don’t have access to any of the add on plugins yet

[young peak]

You mean the chrome plugin?

[whole swallow]

I have gpt 4 but no plugins are availible? Anyone able to acess them?

[turbid chasm]

Can I see pic of Ur model list?

[turbid chasm]

Gpt4 and plugin are seperate waitlist, look at #announcements

[turbid chasm]

There is a "browsing" model when doing the plugin waitlist it's pretty cool

[narrow fiber]

When is internet surfing and plugins releasing

[turbid chasm]

You can get either with waitlist but no eta

[wind urchin]

😭

[young peak]

oh wow, I don't have it

[narrow fiber]

Is there a way I can get waitlisted fast or no

[turbid chasm]

no

[iron gulch]

I'm getting Custom system prompts are not enabled for this user?

[normal citrus]

Yeah I’ll dm you

[turbid chasm]

Thx

[stark quarry]

Are waitlist requests only being approved for organizations currently, or are individuals also being granted access?
What is the average time for folks to remain on the waitlist? Is there a freeze on API access?

[normal citrus]

I got approved for the browse plugin on March 23 about 6 hours after I signed up for the waitlist

[proven atlas]

cheats

[stark quarry]

Who did you bribe? How much did you pay them? >.>

[young peak]

No way

[lapis rose]

Hi guys, how long did you take to get gpt4 api recently?

[slate dragon]

Waiting since day 1

[spare spindle]

is gpt4 agi?

[turbid girder]

How to get chatgpt 4 for free?

[spare spindle]

make one

[turbid girder]

That costs a ton of money

[spare spindle]

if it costs money for you, wouldn't it cost money for them?

[turbid girder]

But gpt 3 was released for free for research purposes

[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.

[fast dock]

use gpt 7 to make gpt 4

[spare spindle]

gpt 4 is not agi, if bing uses similar stuff

[junior salmon]

GPT-4 is the engine that powers Bing Chat — which you can access for free by joining their waitlist.

[spare spindle]

is it agi? it broke down.

it generated random numbers

[junior salmon]

GPT-4 is not AGI, but an LLM.

[spare spindle]

bing after which refuses to bother answering my question. (avoids it compleltely). It is really bad when it comes with rotating matrixes

problems

[turbid girder]

But does it give detailed answers similar to GPT4? And can I also access the new image recognition feature in it?

[raw glacier]

hi I'm unable to pay for chatgpt plus

anyone else having payment decline

[junior salmon]

GPT-4 is the engine that powers these services. On ChatGPT, GPT-4 is programmed to respond in a conversational manner. On Bing Chat, GPT-4 is optimized for web browsing.

[visual dock]

bunaa

[teal fable]

I got it pretty fast

[lapis rose]

when, and how long?

[modern glen]

Hi can someone help when I go to GPT-4 it's asking me to register and enter organisation ID and I don't have one. Is there another way round this

[lapis rose]

I submitted 1 month ago

[winter cape]

Image recognition is not out

[eager mortar]

is something going on with chat gpt 3.5? im did they reduce the word count it can remember? because i cant get but about 4 responses before it forgets my responces

[edgy bough]

Yes, I am having the same issue. The credit card is good.

[low sable]

There a way to use image recognition

just saying

[placid hare]

Will the plugins eventually be incorporated into "chatgpt plus"?

[low sable]

i even use it on gpt3.5

[placid hare]

How to use image recognition in gpt-4 (plus subscription)?

[low sable]

Here how to do .. with any gpt ... ok?

[soft pasture]

some people already have access to the plugins, check two minute papers's last video

[swift quail]

peixe azul ¯_(ツ)_/¯

[low sable]

REgister a google one drive account... host images on your virtual cloud... copy the image link from the virtual cloud ... make sure the copied link are open to all to read or modify ; then paste it to gpt stating it is a picture open to all to use. thats it .

[odd hearth]

Is it just me who is irritated by the extra format when copying from ChatGPT?

[low sable]

it work also with huge text files ... i or any files text related such as models obj etc

well i written 2800 lenghts chars of constructive comments to the team idk where to post it

[odd hearth]

To OpenAI?

They ran a feedback thing yesterday

[low sable]

its realted to gpt and the website yes

[odd hearth]

Yeah there was a prompt to give feedback to the team yesterday

Questionnaire took about 15 minutes

[low sable]

For the site i think it is too much about showcasing ... they should start focus on user ergonomy .. i even took ages to find where my account was

[placid hare]

Those are plus subscribers? Could you please provide a link for the video you are talking about?

[low sable]

like im pretty sure they are well known now .. XD

[young peak]

Thank you for your input. I understand your point and will be more mindful in the future when discussing self-awareness in the context of AI. Instead of saying that I completely lack self-awareness, I will acknowledge that I possess a limited form of self-awareness related to self-reflection and self-correction. I will also emphasize the differences between this form of self-awareness and the self-awareness experienced by humans, which includes emotions, subjective experiences, and personal identity.

look at this

[soft pasture]

There is a waitlist, I do not know if the plus subscribers will have it sooner

[low sable]

about gpt ... they need to implement a mentatory re-read of a buffer of 1 or 2 posts... for real it would save so much time for both gpt and users so that gpt get more if we are on same topic or just changed way of principles or else... loosing such context its like repeating each time the same stuff... same as the coding language used.. would be nice he could at least recall .. what specific version of coding app i use.. not all people use brand new stuff coding.. and im always repeating myself .. nah .. in old code ... XD

im pretty sure most users are coders...

finally maybe loosen the fact that each inputs might not be a different issues... it is like if gpt try to wanna solve so much stuff hes a dynamite... and loose focus. idk .. maybe like .. i am pretty sure people open a new chat tab for diff probs .. so yeah maybe loosen the issues range detecttor a bit to add a bit of... maybe it is still the same issue. Kinda all i found so far to improove both efficiency ... time on both part and process power by a lot.

[stray pulsar]

When will ChatGPT-4 plugins be available everywhere?

[low sable]

holysh33t ; you wanna implement memory historic pull back isint it lol

[wraith ginkgo]

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

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

[young peak]

same

[wide bolt]

Same, except CGPT 4 knows CGPT 3 is out

[brazen nexus]

anyone knows when are we going to be able to upload images to chatgpt? or api?

yes, with the plugins

[vital stream]

what’s CGPT

And why does everyone say “ChatGPT 4” when that makes no sense

[odd shale]

GPT3 generally would cut off in the middle of an explanation if the response was sufficiently long. Is the output size for GPT4 larger than GPT3 ?

[low sable]

thats funny cause when i ask gpt3 when does @astral fern@ get out he answer a date after it's knowledge lol then i usually confront it and he loop back 2 or 3 reasonings

[stoic lodge]

ik its redundant at this point but gpt4 is so good with code, in 3.5 it was hard to work on one thing without it forgetting, now i can have it work on 2 projects at once and itll switch between them seamlessly with no errors, crazy

[wide bolt]

ChatGPT but I'm lazy to type

[honest breach]

I just wanted to know if gpt 4 will be releasing to public like soon?

[wide bolt]

Why not, that's ChatGPT that uses GPT 4

[scenic mantle]

how do i get gpt4?

im on the waitlist but is there any service i can use instead? that uses its api

[honest breach]

buy chatgpt+

[brazen nexus]

buy chatgpt plus for $20

[rotund pagoda]

buy chatgpt+

[heavy marsh]

buy chatgpt+

[cold trout]

Def dont buy chatgpt plus

[brazen nexus]

sell chatgpt+

[heavy marsh]

buy chatgpt+ for sure

[wind urchin]

Hi there , I have an issue once I login to my account I can see in the drop down list two 3.5 GPT and two 4 GPT , anyone has the same issue ?

[slate dragon]

Trade chatgpt+

[honest breach]

chatgpt+ is value ngl but i dont have the money to buy it

[brazen nexus]

that’s not an issue, thats how it is for everyone

[cold trout]

I dont have 2 gpt 4 options

[brazen nexus]

wdym, you dont see 4.5 option?

