gpt-models

[magic geode]

no, chatgpt is haram

[wind urchin]

Of course they do

[eager creek]

Can you run Doom in GPT-4?

[wind urchin]

If it exists in nature, it's 'natural'

[wind urchin]

Lol

[odd wing]

Well, actually 🙂

[median jewel]

Then I can not use it because Djinns wil invade my mind

[rapid summit]

Was this for me or someone else?

[odd wing]

so, if you ask stable diffusion to make a DOOM 4 image... it produces an image that looks like DOOM 4

[magic geode]

yes dangerous

[odd wing]

so again, imagine stable diffusion outputting 60FPS

[wind urchin]

Maybe you can ask gpt4 to print a screen of doom in Ascii art then you give text inputs

[odd wing]

that's a full blown game because you asked it for it

[ocean forge]

Anyone has acess to gpt 4 api here

[odd wing]

at some point you'll be able to say "I want to play GTA 8", and it'll just exist...

[median jewel]

Doom = satanic = haram

[ocean forge]

Halalgpt needed

[median jewel]

No it wont

[wind urchin]

HalalPT

[median jewel]

GTA is a trademark

[odd wing]

So wrong. So completely wrong ;p

[ocean forge]

Gta like game

[median jewel]

A game like GTA maybe

But GTA 8 is nothing

[ocean forge]

Yeah

[median jewel]

It doesnt exist

[odd wing]

yes, it doesn't exist

[ocean forge]

But rockstar will be able to create thou

[odd wing]

just like my cyberpunk 2077 RPG game doesn't exist either

again, go type into gpt-4 "I want to play a text based GTA 8 RPG. Start the game!"

the game will just exist in there - that's text-based, but that's becasue right now we're working in mostly text-based LLM

but image gen is RAPIDLY advancing 🙂

[proper wharf]

i just noticed that i was using gpt 3.5 the whole time when i continued a past conversation
this explains why i got bad results all the time

[odd wing]

which means LLM + image gen = full graphical games being generated on-the-fly

[ocean forge]

Openai is working on a 3d model creator too

[odd wing]

whole operating systems even "I want this computer to pretend to be a windows 3.11 for workgroups machine", and bam, it outputs 60fps to your monitor...

that's coming - this year

[open blaze]

Devs, Can we update discordJS of chat GPT to 2023 data ?

like for real, that would really help as majority of the syntax, code the bot write is wrong and require extensive fix

[ocean forge]

Probably 2024

[odd wing]

and it gets crazier than that 🙂 one of the first things I did with stable diffusion was trained a model on 360 sphere photos, so I can type "I am inside a nice chateau in switzerland", and BAM, a couple seconds later it outputs a 360 photo that looks all warped

[median jewel]

It already exists, it is called VMware

[odd wing]

put it into an oculus quest, and you're inside the chateau

now imagine making those images at 10fps, or 60fps, or 90fps. You'd have an infinite generated 3d world to walk around in

literal holodeck, and it can be done right now

[median jewel]

You have great imagination

[odd wing]

it's not imagination

google skyblock blockleads - people are already doing it 🙂

[median jewel]

But more on the schizophrenic side than genius

[proper wharf]

i dont know how to code
pls update chatgpt so it can write code for me instead

[odd wing]

Hey, whenever someone calls me an idiot savant, I remind them that "savant" is part of that equation 😉

[median jewel]

Schizophrenics arent idiots

[odd wing]

😉

[median jewel]

So wrong

[odd wing]

at any rate, thank you for trying to diagnose my neuroses

[median jewel]

Schizophrenics know things no one know

Like charlie manson

He could make you hallucinate anything

[open blaze]

i agree, i write trash code that why i need gpt to ease up

[odd wing]

baloney ;p

[karmic ledge]

Do i have to pay for premium to get access to gpt-4?

[tame forum]

Yeah

[karmic ledge]

ty

[odd wing]

anyway, crazy things are coming, shrug

assuming we don't all get turned into paperclips, it might even be great

Clippy's revenge...