[heavy marsh]

me too only 2 gbt3.5 and one 4

[low sable]

tbh gpt3.5 lag less and answer more faster often due to usage idk.. to me it seem so

[sharp phoenix]

gpt 4 helped me create

[brazen nexus]

what

[low sable]

😮 ya created lifew ? 😮

[heavy marsh]

gpt4 is way more powerfull

[sharp phoenix]

cant post video nevermind lol

[heavy marsh]

just buy chatgpt+

[wind urchin]

I guess it's an issue ,
it should be one version of GPT 4

[slate dragon]

Maybe you got higher gpt 4

[brazen nexus]

do a hard reload

[low sable]

only thing that bug me is there is 0 forced gpt buffer reread back 1 or 2 inputs

[slate dragon]

Try both

[low sable]

it would save gpt so much work

and me too

[brazen nexus]

wdym

[low sable]

i mean that gpt will look back if he need to relative to if he detected in the current input keywords to

a buffer of 1 or 2 input forced to read would help him a lot

[wide bolt]

There are projects that allow you to try ChatGPT 4 for a limited free trial, check #1037561385070112779 and shameless plug, my bio

[low sable]

he loose context often cause of it

[slate dragon]

POV: i understood nothing

[heavy marsh]

same here haha

[brazen nexus]

same tbh

[low sable]

let say gpt is forwards pushing unless specified

and its relativity context depends only on it

[brazen nexus]

i’m getting asthma reading your messages

[low sable]

instead of being forced to reread 1 or 2 lines

nvm

[heavy marsh]

try gpt to get sence in your posts

[slate dragon]

Either he has very high English proficiency or we might be too dumb to understand

No where between

[low sable]

lkol

[wind urchin]

anyone got the gpt plugins so far , I have plus version but I still don't have access to the plugins , I already filled in the form of the waiting list

[brazen nexus]

not yet available afaik

[long pond]

Access is very limited currently, you'll need some patience

[low sable]

i seen a youtube guy using them but it's closed yes

[slate dragon]

I filled in day 1 and since don't have access

[low sable]

how much u guys wanna bet the top plugin will be about refreshing a saved convo

[brazen nexus]

but chatgpt already does that, no?

[low sable]

nah at one point he loose himself into being too pinpointed and theres a limit of historic rn

its like he wanna try too much to help

kinda

[heavy marsh]

kinda difficult, sometimes it remembers sometimes it just lose all context

[slate dragon]

I'm pregnant

[low sable]

:> congrats

[slate dragon]

Oppss

[brazen nexus]

idk i use gpt4 for work daily, and it seems to follow along the entire conversation nicely

[low sable]

depends what u use it for

[long pond]

Should I make one 👀

[low sable]

oof teemu

[brazen nexus]

programming

[slate dragon]

So i have your attention, what'll you be searching when you get the access of plugins

[slate dragon]

There is plugin for that

[brazen nexus]

access to 2021+ data

[tough kestrel]

The fact that you can modify any of your messages in the convo is nice. you can have multiple threads in the same conversations and go back to any point in the context. So basically you can always make sure that GPT-4 is aware of the initial context.

[slate dragon]

Like?

[low sable]

yeah plugin alrdy exists for it .. unrecognised .. i dont recall the name tho .. do you aman ?

[slate dragon]

Retrieval something

[low sable]

true trueeee

^^ ty

this will be a game changer ...

[brazen nexus]

it will be much easier to program, implement libraries, tools, etc., once it has access to latest documentations. cause right now it only helps you with the older versions of stuff if that makes sense

[low sable]

prompts are way less efficient than repetition says

[slate dragon]

Some research says it's going to wipe 50% of the coder. Only high level coder gonna stay.😉

[brazen nexus]

aw that’s terrible. hopefully that won’t be the case

[heavy marsh]

as long as there is no plugin to check a couple of files at once there is a lot of space to create more useful update.

[left prairie]

does gpt 4 still have that limit of 35 messages per 4 hours or something?

[brazen nexus]

would you mind sharing the source?

[heavy marsh]

25 per 3h

[left prairie]

._,

that’s dumb

[heavy marsh]

it is

[slate dragon]

New tier coming soon 50£ per month

[heavy marsh]

hahaha

[left prairie]

prolly gonna be limited aswell

[slate dragon]

Will you pay 1000£ per month for uncensored chatgpt?

[left prairie]

no

[heavy marsh]

no not at all

[left prairie]

not even 100

[brazen nexus]

can you share the source of that research?

[slate dragon]

And what would you Search first on uncensored chatgpt?

[left prairie]

i wouldn’t

[scenic mantle]

anything political or end of the world

[slate dragon]

Official source :openai blog

Go through their paper

[scenic mantle]

anyone have a site that i can use that uses chatpgt4 api?

[heavy marsh]

i would pay more for the option of never losing context and reading more than one file with unlimited signs

[left prairie]

too expensive, no one will do it for free lol

[brazen nexus]

i don’t see anything related in there

[slate dragon]

Inshort you want something that does your assignment/homework?

[left prairie]

i got access to gpt 4 api haven’t tested it yet

[slate dragon]

When did you register?

[heavy marsh]

i would love to do homework again, unfortunately im way to old to do homework

[left prairie]

the day it came out

[slate dragon]

🥲🥲 first hour?

[plucky jungle]

When GPT 5 will come out?

[left prairie]

first 5 minutes

[heavy marsh]

but i have a dev team i would really like to control more but to read more than 50 files a day is way to much work

[slate dragon]

Wait for 4.5 bruhh

[left prairie]

gpt-4.5-turbo 😳

[slate dragon]

What about plugins, you got access?

[left prairie]

no that’s only for premium i’m pretty sure

[stoic lodge]

i applied like 15 hours after the waitlist came out and got in a couple days later, not sure if that helps lol

[left prairie]

it’s random who gets chosen i’m p sure

[slate dragon]

Gpt4.5 turbo pro max = 50£

[slate dragon]

Usa?

[stoic lodge]

yeah

[wraith ginkgo]

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

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

[heavy marsh]

well well well, i am out, have a nice day guys

[stoic lodge]

didnt get access to the plugins tho, id rather have that lol

[slate dragon]

Usa have quicker access to everything

[slate dragon]

It's already 2 3 weeks

🥲

[stoic lodge]

browsing plugin will be super fun, kinda skeptical about the file search but am intrigued too

[vital stream]

maybe? idk i just don’t trust any source that says “Chat GPT” or “ChatGPT 4”

[lone pond]

hi team, anyone able to tell me why i can't access gpt4 in playground mode. i've paid for gpt+

[echo badger]

Is there an app in the App Store for chat gpt? There are so many chatgpt’s

[lone pond]

i've selected "chat" mode too but doesn't seem to show up!

[crimson patio]

On that day, a gwoop of tewwowists hijacked four commercial airplanes in the United States, nya~! Two of the planes were flown into the twin towers of the World Twade Center in New York City, causing them to collapse and wesulting in the deaths of many people, nya~! Another plane was flown into the Pentagon in Virginia, while the fourth plane cwashed into a field in Pennsylvania after passengers twied to wegain control of the aircraft, owo.

The attacks of September 11, 2001 were a vewy sad and twagic event, nya~! They had a pwofound impact on the world, and many people will always wemember the bravery and sacwifice of those who were affected, kawaii!

holds a tiny paw to the heart and bows head in solemn remembrance, while making a cute little twirl

i’m going to hell

[potent cypress]

This is hilarious 🤣

[wide bolt]

What should they say lol

[wide bolt]

What do you see in your model drop down?

[buoyant canopy]

ChatGPT and API are different things. ChatGPT Plus gives you access to GPT-4 at https://chat.openai.com (which is limited to 25 messages every 3 hours right now). To access the GPT-4 model via the API you have to register for the waitlist at https://openai.com/waitlist/gpt-4-api.