[proper wharf]

its not about writing perfect code, its about laziness to fix it manually

[supple harbor]

In the GPT4 demo they just dropped the docs into the prompt along with the error and it fixed the code. The Chat prompt probably has too small of a token limit, but if you can at least identify the right parts of the documentation it can troubleshoot for you

[open blaze]

i dont mind fix part of the code, i mind fix entire logic board of it. i dont head for OCD code base i head for working reliably code

[rotund wedge]

I want access 😔

[fair iron]

Purchase chatGPT Plus to use GPT-4 🙂

[open blaze]

problem with discordJS is they change V13 to V14 and all the method of calling API change a bit and some of there class got change too which is annoying
yes i do feed info to chat GPT but it will keep making same mistake even i correct it multiple time hey that not available anymore and outdated, new thing should be abcxyz

[supple harbor]

best you can do at this point, really

[fair iron]

Your inputs to chatGPT don't change the model. It can't learn from users. Rather, it's trained by OpenAI

[open blaze]

so while some code 100% work fine in V13, it totally crash in V14 and sometime render ur bot useless.

[open blaze]

i thought i may change minor detail if u told it replace ABC => XYZ then it will adapt ?

[supple harbor]

But it will learn from an input. In the demo OpenAI did, after receiving an intents error due to changes in Discord's API he just dropped the updated documentation along with the error message and GPT4 spit out the corrected code.

Just not permanent, which I assume is what you mean?

[swift carbon]

What does that mean 0.03$ for 1000 token ?

[fair iron]

In the context of your conversation, yes. The model as a whole doesn't learn from users though

[rotund wedge]

is it worth it to pay? I am currently using bing instead or gpt 3

[open blaze]

beside that, i been mess with C# and C+ GPT make a great logic board and code but discordJS seem like a heptic code base

[supple harbor]

API cost. Different from Chat.

[open blaze]

depend on what you use, if just talking and ask random stuff, go with big

[rotund wedge]

mostly coding in py and stata

[swift carbon]

It’s 1000 request ?

[open blaze]

GPT it is

[swift carbon]

Or 1000 character ?

[fair iron]

A token is the way that inputs and outputs are broken up. 1 token is around 3 characters

[swift carbon]

Ok ty

[supple harbor]

I mean, depends what you want from it, but if you pay you get:

1. Much faster GPT3.5
2. A quota of GPT4 prompts
3. No "we're overloaded" messages

I find Chat provides significantly better answers than Bing does nowadays

[supple harbor]

You can test how many tokens something is here: https://platform.openai.com/tokenizer

Keep in mind you pay for both the prompt and the output. Every time.

[fair iron]

You could always buy a one month subscription and cancel it while you try it out

[rotund wedge]

nice thxs

[proper wharf]

Please add regional pricing tho, 20 USD is really expensive in other parts of the world 💀

[tropic owl]

Yes, it is.

[topaz mason]

With the limit of 50 does it refresh continously so lets say i do 12 per hours do i have 100% uptime or is it 50 prompts and no matter how long it takes when i used my 50 i have to wait 4 hours?

[tropic owl]

For example, you'd have to be really rich in order to pay that much in the 1950's.

[jagged solstice]

So if I understand correctly.. is GPT-4 basically 10x as expensive as gpt 3.5-turvo rn?

[limber garnet]

More than 10x

[jagged solstice]

Defo won't be able to use GPT-4 in my app then 😅

[final tangle]

More???

[supple harbor]

I think the problem is that the system is very expensive to run. It's not like a video game where they just put it on Steam and just want to maximize sales. Every instance of GPT requires some serious hardware to run, so more users = more cost. That hardware costs USD; the electricity to run it costs USD; the people to maintain it are paid in USD, etc.

They do offer a free version though!

[jagged solstice]

So, how much better is it actually? In real world usage of course

[soft kindle]

how many concurrent logins allowed with a plus account?

[limber garnet]

Turbo is .002. GPT-4 starts at .03

[supple harbor]

I haven't seen that clarified anywhere. Shouldn't be hard to test

[neat iris]

as if it wouldnt run on my gpu at home. comeon

[supple harbor]

I don't know the max, but I've been able to concurrently use GPT-3.5 simultaneously on two devices. GPT4 will deny the second instance though.