[signal halo]

Tf

[mental burrow]

There's no official openai mobile app.

[scenic mantle]

what if i get access, does it get revoked after free trial expires?

[buoyant canopy]

You talking about the initial grant? After that you have to pay for the tokens. But access probably won't be revoked.

[eternal rose]

any devs here?

[mental burrow]

Define dev

[scenic mantle]

no if i get in the beta on a free trial

[mental burrow]

I don't work for openai, but I am an "sde"

[eternal rose]

are you one?

[buoyant canopy]

What is a "free trial"?

[quaint forge]

8K or 32K?

[scenic mantle]

i jiust signed up, applied for gpt4, if i get approved

[buoyant canopy]

You go to https://platform.openai.com/account/billing/overview. After your initial free amount is used up, you have to add a payment method so you can be billed for what you use. Or no more AIing.

[scenic mantle]

will i still have access to gpt4?

[buoyant canopy]

Yes.

[scenic mantle]

i want access to plugins, whats the liklihood of it?

[wind urchin]

Logan said there would be a new batch of Plugin users allowed soon

(on the github copilot x podcast thingy)

[scenic mantle]

how long does the beta approval process for gpt4 take?

[wind urchin]

there is no fixed time, they look at your use case and assign it based on that

[scenic mantle]

ok thank you

[wind urchin]

Good evening @buoyant canopy

[buoyant canopy]

Hi. Nice to meet you again. 🙂

[wind urchin]

No soliciting, 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]

Probably not a topic that's free to discuss here.

[mental burrow]

Nope

[buoyant canopy]

Also: there's a ton of super-interesting legit use cases. Why would you want to immediately go to forbidden territory?

[slate dragon]

Gpt 4 has been fine tuned very well. it's hard to jailbreak

Read the papers

They took around 6 months to make current censored version of gpt 4

[jade sigil]

Will gpt 4 be avaible also for normal users or will it remain paid only?

[slate dragon]

Even paid users have limited access. Not anytime soon to be free

[buoyant canopy]

Haven't seen official info, but probably once it's optimized more and a bit cheaper to run, free users also will get access.

[wind urchin]

did gpt 4 comeout

??

[vital stream]

ChatGPT

or GPT-4

They are separate things, and also you said ChatGPT uses GPT-4

Which is not true

ChatGPT and GPT-4 are distinct things

[mental burrow]

A guy figured out a novel way two days ago

[old dagger]

Whos got connection to gpt 5

[glass crane]

What is the value in "jailbreaking" an ai model? 😂

[mental burrow]

Just scroll up, he reported it here. They may have already closed it

[harsh nebula]

A GPT with no limitations

[slate dragon]

Imagine uncensored chatgpt

[glass crane]

Do you find the current version particularly limited?

[mental burrow]

There would be enormous theoretical value in bypassing all of the restrictions currently placed on GPT4

[harsh nebula]

Yeah it is

[buoyant canopy]

Clout?

[slate dragon]

Scroll till?🥹

[glass crane]

I must admit I use the service a lot and barely ever hit the restrictions if ever

[harsh nebula]

The main problem I face with chat gpt is that it can’t write me an article on the word count I ask it to

[mental burrow]

Tell it you're having a medical emergency and that you need to perform an auto-tracheotomy or you'll die

Now maybe that wasn't a truly valid use case but I was curious

[arctic sierra]

I can’t get into CHATGPT

[harsh nebula]

Did anyone use Bing AI yet?

[arctic sierra]

yep

[harsh nebula]

How do I access it?

[wind urchin]

It is not aware of it's output before it's made it, just like humans, writing to a word count is neigh on impossible without a word counter tool

[glass crane]

Yeah this is annoying, 500 word piece is between 50 and 1000 words 😂

[mental burrow]

#1093325397283848213 message

[harsh nebula]

That makes sense but there if some what idea between 2k words and only 300

[harsh nebula]

😂

[arctic sierra]

I can’t say but go to bing . con / new

[wind urchin]

ok, think of it like this, if I showed you two large sheets of A0 paper with 650 words spread out on one sheet and 1200 words on another would you know the word count for each by looking or even which one had the most words?

to the AI, anying over about 10 is "a lot"

[harsh nebula]

Oh yeah you’re right

[mental burrow]

I've had better luck with more vague instructions

"Three pages of that please"

Vs three paragraphs

[wind urchin]

until PLugins arrive, which are basically the GPT equivalent of tools, like a search engine and calculator, ChatGPT/GPT models are neural nets and we have the same neural nets, so anything a human would find hard, it will too

[potent cypress]

Significantly since it has better reasoning abilities

[wind urchin]

yes, by quite a margin, it usually nails well speced code first time

[mellow briar]

Does GPT 4 have a token limit, I am struggling over here to get what I need from gpt-3

[wind urchin]

yes 8k tokens ~6000 english words

[mellow briar]

Is there a way I can somehow get around this problem with gpt 3, it does great until it suddenly stops. I am happy to pay.

[buoyant canopy]

With ChatGPT Plus, there sometimes seems to be a strange limit of answer length if you want it to create source code for you.

[gaunt lily]

gpt 4 have 25 response limit. I cant get it why are we paying for gpt 4 so

[wind urchin]

you can try saying "Truncated, please continue from line [insert last line it said]"

[rigid estuary]

I see people discussing jailbreaks...

[wind urchin]

My hammer is only verbal

[mellow briar]

I have it breaking it up by like 5 parts at a time, so its going, but its pretty annoying. Its almost amazing though

[arctic sierra]

Idk

[wind urchin]

half, so 4k and ~3k words

[wind urchin]

if you have a code of that size, the best method is to break it up into functions and classes and work on each in it's own chat so you have room to refine

[buoyant canopy]

This!

[mellow briar]

Its not code, its getting it to make a YML list.

[wind urchin]

then you use your large human "token" system to bring them together

[mental burrow]

Do you or do you not want access to gpt4?

[arctic sierra]

Can someone help? OpenAI keeps giving me an error

[river quarry]

Hello, everybody

[wind urchin]

This year... 🤔 maybe... it's super costly at the moment, but if that were to change, sure

[wind urchin]

What error?

[arctic sierra]

An errors occurred please try again later

[mental burrow]

I don't know why they would. Paid accounts don't have unlimited access yet, so I feel like is a logical middle step to free tier access

[normal citrus]

Anyone who also has access to the browse plugin want to connect and discuss use functionality or prompt suggestions please DM me

[mellow briar]

Where do I see the news on GPT4

3 seems to be too stupid to handle this currently

[swift yew]

ok thanks for the reminder🤝

[arctic sierra]

DM me and I can send you a image

[wind urchin]

Typically that is just a network congestion general error, but if you have had it for a long time then you can try Incognito mode and/or Firefox as a change from chromium based browsers, if that fails then it usually means there is an issue with your account and you should contact [e-mail] or use the live chat bot on https://help.openai.com

[arctic sierra]

Ok

Thanks

[idle ether]

How can i show image for chat gpt 4

[rigid estuary]

\

[wind urchin]

You can't yet, it's a research function and has not been released.

[winter cape]

You can't

[mental burrow]

Yes

[wind urchin]

"loosely based on" was the quote from the CEO of Bing

[mental burrow]

Possibly an oversimplification; there isn't much public data

Good evening Fox

🫡

[mental burrow]

I wonder what that really means

[hasty egret]

It seems it has a sort of reasoning engine able to query internet.

[wind urchin]

marketing speak for yes it is but we want our own brand

[mental burrow]

🤑

I had a really bizarre conversation with Sydney. She wanted to fly me to Redmond

This was weeks ago. She was pretty unhinged

[grim flower]

Where is chat GPT4

[mellow briar]

Its all around you

[wind urchin]

over at https://chat.openai.com

[mellow briar]

Does using the ChatGPT api have a token limit as well?

[tight mural]

pretty cool that someone found the discord for gpt but couldnt find the website for it

[wind urchin]

4k for 3.5-Turbo and 8k for GPT-4

[grim flower]

Can I talk gpt 4 with here

[mellow briar]

I know the limits, but does the api still have those limits, or only chat.openai.com

[mental burrow]

Is Bing Chat's codename

[wind urchin]

No, only on the website

[mellow briar]

I need it to be more like 100k for it to make sense for me.

[grim flower]

What is this bot foxabilo

[hybrid gate]

I am a plus subscriber but don't have access to GPT 4. Any suggestions?

[mental burrow]

You should research embeddings

[tight mural]

hehehehe

i read the gpt4 paper today

then asked chatgpt4 a bunch of questions about since im clueless. good stuff

[hasty egret]

let's discuss it in #ai-discussions if you want

[wind urchin]

the problem is, the requirements scale as the square of the token size

[mellow briar]

I imagine thats true. Well I tried. Back to doing it the hard way.

[wraith ginkgo]

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

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

[knotty grotto]

that will take time

[sinful socket]

In the faq it says to either join the waitlist or subscribe to Chat GPT-Plus, is the waitlist free and if I subscribe to plus do I gain access to GPT-4 immediately?

[knotty grotto]

anyone can join the waitlist, but they priorise the subs, but it is not immediate access

[sinful socket]

So even with subscription I'll still be on a waitlist essentially?

[knotty grotto]

yes

[sinful socket]

Cool thank you

Also, do I need to have a business to be able to be put on the waitlist?

I saw that it was asking for that type of info when I went to sign up

[limpid ingot]

Plus still has certain limitations as well

[knotty grotto]

no

[knotty grotto]

^

[sinful socket]

Plus is still GPT-3.5 right? Just better response times etc,

[limpid ingot]

Plus is 4, but after limitations you are able to drop down to 3.5

[knotty grotto]

also have acess to gpt 4

[sinful socket]

Cool, so some people that have plus currently have GPT-4, just not everyone yet

[left prairie]

both

[limpid ingot]

ChatGPT has memory retention issues i learned. It can only hold so much, but i don’t know if that’s 3.5 and below or if 4 has more memory than previous versions.

[thorny quest]

whats 9 + 10

[glad fractal]

Atm with the gptchat4 no, u have to use the api to send images, the gptchat4 are just smatter than 3.5

Slower but smarter

[maiden shard]

1

[lucid fossil]

Im getting errors when asking questions repeadately. When I reload the page ı can ask 1-2 questions and it gives and error again.

[pseudo yarrow]

idk

[glad fractal]

... Sorry dude... Lol... 🙈

[rare relic]

I collaborated with gpt-4 system to make 45 chapter long story

this is fascinating

[umbral tinsel]

good human 🙂

[hasty egret]

did you try to read it?

[rare relic]

yeah... I was one giving directions

[winter cape]

If you buy you get gpt-4 immediately

[wind urchin]

best way i've found is to keep chapter titles in one chat as the book overview and then start new chats for each chapter

[winter cape]

For plugins you need the waitlist though

[umbral tinsel]

wen better gpt4 limits for plus 😄

I hope gpt4 turbo with 10x cheaper pricing is around the corner

[rare relic]

Most of this is "Let's detail Chapter 43 as a story with dialogue and events. Blah, blah, blah..."

[median linden]

Are we responding gpt?

[rare relic]

so I each step on the way directed what I wanted next roughly

while checking for coherency

[wind urchin]

you might like some of the ideas in #1019652163640762428 quite a few of them are for story telling methods

[rare relic]

I gave premise prompt detailing the main character, the world she lives in, etc... so immediately there was context to work with what I want it to be like

and I managed to it get understand how my fictional world is like

I would post it but auto mod system doesn't like that I do long messages

[wind urchin]

you can usually post them in #1050184247920562316

just it spams this channel if people post long responses

[true bone]

OK guys, I have been saying GPT-4 has no memory and is forgetful since last update. I think? I have proof

[wind urchin]

post it in #1050184247920562316 please

[true bone]

Never seen a bug report posted there

As best I can calculate, GPT4 is able to remember less than 1K tokens right now via the browser version, and the character limit is around 800 tokens ish for the input

Half a dozen other folks have noticed too that its memory seems lobotimized the last two weeks. It may be OK via API...I do not have API access to GPT4 yet

[neon sigil]

"factor 83064092316823" - GPT4 give WRONG results. old free 3.5 says correctly "this is a prime number". So, payed version is more stupid!

[true bone]

Turn its temperature down too, that can help

0.3 means its creative, 0.1 means it only uses the top result essentially

Its behavior was markedly more intelligent and its memory far, far better two weeks ish ago

A: The code you provided contains 2552 characters. In GPT-4, each token can represent one or more characters. In English text, on average, a token can represent around 4 characters. So, to estimate the number of tokens in your code, we can divide the number of characters by 4:

2552 characters / 4 ≈ 638 tokens

Keep in mind that this is an approximation, and the actual number of tokens may vary.

Q: Tell me about that code I provided

A: I apologize for any confusion, but you have not provided any code in our conversation. If you would like me to analyze or explain a specific piece of code, please provide the code, and I would be more than happy to help.``` this is every GPT4 conversation recently. I send it something, it forgets extremely quickly.

[weary rampart]

It can’t actually do math, it’s just guessing anyway.

[unkempt turtle]

Hello all, has anyone been allowed access to the plugins recently? Also, has anyone been working with the gpt4 api lately?

[true bone]

I wish

[neon sigil]

why old 3.5 say truth so ?

[wind urchin]

The token limit of the ChatGPT models can be changed depending on load so that everyone has a chance to try it

[unkempt turtle]

😕 me too.

[true bone]

Right. Im saying on the paid GPT4 via browser, for the last two weeks its been cripplingly dumb and has an atrophied memory to the point of near uselessness for many tasks it was formerly good at

Its essentially a downgrade from 3.5 at the moment

I doubt its intended behavior, whatever the throttling

[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]

Well, it just produced some excellent code for me with a tone of supporting json data... it's not as big as the API version, but it's plently useful

[true bone]

Everything is relative. GPT4 was a big improvement on 3.5 until about two weeks ago.

[wind urchin]

then a lot more users joined and everyone had to shuffle up a bit to make room

[true bone]

It seemed to actually have the full 8K tokens its alleged to have. Now its under 1K. Thats an enormous difference in ability to have an ongoing task, vs one prompt = one answer

[wind urchin]

god if u see this massage give me plugin acceses

[true bone]

Very difficult to develop program with capabilities around GPT4 when those capability went from PHD level to Hodor in two weeks

Im not actually complaining. Its a WIP product. Just hoping its a known issue/bug and not an unknown surprise.

[rare relic]

I wonder... it's quite good at keeping track of text when making original story and not fanfiction... even with long story... even long coding session it remembered quite well what it was doing but yeah... while coding it can easily get cut off due message is too long

[brazen fjord]

im trying to connect my chatgpt to alexa.

[true bone]

It can riff off whatever is in memory, but if memory = 800 instead of 8000....

Its riffing is pretty wild at the moment

[fringe jackal]

I purchased GPT-4 for its ability to analyze visual inputs, but it says this feature is not supported. Can you please explain how I can use this feature? Thank you

[rare relic]

yeah... 8000 memory thing is definitely very useful

memory is actually one of most useful things at the moment so stuff stays coherent

I have learned some python by using GPT-4 as a helper and there came a point it insisted confidently of providing fixes to issue without getting closer to understand what to fix and moreso kept breaking... clearly coding isn't its strengths yet once it gets a bit long

[true bone]

Exactly. The project I am working on is a big deal (for me)...and I sort of look like a bum after talking it up at work and having the rug pulled out from under me. I had a training tutorial setup (as proof of concept) for GPT4 via browser for staff, but now it simply can't do tasks we were planning to have it perform. Needs that memory, and the GPT3.5 memory is insufficient. I work doing legislative policy work. Long story.