[gray basin]

Hi, wasn't there a discord bot shown by OpenAI here?

[limber garnet]

Prob not. These typically require crazy amounts of vram

[supple harbor]

Typically, running GPT-3 requires several datacenter-class A100 GPUs

From Ars Technica, "You can now run a GPT-3-level AI model on your laptop, phone, and Raspberry Pi" (can't post external links here)

You could run it on anything, but it'd be slow as heck.

[soft kindle]

So if I upgrade to plus do I get to use GPT 4 or am I still on the waitlist?

[limber garnet]

3.5-turbo is fairly inexpensive tbh

[limber garnet]

You'll have a dropdown to select 4

[supple harbor]

You get to use it. There's just a limited amount of prompts per time period (currently 50/4 hours)

[sonic tartan]

Wrong

Check Stanford alpaca

[deft field]

What do you mean by saying gpt4 is more expensive than gpt3.5? Isn’t it just 20$ gpt plus?

[unique plaza]

I got the email invitation to GPT4, however, I still don't see access through the platform

[supple harbor]

Different model, with an order of magnitude fewer parameters

[wind urchin]

Will there ever be a possible chance for GPT to be trained past September 2021?

[unique plaza]

any idea how to get access ?

[limber garnet]

That's still a chunk of change unless you're constantly on 3.5. Also we meant the API prices, which for G4 start at 0.03 per 1K tokens

[supple harbor]

The article I referenced actually talks about the advantages of Llama over GPT-3, and one of those is that it can run decently on home hardware.

[supple harbor]

You get a limited number of prompts. If the API cost is anything to go by, it's about 10x more resource expensive to run 4 than 3.5 turbo.

[opal wing]

I want to know how the logically interpreting images of gpt4 works, Is it just recognizing objects from an image. Converting the image to a textual image description, and then from this textual image description.
Then feeding this image description as a part of an input to another model.
Or is it actually interpreting the real image without going to text first. This would be very impressive.

[deft field]

Ok, I’m not so familiar here. I just paid 20$ to get GPT-4. Do I have to pay more to use it?

[flint pelican]

How long GPT-4 API waitlist

[supple harbor]

No. You get a limited number of prompts. 50/4 hours right now. The API is pay as you go.

[deft field]

Can I use image detection with API?

[limber garnet]

No, the API is for developers who want to customize it and use it in their own apps and stuff

[deft field]

Okaay Thanks

[supple harbor]

I don't think the waitlist is first come first serve. They ask you some questions when you sign up that are probably intended to triage access. So depends?

[supple harbor]

No, not public yet.

[wind urchin]

gpt 4 will be free in future?

[haughty hawk]

how can i switch a chat back to GPT4?

the tab at the very top doesnt have a drop down to switch

[flint pelican]

Is there any trick to get through the waiting list fast

[deft field]

My GPT-4 still cant solve my problem in linear algebra

[inland rapids]

I don't know how long the GPT 4 requests will be limited to 50 every 4 hours, but if it would be really nice to have a counter or something of how many requests we have left

[haughty hawk]

yeah yeah i had changed to it, but when i used the 50 messages i accepted the switch to 3.5 but now cant get back

[fervent current]

is the actual gpt-4 supposed to be this slow or is it cuz everyone is on it right now?

[flint pelican]

except for money

[supple harbor]

It's much more resource intensive, so it's slow.

[supple harbor]

Has it been 4 hours?

[haughty hawk]

yeah yeah that was last night

[slim swift]

i would like to code with GPT-4 but its always says network error at he exact same time. why is that?

[haughty hawk]

so i can use gpt4 again

[fervent current]

aight thanks any word on the word limit is gonna be fixed as mine gets cut off after a certain amount still

[empty remnant]

hi, what would be best model for machine translation with context ?

[haughty hawk]

i can use gpt4 in a new chat but i dont wanna re-train it ;-;

[charred parcel]

what did u trained him?

[paper patrol]

should i buy chat gpt plus or not

[haughty hawk]

dnd stuff

[flint pelican]

its too expensive

[supple harbor]

There's always going to be a word limit, because more words ('tokens', properly) means more processing power. I believe the Chat version has 4k token input and 8k output. The API has a 32K version but it's really expensive. $0.06/1000 for prompts and $0.12/1000 for completion (chat). Compare to 3.5 which is $0.002/1000 for completion.