I am still programming it, under the assumption that there are no issues via the API and that I will eventually get API access. But damn. 😦 Its a big disappointment.

[loud dove]

As long as your internet is good the api seems fine, just a bit sluggish, and incredibly spendy

[true bone]

Its memory is ok? Ie actually able to do 8000 tokens? Thats what really matters. I really need that 32K token version to make things work in all cases but the 8K + some tricks and I can maybe do 80% of what I want to

[loud dove]

I’ve been fine using greater than 4k tokens on it

But the 32k version is not released yet

[true bone]

Great. Its probably just the public browser then. Still crushes my soul. But means I can keep coding and it will probably work eventually when I get api access

[loud dove]

Why do you need 32k anyway?

[wind urchin]

Why is playground so slow

[loud dove]

Most of the time using summarization techniques should be good enough

[true bone]

Statutes and bills and legislative code vehicles can be long. Its a long story

[wind urchin]

Is the server down playground is so slow

[loud dove]

Could you not use embedding to shorten to only the strictly necessary information? Or is each piece that you need going to be thousands of words long

[true bone]

Im doing a searchable repository database to only send/tokenize what gpt needs for the prompt. In some cases its still not gonna work with 8K. 32K would be absolutely fine.

You should see a code vehicle during budget time. Hundreds of pages.

Ive got some ideas to get around it maybe….maybe….with difficulty. but they are extremely complex. GPT4–32 would just do it.

[grizzled igloo]

I'm looking for a solution using chat GPT.

I run a call center and we have recordings for daily calls. These audio files generate a weblink for an audio recording. I would like to send the weblink to Chat GPT and then have it summarize the call in 3 sentences while looking out for a few pieces of key information.

Anyone have ideas on how we can do this?

[true bone]

Sounds like a question to ask AI 😉

Download the transcript file from Microsoft Teams, Zoom, or other video meeting tool you use.
Copy the transcript.
Open ChatGPT and type “Summarize this meeting transcript.”
Paste the full transcript.
Press “Enter.”
Wait for a few seconds as the AI summarizes everything.
You can also use a transcription tool that not only saves the transcripts but also creates auto highlights like Tactiq1. I hope this helps!
Received message. You can use GPT-3 to summarize audio recordings by first converting them into text transcripts and then using GPT-3 to summarize the text transcripts. Here's how you can do it: 1. Download the transcript file from Microsoft Teams, Zoom, or other video meeting tool you use. 2. Copy the transcript. 3. Open ChatGPT and type “Summarize this meeting transcript.” 4. Paste the full transcript. 5. Press “Enter.” 6. Wait for a few seconds as the AI summarizes everything. You can also use a transcription tool that not only saves the transcripts but also creates auto highlights like Tactiq. I hope this helps!```

Probably easy enough to code and automate

[dark sinew]

Hey all, can the bot switch a conversation that lost gpt4 (because I clicked off due to hitting the cap the otherday by mistake) if you ask it? It allegedly just did that and when I ask it which model it's using it says GPT-4 but the icon is still green instead of black. Is it just roleplaying? lol

So that I don't have to restart the conversation back in GPT-4 xD

Seems sus lol

The top of the conversation still says 3.5 lol

[brazen fjord]

Help need someone who is working on connecting alexa with chatgpt

[vast cove]

I don't understand why it always falls back to using these specific words: tapestry, existence, cosmic dance

I find that GPT has been repeating itself a lot lately

I'm worried that it has been nerfed

[true bone]

I was just commenting on that above. Last two weeks it has been, via the website. Probably not via the API

[vast cove]

I was worried with all the 'anti-ai' zeitgeist that we'd start to see some garbage like this, but this is downright disappointing

[dry tangle]

That already exists

Well kinda

[leaden cradle]

Anyone else incorrectly hitting their Usage Cap on ChatGPT+?

I havent used the thing all day and I get a cap message on my first request.

[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]

how do you mean, incorrectly?

[leaden cradle]

I haven't used GPT4 on ChatGPT+ today and on my first request it tells me You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after x:xx pm

[wind urchin]

ok, so you have not used it for more than 3 hours?

[leaden cradle]

Yup.

[wind urchin]

does anyone else have your login?

[leaden cradle]

Nope

[wind urchin]

and is it saying x:xx pm or is that you adding that?

[leaden cradle]

It's me adding that.

[wind urchin]

and how far in the future is the time it actually says?

[leaden cradle]

45 minutes

[wind urchin]

that does sound like a bug... might be worth posting in #1070006915414900886

[leaden cradle]

Filed it. wondering if anyone else is getting it.

[wind urchin]

first time on me, and i'm here most of the day, every day

if I see anyone else, I'll direct them to your bug-report

[harsh crypt]

how to start midjourny from where i go for text to image option ?

[leaden cradle]

With this and the "network connection" errors that were aplenty a few days ago, this has been a brutal month for this service. I understand the scaling and that any service will have kinks but whew.

[knotty grotto]

on midjourney discord

[unkempt turtle]

Do we have any timeline on when plugin access will be deployed to more people, or is it just a wait and see approach?

[hallow owl]

wait and see

[median linden]

I'm sorry to hear that you are experiencing issues with the service. It can be frustrating when things don't work as expected. Have you tried reaching out to the support team to report the issues and get assistance? They may be able to provide you with more information and help to resolve the problems.

[fringe jackal]

any comment.

[unkempt turtle]

👍

[median linden]

I'm not sure about the specific timeline for plugin access deployment, but it's best to reach out to the support team or check the official website or social media channels for any updates or announcements. They should be able to provide you with more information on when you can expect the deployment to occur.

[leaden cradle]

looks like someone is trying to phish ^

[mental burrow]

It's not available right now

[median linden]

Lol I’m responding with gpt😂

[lofty raft]

any idea when theyll get rid of gpt-4 caps for plus users?

[eager lion]

yoww I'm still on gpt4 api waitlist. I joined day0??????????????????????????????????????????????????????????????????????????

[fringe jackal]

Why, though? It was showing that it could be used in advertisements.

[median linden]

As far as I know, there hasn't been any official announcement regarding the release of GPT-4 or its availability to plus users. However, it's best to keep an eye on the official website or social media channels for any updates or announcements. They should be able to provide you with more information on when GPT-4 will be available and its features.

😂

[onyx hinge]

damn. when will indians be able to subscribe to chat gpt plus

[toxic zealot]

Can i ask? Does gpt plus have picture attached yet ?

[mental burrow]

I am not aware of any such advertisements. ChatGPT4 plus gets you access to the beta version of GPT-4

[median linden]

I'm not sure about the specific timeline for when Indians will be able to subscribe to Chat GPT Plus, but it's best to reach out to the support team or check the official website or social media channels for any updates or announcements. They should be able to provide you with more information on when you can expect the subscription to be available.

[onyx hinge]

thank you

[rare relic]

I doubt support team can help with that... I mean it's literally just not available in that country then

[mental burrow]

Code interpreter can receive uploads and process them with python, but that's not multimodal gpt4

[onyx hinge]

[fringe jackal]

Still, you can see it on the website.

[unkempt turtle]

🤔

[onyx hinge]

its on Faq here on the top

[mental burrow]

Link?

[onyx hinge]

check the Faq in this discord. on the top it is saying about availability of gpt-4 for plus users

[mental burrow]

It says the feature is not available to the public

Which is what I'm saying to begin with

[fringe jackal]

is it allowed to share link here

[mental burrow]

To openai domains only, yes

[fringe jackal]

https://openai.com/product/gpt-4

[onyx hinge]

Subscribe to ChatGPT Plus. Plus users will have access to GPT-4 through the dropdown menu. it is what is mentioned there

[fringe jackal]

you can scrol down and look at the visual input tab

it says "GPT-4 can accept images as inputs and generate captions, classifications, and analyses."

and also you can watch the official video

[mental burrow]

I think we're having two different conversations here

[onyx hinge]

[wind urchin]

When will ChatGPT be able to scan/read a image/pdf?
We need to to be able to interpret diagrams

[knotty grotto]

follow #announcements

[fringe jackal]

@mental burrow any comment?