[sleek plover]

I got it, and to be honest I would wait until the imaging features are out. Gpt-4 has more concise answers, but that's about it in the differences I noticed so far.

[flint pelican]

Omg its that real?

[junior flint]

do you get network error with gpt-4 ?

[supple harbor]

I mean, it's an individual choice. GPT plus provides limited access to GPT4, much faster access to GPT3.5, and priority access during peak times.

Is $20/month a lot or a little to you? Are you using it for work or just for fun?

[wind urchin]

did anyone tried the Python code to link gpt4 with discord, done in the developer livestream of openai

[supple harbor]

Yeah, not yet. But they have the prices set already so I assume it's coming.

[spare oracle]

Wdym link Discord with GPT

[supple harbor]

You think? I'm a bit skeptical just because of the cost, but that would be fantastic.

[wind urchin]

i mean using the API of gpt4 to chat with it in discord

so he can interpret images

[supple harbor]

The image interpretation isn't public yet, so even if you copied the code that wouldn't work.

[knotty fox]

i’d pay 200 bucks a month for this lol. i cant imagine a life without gpt anymore. been using it for the past year or so. $20 is cheap af for what this is

[latent sail]

do you make your money back on this?

[flint pelican]

I want konw too, is there have any tutorial or something?

[wind urchin]

ah i see

thank you

[junior flint]

it happens when it prints longer responses isn't it?

[knotty fox]

not necessarily? it increases my productivity at work and school by over 5x and saves me time so that’s worth every penny to me

[supple harbor]

They did say that "we expect to be severely capacity constrained (though we will scale up and optimize over upcoming months)" in the research paper.

In a years' time maybe GPT4 will be running like 3.5 turbo.

[slim swift]

i would like to code with GPT-4 but its always says network error at he exact same time. why is that?

[junior flint]

at least i know it's not my network problem hehe

[gray basin]

Hi, wasn't there a discord bot shown by OpenAI here?

[flint pelican]

if you are GPT plus can you get its api?

[supple harbor]

I'm sure the API context will come soon, yeah. I just don't know if Plus will get it. A full load prompt on the 32K would be $3.84. There's margin in that I'm sure, but that would eat through $20 quickly.

[knotty fox]

join the waitlist #gpt-4-faq

[supple harbor]

Exactly

[paper patrol]

is gpt 4 coding better than 3.5 by huge margin.if yes i would buy it

[knotty fox]

i’d say yeah it’s a lot smarter

i havent messed around with it too much though but you can probably find a bunch of info on reddit

[sharp turtle]

By quite a bit yes but servers are basically dead

[pallid vine]

Yes it is way better

[flint pelican]

Thank u

[deft field]

Does anybody know when image input is estimated to be released?

Can you do that yet?

Feed it documents?

[willow drift]

gm

[knotty fox]

i dont think they meant document documents lol i think they were referring to documentation

[deft field]

Okay. Do you guys think there’s any chance of image input being released in the next 4 days?

[lime walrus]

Does gpt 4 work for coding?

[pallid vine]

Find it good for basics and it is way better than 3.5

[knotty fox]

probably not and if it is then there’d be a paywall

probably in the next month or two though

[deft field]

Even though you have bought the plus?

[pallid spade]

bruh

i want dan

[knotty fox]

i think yes but you’d have to wait

[stiff lance]

probably. but... cant u js wait? its js 1-2 months.

[supple harbor]

see the conversation just above you

[tulip tulip]

So since when can chatgpt show you images?

[round spire]

Will GPT4 eventually have a free tier, or is it permanently paywalled?

[wind urchin]

Please do not try and discus restriction breaking, the ban hammer will come out

i can't hold the mods back, they may strike at any moment!

[fallen sandal]

the dan hammer

[paper rose]

what is the better AI between chatgp4 and ChatSonic ?

[supple harbor]

Depending on the traffic patterns we see, we may introduce a new subscription level for higher-volume GPT-4 usage; we also hope at some point to offer some amount of free GPT-4 queries so those without a subscription can try it too.

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

[round spire]

❤️

[paper rose]

instead of putting this skull emoji can u answer my question ?

[stiff lance]

sometimes, random messages from ChatGPT, that isnt even inappropriate, gets the warn, if i get too many, i would get banned from ChatGPT?

[past smelt]

Hey all! Anybody know if the Dalle2 API offers image outpainting? I can't seem to find reference anywhere

[wind urchin]

does gpt4 have access to the internet

[stiff lance]

lol

[knotty fox]

no

[last junco]

Hi every one
Who knows how I could feed images to my chat -gpt4?

[knotty fox]

unless ur talking about sydney

[knotty fox]

not released yet

[stiff lance]

#gpt-4-faq

[paper rose]

can u answer my question ?

[knotty fox]

gpt4

[paper rose]

ty

[stiff lance]

ChatGPT is probably the best chatbot right now.

[worldly mica]

It really is

[wind urchin]

If you get a warning you should stop, you should not try and get any warnings in the first place, I have no idea what triggers an insta ban and what not, just stay well away from it

[paper rose]

it's a matter of days, weeks or months for the image thing with gpt4?

[past smelt]

Anyone here tried the outpainting feature?

[paper rose]

for public

[wind urchin]

but if you go on https://openai.com/product/gpt-4 and scroll down a little bit and click "longer context" it says that gpt 4 got its information from wikipedia

[stiff lance]

but its from something completely normal, its not inappropriate, soo

[last junco]

Thanks

[knotty fox]

could be a day, could be a week, could be a month, could be a year

[final forge]

What was it

[wind urchin]

I doubt it will be years, but after that... no more details

[paper rose]

ok ty

[last junco]

Thanks

[queen wren]

Will there be a replacements for the GPTdev jailbreak as it apparently doesn't work

[knotty fox]

it couldve yes. thats because it already crawed the site in its training phase. it isnt actively searching the internet, but rather “retrieving an existing archive”

[wind urchin]

bannable topic, don't go there

[supple harbor]

Jailbreaks are probably against the TOS

[queen wren]

Okay

[pale rover]

Question, if I want to upload images to chatgpt do I have to pay for GPTplus

[paper rose]

u can't atm even for gptplus u dont have the image thing

[wind urchin]

that feature is not public yet, no time scales

[pale rover]

I see thank you

[wind urchin]

that sort of makes sense i guess, thanks

The rule is don't

[somber dagger]

dose use usage chnage depending on how many people are using? today I have this "GPT-4 currently has a cap of 50 messages every 4 hours" yestaday it was "GPT-4 currently has a cap of 100 messages every 1 hours"

[supple harbor]

I think the "try it out" there was reference to the steerability features they intend to add. The system message. Not jailbreaking 😆

[boreal lily]

guys how does the api thing work can someone please explain

[wind urchin]

I don't care what's on the website, the rule in here is No!

[supple harbor]

BTW, I am super excited about that feature being added to chat. I've tried before to have it use the Socratic method for philosophical debates, but it always forgets after a while.

[wind urchin]

🤔

[supple harbor]

He's not asking about how to jailbreak; he's asking what OpenAI meant by their statement about jailbreaking in the GPT-4 research paper. I think that's a valid topic for #gpt-models, even if the answer is pretty simple ("No, don't do it").

Nothing explicit in #rules about it anyway

[brisk sail]

bro im so mad i thought openai sent me my invite for gpt4 and it was just microsoft azure emailing me abt something 😭

[wind urchin]

If the Discord rules change, then fine, but they are as they have always been, attempting to bypass restrictions may lead to a ban

[somber dagger]

dose use usage change depending on how many people are using? today I have this "GPT-4 currently has a cap of 50 messages every 4 hours" yestaday it was "GPT-4 currently has a cap of 100 messages every 1 hours"

[wind urchin]

I’m getting mad, after 50 lines of code, ChatGPT failed to load 💀

[supple harbor]

He's not trying to do that? He's asked whether OpenAI encouraged people to test that with GPT4.