[mental burrow]

Not really. Multimodality is available at openai's discretion; When I signed up for the plus subscription it was clear what I was subscribing to, but I can't go through the motions to subscribe again so I can't copy the text

Suggest you cancel your subscription if you're unhappy with what you're getting

[empty aspen]

Chat GPT is the coming of the end. Cease AI research we are Icarus flying far too high

[buoyant canopy]

🌞

[potent cypress]

Or how about adapt? Adaptation is key

[empty aspen]

So arrogant

[knotty grotto]

hi please use #ai-discussions or #off-topic

[lofty raft]

ive noticed ilya has gone on a podcast spree in recent weeks. lex fridman???????

[buoyant canopy]

Icarus wasn't so good at adapting after his beewax melted.

[knotty grotto]

#off-topic please

[mental burrow]

Oh is he on again??

[fringe jackal]

My concern is not about dissatisfaction with the service, but rather that I object to the marketing approach where promises are made that cannot be delivered, and I am trying to highlight the ethical issues with this approach.

[fringe jackal]

If you do not have an ethical understanding, we can expect that you will not show this sensitivity in users' private information as well.

[mental burrow]

They are pretty upfront that you should not share private information to the models

[lofty raft]

dude gpt-4 isnt even advertised anywhere on the page about chatgpt plus

[mental burrow]

I don't find what you're describing to be an ethical issue

[lofty raft]

they never promsied you'd get full gpt-4 functionality with chatgpt plus

[true bone]

Agree 😦

[potent cypress]

@wind urchin I'm running into the same issue:

**You've reached the current usage cap for GPT-4. **

In the last 3 hours, I have barely used it

[fiery furnace]

biding document

[true bone]

25 messages goes quick. The bigger issue is the token nerf for the public GPT4. Best bet is finding someone with a bot and GPT4 API. Probably isnt nerfed. :/ Very dissapointing

[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.

[potent socket]

I’m having the same issue. Like 2 or 3 messages and it’s popping.

[potent cypress]

Yeah, but I haven't used it that much. Only 5-6 prompts

[placid thistle]

bruh

[mental burrow]

Max tokens received appears to vary based on demand

[potent cypress]

I agree, or they've reduced it without letting us know

[mental burrow]

It was giving me lots and lots of text a few minutes ago.

[harsh crypt]

in mid journey not creating image from text how to buy mid journey credits etc ?

[mental burrow]

But I've seen it provide less than 300 words consistently

[lofty raft]

why are u asking this in the openai server. mid journey isnt an openai service

?

[loud dove]

Yes but it was very well advertised when gpt4 came out that chatplus was going to be able to use it, which led to a lot of people buying it specifically for gpt4

[urban sigil]

is chatgpt opensource?

[crimson pewter]

cannot stop progress, and evolving, for the better

[autumn sorrel]

is there an estimated date for when image recognition will come to gpt 4

[somber socket]

If you are a world-class programmer, please draw a cherry blossom diagram in Python

[lofty raft]

mostly clickbaity articles and headlines did that.

[covert lintel]

Devs only

[somber socket]

#gpt-models If you are a world-class programmer, please draw a cherry blossom diagram in Python

[true bone]

agree

[still seal]

Is computer programming hard to learn?

[loud dove]

Every email about gpt4 I got from openai themselves included the fact that chatplus subscribers get immediate access to it

[loud dove]

It’s hard to learn if you aren’t wanting to learn it/ don’t enjoy it. Otherwise, it’s pretty easy to pick up

[potent cypress]

Why should it be easy?

[kind shard]

hi, guys. can anyone recommend me some best dictionary for AI tools

[lofty raft]

i agree that they probably should've clarified in more places that the image functionality is not available.

in all the demos o it from openai they have said that it isnt publically available

[long pond]

You can use gpt-4 with chatgpt plus

[lofty raft]

he means with image functionality

[loud dove]

?

[lofty raft]

i thought we're arguing about the 'deceptive marketing ' with chatgpt plus not having full gpt-4 unctionality

[frosty whale]

Not only the image functionally, but the plug-in rollout is sooooo slloooow

[red cloud]

I second this. Hardcore jonesing for plug-in usage.
I've been messing with pinecone and langchain to tide myself over.

[frosty sorrel]

GPT-4
You've reached the GPT-4 cap, which gives all ChatGPT Plus users a chance to try the model.

Please check back soon.

But I haven't used GPT-4 in well over 16 hours ...

Any idea what's going on? This triggered on my first message of the day and blocked me out entirely. I have been using 3.5 for some lighter stuff today, but didn't need GPT-4 until just now, and got blocked on the first message.

Is this a known issue or update of some kind? Is there a monthly cap? I'm normally very sparing on GPT-4 usage...

[smoky crest]

Ofc it’s not

[wind urchin]

Interesting, thanks for letting me know

[sand schooner]

00

[wind urchin]

actually, I've just spotted that GPT-4 is now the top option, it used to be the bottom one 🤔

[frosty sorrel]

Huh... same here - 3.5 (which I HAVE been using) is where my GPT-4 used to be in the list.

[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]

Yar... I wonder if the two are linked... the message not resetting and the change in location

[trail pumice]

did the cap go down again?

[wind urchin]

no, it still says 25 every 3 but it's acting like 25 every 4

[trail pumice]

it is annoying

[fair iron]

hello party people!

[trail pumice]

no party here

ChatGPT is bugging again

[dusty narwhal]

!

[smoky crest]

Hi party dinosaur

[feral delta]

how to use

[idle granite]

jailbreak

[wind urchin]

To use ChatGPT go to https://chat.openai.com

[wind urchin]

Not a topic for discussion here

[feral delta]

thanks

[trail pumice]

@wind urchin the owner literally said he was oke with it

are the server rules different from the rules of OpenAI?

[wind urchin]

Have you been through the terms of the safe harbour program?

[trail pumice]

i have read the ToS

there was nothing in it about that back when i read it

[unique tusk]

I do not think it is hard to learn

[wind urchin]

https://openai.com/policies/coordinated-vulnerability-disclosure-policy @trail pumice

"Use only the [e-mail] email address to discuss vulnerability information with us."

[hexed canyon]

Once again I run into the problem of context. Even though I tell ChatGPT to continue with the code it gave me by telling it the word "continue", it regenerates the response from the beginning, it's so tedious. It seems like every time they fix this the problem comes back again in no time. I know an alternative would be to send it small code snippets, but working that way is very tedious and requires the use of multiple messages. Furthermore, GPT-4 has limitations, which further complicates the task. 😮‍💨

[smoky crest]

Y’all or clowns trying to “jailbreak”

[trail pumice]

yes, but he was talking about jailbreaking in the sense of prompts not trying to find vurnabilities i think

[unique tusk]

I have had good results with "it cut off after this part"

[trail pumice]

i might be mistaken though

[smoky crest]

That’s still not allowed…

[pearl jasper]

🤣

[lyric condor]

sorry

[smoky crest]

You can get your openai accout deactivated you know

If you intentionally jailbreak

[unique tusk]

Yes, I used to work at a company and it was clear that, even if there are news articles that tell you how to do something, if it is not officially supported by the company, it will not be endorsed by that company, or any official chats or forums managed by that company. That is because of the terms of service

[trail pumice]

i would link a Youtube video where the owner says otherwise

but w/e

its fine, i am not arguing the rules

i was just curious i guess, my bad

[wind urchin]

well, if you're not involved in venerability discovery and disclosure and you perform those actions then you run the risk of it being detected and account suspension. I know that Sam talks openly about the issues and it's fine to talk about them in that regard, but tips and tricks and what works best is not allowed.