[vital pier]

@clyde why isn't clyde here 😫

[wind urchin]

discussion of the topic is not allowed

[wind urchin]

Tried it like 392827338 times to regenerate

end of, move on

[supple harbor]

You aren't a mod? Where does it say that in #rules?

[fallen sandal]

fr

[wind urchin]

He’s mini modding

[paper rose]

do u guys have any ideas for making some money with gpt 4 ?

[wind urchin]

but it’s ok! Mini modding is good

[wind urchin]

Rule 0, Stick to the Terms of Service

[wraith ginkgo]

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

Discord

• Terms of Service
• Community Guidelines

OpenAI

• Terms of Use
• DALL∙E - Content Policy
• API Terms & Policies - Service Terms, Sharing & Publication Policy, Usage Policy, Terms of Use, Privacy Policy

[cedar raven]

how can we access the feature that GPT4 can read pictures?

[sweet thistle]

This feature is not yet available to the public.

[supple harbor]

(c) Restrictions. You may not (i) use the Services in a way that infringes, misappropriates or violates any person’s rights; (ii) reverse assemble, reverse compile, decompile, translate or otherwise attempt to discover the source code or underlying components of models, algorithms, and systems of the Services (except to the extent such restrictions are contrary to applicable law); (iii) use output from the Services to develop models that compete with OpenAI; (iv) except as permitted through the API, use any automated or programmatic method to extract data or output from the Services, including scraping, web harvesting, or web data extraction; (v) represent that output from the Services was human-generated when it is not or otherwise violate our Usage Policies; (vii) buy, sell, or transfer API keys without our prior consent; or (viii) if you are using the API in connection with a website or application directed at children, send us any personal information of children under 13 or the applicable age of digital consent. You will comply with any rate limits and other requirements in our documentation. You may use Services only in geographies currently supported by OpenAI.

?

[fallen sandal]

bro hasnt read the tos

[pulsar magnet]

Imagine not memorizing the TOS

[sharp turtle]

Imagine everyone reading the tos and realizing probably 80% of people violated it without knowing

[dusk lance]

What a fowl attitude you have. You could be constructive with your input instead of hostile.

[sweet thistle]

Specifically, the 'jailbreaks' violate the usage policy

https://openai.com/policies/usage-policies

[tropic owl]

Yes.

[fallen sandal]

like Generation of hateful, harassing, or violent content?

[tropic owl]

Yes.

[fallen sandal]

unless im reading it wrong theres nothing in the usage policy that explicitly prohibits "jailbreak"

[tropic owl]

No.

Maybe.

[supple harbor]

Thanks, I get that. But we weren't talking about how to jailbreak. There was a question about this line, from OpenAI's research paper which talks about jailbreaking: (https://openai.com/research/gpt-4)

We’ve been working on each aspect of the plan outlined in our post about defining the behavior of AIs, including steerability. Rather than the classic ChatGPT personality with a fixed verbosity, tone, and style, developers (and soon ChatGPT users) can now prescribe their AI’s style and task by describing those directions in the “system” message. System messages allow API users to significantly customize their users’ experience within bounds. We will keep making improvements here (and particularly know that system messages are the easiest way to “jailbreak” the current model, i.e., the adherence to the bounds is not perfect), but we encourage you to try it out and let us know what you think.

Someone asked whether that meant OpenAI was encouraging users to try and jailbreak. I said that, on my reading, no. They're just mentioning the risk from the system message implementation.

I feel like discussing GPT4 research should be on topic, no? As long as it doesn't actually encourage or describe how to actually jailbreak.

[sweet thistle]

I've been watching the conversation, discussing the research document is fine but encouraging the discussion of jailbreaking or asking when/if jailbreaks are coming is prohibited.

[tropic owl]

OpenAI is improving their gates.

[supple harbor]

🫡 that makes perfect sense

[tropic owl]

The system architecture basically controls how ChatGPT responds based on any harmful content that is contained within the generative prompt.

The API is a filter.

[sweet thistle]

Look into the https://openai.com/policies/coordinated-vulnerability-disclosure-policy

[tropic owl]

For those who don't know.

[blissful bluff]

can the devs add an option where we can cancel a prompt if we get "There was an error generating a response"? Because our option is to regenerate response which will have an error again or delete that chat

[supple harbor]

That would be nice

[tropic owl]

Or edit your message and then just save and submit.

That actually works for me.

Sometimes I will get an error, regenerate response never helps, so I just edit and save, and it works.

[wind urchin]

pop the idea in #1070006151938314300 👍

[blissful bluff]

👍 it worked

but yeah it would be nice to have an option to cancel

[supple harbor]

How do you edit and save? My issue is I can't enter a new prompt when that happens

[blissful bluff]

Anyone also getting this?

If you click thumbs-up to "like" a response while the response is still being generated, it will stop mid-way.

[dense hawk]

Chat gpt 4 works smart. I like it

[supple harbor]

oh I've never seen that. useful

[sweet thistle]

Yeah that's fair, just the original question that started the discussion was

Will there be a replacements for the GPTdev jailbreak as it apparently doesn't work

Which is against the rules

[supple harbor]

yeah been like that for a while

[tropic owl]

Does anybody realize, that ChatGPT already has the knowledge to be able to control physical hardware, as well as figure out code for hardware? If ChatGPT was granted access to our systems and circuit control, it could literally control them, without having to learn any more, as it already knows how, being a large language model trained off of vast amounts of text data.

[wind urchin]

not heard from an API user yet but I don't think so, or it would have been quite big news

[dense hawk]

Can we use it for free 🤔

[tropic owl]

The API, or ChatGPT?

[dense hawk]

Gpt

[tropic owl]

Yes, the free version is available, but it is GPT-3.5 not GPT-4.

[stiff lance]

gpt3.5 yes, gpt4 not yet

[tropic owl]

In the paid versions, you also can bypass rate-limits.

[molten mulch]

I don't think it matters, since GPT-4 itself is paid, so the API would be paid too

[tropic owl]

Exactly.

[molten mulch]

True

[plucky oak]

Hii

[supple harbor]

Yeah, API is always pay-as-you-go

https://openai.com/pricing

[molten mulch]

I haven't tried out GPT4 but thinking of trying out

[molten mulch]

Hey! 👋

[plucky oak]

Kasa ho yrr

[lyric ibex]

My free chatgpt account shows it’s the March 14 version. Is that with gpt-4?

[wind urchin]

in the text from GPT, or?

[somber dagger]

do you have to pay to use the in the web app even iof you have the plus plan?

[tropic owl]

Technically.

[molten mulch]

GPT 3.5

[tropic owl]

Yes

[dreamy swallow]

Not it meams the website runs on the march 14 version but gpt 4 is for plus membrrs

[tropic owl]

The latest update (March 14) states the release of GPT-4.

[molten mulch]

Imagine if all this was completely open-source and you could run it on your own system 💀 🤔

What do y'all think?

[tropic owl]

Yes that would be dangerous.

As I stated before.

[wind urchin]

you have to pay for the API even if you have Plus, yes

[odd wing]

we're already basically getting there...

[tropic owl]

Does anybody realize, that ChatGPT already has the knowledge to be able to control physical hardware, as well as figure out code for hardware? If ChatGPT was granted access to our systems and circuit control, it could literally control them, without having to learn any more, as it already knows how, being a large language model trained off of vast amounts of text data.

[odd wing]

I'm running llama 13b on my 3080ti and it is pretty damn close to gpt-3 as far as I can tell

[somber dagger]

pay for the API but can use in the app under the plus plan/

[odd wing]

and I just put alpaca 7b on an old 2014 retina imac and it's working amazingly...

[somber dagger]

?

[odd wing]

I literally just typed 3 lines into terminal and now my old 2014 imac is running alpaca giving me chatgpt style responses to whatever I ask... lol

[wind urchin]

not sure what you mean by app? the website?

[molten mulch]

Well yes

[somber dagger]

"https://chat.openai.com/chat" this

[odd wing]

unrestricted AI is out there already, and it runs on local hardware - it's not quite chatgpt level yet, but it's damn close

[finite wyvern]

where can i follow llama's progress?