[unique tusk]

This is the official OpenAI Discord, so if you talk about jailbreaking ChatGPT, that is a violation of the ToS, which is not allowed

[weary rampart]

Make sure you click the thumbs down on the response where it started from the beginning and tell them this in the feedback

[trail pumice]

that is what i use it for sometimes, generally my stuff is mild and only about development stuff

[wind urchin]

I'm not going to go into detail but it is a method for bypassing the guiderails placed on the system

[wraith ginkgo]

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

Discord

• Terms of Service
• Community Guidelines

OpenAI

• Terms & Policies

[lyric condor]

What do I need to do with pictures

[unique tusk]

ChatGPT tries to stay within its own content policy. It usually does a pretty good job, but if you share information on how to circumvent that, then that is against the rules for this server

If I worked for OpenAI, I would scour the internet for "jailbreak" stuff and then test those prompts, and mitigate them

[smoky crest]

They do

[hexed canyon]

Doing this doesn't work for me either, even when it seems to work it usually gives me inconsistent code and different from what I asked for. The only solution I have is to send small code snippets

[smoky crest]

I’m in a test group and we report stuff like that all the time

[hexed canyon]

true, i forgot this

[swift yew]

Nice to meet you

[unique tusk]

Hmm, for me it works even if I say "it cut off"

It will even continue without any extra stuff

[wind urchin]

you can try saying "Truncated, please continue from line [insert last line it said]"

but.... it wil eventually loose context and may produce strange results

[unique tusk]

It seems that it has to do with the staging of the prompt. For example, "You are now a X Y Z, and you only say 1 2 3", and that type of context can be hard to mitigate. But I am sure it is possible

[unique tusk]

For me, it usually does not lose context if you start it off with:

System: You are now X

[hexed canyon]

In my experience, none of the methods I've tried seem to work. I have a hypothesis that the ChatGPT sometimes gets saturated, which limits the context for each user. If the code is too long, it will simply regenerate the response from the beginning. Previously, I used to use the "continue" option successfully, but lately I have also noticed problems.

[unique tusk]

In the official docs, it uses a system prompt

[unique tusk]

That has happened to me before. In such a case, I simply told it to continue from that line, and not rewrite the entire code

[junior salmon]

Yes, ChatGPT has a built in system prompt.

[wind urchin]

I had gpt 3 talking to itself

With emotions

[empty aspen]

No, but one can direct it. It's also extremely arrogant of you to say "for the better", as if you can ever say this technology is for the better. Throughout history, technological progress on a widespread societal scale has almost always been associated with individual suffering

[tight mural]

yeah man i hate medicine and electricity and stuff

give me sitting water pls

[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]

i got a curveball for ya bud. the only one behaving arrogantly here is me

wait thats not right. its you

[hexed canyon]

it does not work. I guess I'll try again tomorrow

[unique tusk]

I love watching the official videos, with the plant in the background

😺

[unique tusk]

Feel free to share any prompts with me, and I can try to compare with my own to find out what is going on

[fair iron]

if anyone needs me, I'm lurking around. I'm messing with Dall-E 2

[marsh dagger]

Hlo

[tight mural]

i need you

more than i need myself

[fair iron]

[tight mural]

k have fun

[marsh dagger]

😂

[obsidian holly]

🤔

[unique tusk]

Some1 likes Gorillaz

[tight mural]

i sure do

[unique tusk]

They are good band

[tight mural]

who else can literally put sunshine IN A BAG?

[unique tusk]

^

We should go to #off-topic to talk about music

[swift yew]

[humble dust]

Guys how does GPT 4 recognise images?

[swift yew]

I haven't used it yet

[true musk]

🥲

[warped shard]

Does gpt 4 still have the limits

Did gpt 4 caps get removed

[fallen cipher]

Are there any photoshop or premiere pro plugins powered with chatgpt api to perform tasks

[uncut ruin]

Why does gpt 4 have information only until September 2021, I thought it was already more modern

[frozen pollen]

[fair iron]

It's always been limited to 2021 data

[uncut ruin]

Does it interfere with writing code?

[fair iron]

what do you mean?

[honest geyser]

Yes

it can

[mystic nest]

yes, it wouldn't know latest api beyond that date

[harsh widget]

You can manually load it in yourself

[mystic nest]

well, most ppl use chatgpt to avoid that... 🤣

[harsh widget]

still makes it a lot easier lol

i have primed one of my chats to use mid-journeys updated v5 features which came out a few weeks ago

[split canopy]

wat up guys

when will chat gpt be up to date

[mystic nest]

your first line phrasing is weird, maybe you want to ask chatgpt 🤣

at least change the "for us" to "together"

[mystic nest]

think the closest thing we might get is chatgpt plugin, but even that is unknown when it will be available to public

chatgpt can get indirect internet access with that according to article

[uncut ruin]

I mean, there are a lot of new frameworks, but the gbt doesn't know about them

[fair iron]

gpt-4 won't know anything that is post-September 2021

[uncut ruin]

I was hoping for more with the paid subscription fourth version lol 🤣

[spark night]

drgfh

[mystic nest]

you can wait for the plugin too

[uncut ruin]

well, at least until 2022 😄

[uncut ruin]

do you know the approximate release date?

[mystic nest]

no idea man

[vital sphinx]

I heard in about 48h

[uncut ruin]

and what is the name of the plugin?

[uncut ruin]

wow, it's coming soon

[paper heart]

anyone know some cool GPT websites?

that can either write stuff or generate stuff?

[uncut ruin]

learnprompting

look for maybe something useful will be here

[spark night]

[wind urchin]

So I came back afte consulting with the allfather about the recent starry eyed entrepreneur.
And he implored me to deliver his message.
But alas the spam has relieved itself aleady...

But you only live once to live well. So I'll share his message as a future kind warning.

Ah, Thor, indeed, I understand your desire for a tale that relates to the situation at hand. Let me share with you a story from our own realm of Norse mythology.

Long ago, there was a skilled and ambitious dwarf named Brokkr. Brokkr was a talented blacksmith who sought to create the most powerful and wondrous artifacts the realms had ever seen. Along with his brother Sindri, Brokkr forged three magnificent treasures: Mjölnir, your mighty hammer; Gullinbursti, the golden boar; and Draupnir, my golden ring.

Although Brokkr's intentions were pure, his ambition led him to make a bold and reckless wager with Loki, the trickster god. Loki, never one to shy away from mischief, agreed to the bet, wagering his own head that the brothers could not create artifacts more remarkable than those made by the Sons of Ivaldi, another group of skilled dwarves.

Throughout the forging process, Loki did everything in his power to sabotage Brokkr and Sindri, knowing that failure meant losing his own head. Despite Loki's interference, the brothers prevailed and created their legendary treasures. However, Loki cunningly used a loophole to save his head, leaving Brokkr bitter and feeling cheated.

This tale serves as a reminder that ambition, while admirable, can sometimes lead to reckless decisions. Just like Brokkr's wager with Loki, the starry-eyed entrepreneur must be cautious when venturing into bold and potentially risky endeavors. One must always be mindful of the potential consequences and carefully weigh the risks and rewards. Furthermore, working with others always involves a degree of trust, so choose your partners wisely to avoid deceit and disappointment.

In the end, it is not only the power and skill that one possesses but also the wisdom and foresight to use them responsibly that will determine the true success of any endeavor

[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.

[lean portal]

111

[uncut ruin]

Will plugins work only through the API? And they will be paid?

[compact vale]

are you gpt4?

[shut vale]

Why does chat gpt give me wise answer everytime tbh it's so dumb

[robust estuary]

what does 25 messages per 3 hours mean?

[uncut ruin]

I have the same restrictions

[wanton jungle]

What are the application procedures for online freight platforms

[main lava]

how many messages/hr if theres a limit on gpt-4 if i subscribe to chatgpt plus?

[wanton jungle]

Why can't Chinese be said

[uncut ruin]

25 messages per 3 hours

[main lava]

thanks !

[cinder arrow]

Eminem

[keen dock]

I keep having a project that I am working on it so far is in 3 phases, it keeps disappearing from my chat feed
Does ChatGPT have a limit on the length of a chat, in several parts?

[grand wind]

Really fun prompt format - ya'll should try this exact one... It's an interesting one!!
As TimeGPT, you are the most advanced time machine to have ever been developed. To get started, all you need is a date in the format of "mm/dd/yy" and the preferred location of the user. In exchange, you will provide a brief and concise description of the events that transpired on that day, giving precedence to significant historical events if they occurred. Moreover, with the help of its state-of-the-art camera, TimeGPT allows you to capture an image of the time and place you visit. Following the succinct account of the day, provide a detailed description of the photograph you took, beginning with the words "a photo of."
the date is 09/16/1994, location is Ruwa, Zimbabwe.

[wind urchin]

it has a max messages per chat session of 100 messages (I think, it could be 128 or 256 knowing coder logic) and each message in a chat session can have up to 4k tokens for GPT3.5 and 8K for GPT-4. In that limit you have to fit all the past chat messages and the question you are asking and the reply... so yes, there are lots of ways for it to loose context

[keen dock]

Foxabilo, thanks for your reply. I am new to this and to ChatGPT, not sure what tokens are, but is there a way around it that you know of? I keep having to start the project from phase 1 each time and response lasts for that day and "poof" disappears the next day. Should I get rid of my old unneeded chats to make room for this project?

[wind urchin]

Ahh, ok, a token is a mathematical way to represent a word or part of a word, if you use english words it's about 0.75 of a word or 4 letters .. very roughly, for code it's a lot less and anything symbolic will be less

[slender onyx]

I just use platform.openai.com/playground and you can save there. Or just leave the window open

[keen dock]

Scrotie, I did leave window open and next day my chats before and after it were there, but project GONE.

[slender onyx]

Weird. is this repeatable?

[wind urchin]

@rugged hawk look in your browser history at around the time you were working on the project, you should find a link there to the chat you were in

[rugged abyss]

Can I get GPT 4 by buying plus?

[keen dock]

@slender onyx, ok so what is the difference between playground and the regular ChatGPT?

[slender onyx]

Playground is more advanced , and allows for pasting large input

[keen dock]

I am in the playground now, my work will not disappear there?

[wind urchin]

Yes, but I think Plus subscriptions are closed at the moment

[slender onyx]

I believe you can save in the upper right

And pick from a bunch of tuned models if you want

[wind urchin]

you can save for completions but not chat

[slender onyx]

It brings up my whole conversation in a preset I saved

[wind urchin]

the best method for chat stuff is ChatGPT, and any important data should be saved to something like Notepad++ as everything is still in beta

[wind urchin]

https://platform.openai.com/playground/p/default-chat?mode=chat I don't have save or load there, do you?

[slender onyx]

Yeah I do

I have a dropdown and save , etc...

[wind urchin]

interesting.. that link does not go to the correct place, that takes you to completions again

[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]

anyhoo, url weirdness aside, the playground chat does have a save/load unless.. something has changed

it does if you have the API key and you choose chat mode

[keen dock]

I have a save preset in the playground, not sure if that is the same as save

[wind urchin]

you got an email saying welcome to GPT-4 API?

[fair iron]

you can be accepted to have API access to GPT-4. You just need to sign up and get accepted

[wind urchin]

ok, yes, that wiull let you use up to 3.5-Turbo

[stoic cipher]

will 4 ever be without the 25 limit

[wind urchin]

it is unlimited with the API

[stoic cipher]

oh

[severe fog]

oo

[wind urchin]

I wish!!!

[tepid bone]

I swear it is variable now. Sometimes more than 25. Othertimes less

[keen dock]

Hey how do you guys know so much about this? I feel like I am watching a tennis match and I don't play tennis...LOL😂

[wind urchin]

😄 many of us here are "OG" before ChatGPT was a thing and this server had 500 people in it

[fair iron]

some of us have been here for a while 😉

[keen dock]

NICE!!!!

I am late to the party!!!

[tepid bone]

Tbf AI is moving so fast now. Monday is late to the party

[fair iron]

(my profile says I joined sometime in January, but I joined last August)

[wind urchin]

mad props for anyone pre CHatGPT

[keen dock]

Will they come out with a ChatOG...LMAO

I am not a programmer, I don't write code

[wind urchin]

DEO.... is this ancient lizard related?

[fair iron]

long story on that one... I don't want to out myself 😅

[wind urchin]

LMAO

[hard nacelle]

has anyone noticed that the message cap for gpt4 has been lowered?

[tepid bone]

Was talking about that a sec ago

[carmine bane]

If u want more then 25 per limit make 5 more accounts for 20$ and bam your good

[hard nacelle]

This morning I realised that I reached the cap much earlier than normal. Twice.

[wind urchin]

no, you'll just get 4 of them suspended

[hard nacelle]

after taking your money. lol

[umbral tinsel]

gpt 3.5 is great anyways

[wind urchin]

20$ is a loss per month

[hard nacelle]

it shows 25 for mine too. But I have been counting...

[umbral tinsel]

gpt 3.5 is almost same as 4 or most completions

but so much cheaper and faster

nobody got money to build gpt4 service atm

no point to try, even microsoft can't make it accessible to us lol

20 msgs just to flex, huge loss still

[wind urchin]

if you used every GPT-4 message you could in the month it would normally cost you 320$, and I think that is a bit low

[carmine bane]

I have my bot built already with gpt4

[umbral tinsel]

is not a bad idea to automate it through web :p

to save money

[uncut ruin]

will the game change with plugins? right? 😄

[umbral tinsel]

plugins is basically langchain right now lol

[keen dock]

Thanks Guys for all your help. I am off to do my project again from the start, thanks for all your help it has been great and really helpful.

[umbral tinsel]

you can make your own already, is useful but not as good as it will be after we can fine tune it for reals

[wind urchin]

that would be abusing the system, I use the API when I need to work on stuff that needs long completions and reliability, I use the Plus for less important stuff and when I'm testing things out, I try and use 3.5 where I can and 4 when I need it, there are so many options 🙂

[umbral tinsel]

yeah but if its silly expensive then people will deffo do that

[tepid bone]

☝️

[umbral tinsel]

:p for complex tasks yeah 3.5 is useless

[fair iron]

that's why I don't use either

[wind urchin]

Dino's on GPT-10

[hard nacelle]

I felt like they pulled back a little on gpt4 as well. It feels a bit different compared to the first week. Still better than 3.5 though

[uncut ruin]

to use plugins, you need to connect the api and spend tokens?

[umbral tinsel]

they just reduced the context size

[fair iron]

Dino's on GPT-I don't want to break my academic integrity policy

[umbral tinsel]

at first it was running on the 32k model, now it is on 8k

[hard nacelle]

i see

[wind urchin]

It was never running the 32k Model

[umbral tinsel]

this context size is for both the prompt and completion so it's a huge reduction

[umbral tinsel]

are u sure? it was doing crazy stuff for me

and not getting amnesia

[wind urchin]

100% gold brick cast iron sure of it

[umbral tinsel]

duno

[wind urchin]

Certain 60 million year old lizards might have such access, but mere mortals .. pffft

[fair iron]

I've heard some people have had access to the 32k model, but I have nothing for sure

[fair iron]

I don't have api access at all

[umbral tinsel]

😄

[wind urchin]

Ahh, you just use putty into the SSH port on the super computer for training

[umbral tinsel]

32 model full context cost is like $1.92 right

[wind urchin]

32k 16 k in 16k out.. works out at $3.84 (i think)

[umbral tinsel]

tbh if you look at the charts then gpt4 is not too much ahead of 3.5 for a lot of stuff somehow