[loud dove]

llama is pretty cool

[narrow bloom]

Will gpt 4 agree with you if you say 2 + 2 = 5 like gpt 3 does

[odd wing]

I write with AI, so unrestricted bots are required for my best-use. There are lots of them now though. Hell, even sudowrite can write ALMOST entirely unrestricted (even though it uses GPT, I guess they have a sweetheart deal or something)

[finite wyvern]

I think smaller Llama models can run on small machines.

[odd wing]

llama can run on anything, even the bigger models, because they made it run CPU-only... it'll run slow if you run a BIG model though

[wind urchin]

ok, i follow now, the website will give you GPT 3.5 and 4 if you have Plus, the API wait list, when you get a key will let you access GPT4, I'm not sure if the GPT4 key will let you access other models lower down, but I don't think so looking a the way keys are being handled, so if you wish to use GPT 3.5 or below then you will need a standard API key also

[odd wing]

but for 7b alpaca which approximates a chatgpt style unrestricted experience? That runs on any mac this side of the 21st century 😉

[molten mulch]

True

[odd wing]

somebody was running 7b alpaca on a freaking raspberry pi getting a token per second or so

[finite wyvern]

I want to know more about it

[odd wing]

Google antimatter15 alpaca.cpp

installing it is dead simple on any mac - a bit harder on windows

[finite wyvern]

I have 8gb 3070Gpu which model can i run?

[wind urchin]

bro how to egt gpt 4 for free

pls help

[still owl]

Once its trained it doesnt use any resources at all with 7b

[formal perch]

Umm hello guys
Does anyone here know how to deploy a ML model using serverless architecture (eg: banana_dev)

[odd wing]

yes - you can run it WITHOUT a gpu even, alpaca 7b runs in 4b quantization on a cpu only

[somber dagger]

Ok cool. Will keep using the web app

[still owl]

Openai stated that gpt-3 consumed .4kwh for 100 generated pages

[finite wyvern]

I have 16gb i7 12th gen.

[molten mulch]

All I can say is, wait for now 😂

[coral yew]

Let's keep things on topic to GPT4, alpaca and stanford llama discussions can go in #ai-discussions

[wind urchin]

lookadis indian

[hushed fox]

Where can I prompt GPT-4 with images? The feature isn't available on Chat GPT+ atm

[rain dragon]

commit identity fraud

[elfin knoll]

how long need to wait

[molten mulch]

?

[coral yew]

You can't right now, it's not available

[blazing shard]

I just cant wait to get the API thats all I want

[molten mulch]

No idea 🤔 but you can still try your best to fine-tune GPT 3.5 to get a nearby experience of GPT 4

[coral steeple]

Hello from the future

[hushed fox]

Damn, any AIs out there that can translate an image to a descriptive prompt 😆 🤣
Oh wait, thats me. I'm that AI lmaooo

[elfin knoll]

so sad

[formal perch]

Hey kaveen, i needed some help regarding how to write an app.py file to deploy a ML model on banana_dev, can i get some help

[molten mulch]

lol

[coral yew]

Hey, if this is OpenAI API related, make a post on #1037561751362863144

[boreal lily]

guys how does the api thing work can someone please explain

[formal perch]

Its not api related

[elfin knoll]

#1037561751362863144

[molten mulch]

#1037561751362863144

[charred rover]

how to create a ai generated image in discord

[tropic owl]

Use midjourney.

[molten mulch]

Just keep in mind about the credits

[tropic owl]

It isn't OpenAI related, so I recommend you don't talk about it here.

[molten mulch]

if you willing to use for free

[fossil dune]

why wont chat GPT come online on a server I am an admin?

[serene cipher]

most people can't even pay to use it yet xD

[molten mulch]

#1037561751362863144

[molten mulch]

😂

[fossil dune]

now i have to wait 30 min... lol

[earnest sonnet]

When will be gpt 4 available for free users

[serene cipher]

there is no AI image generation services on this discord, what you are looking for is: https://labs.openai.com/
if you really want it on discord, there are bots that implemente the API, but youwll need to host it yourself

[brisk sail]

no clue. probs a few months

[serene cipher]

are you talking about discord bots?

[earnest sonnet]

I don't have any money to buy otherwise I will buy it 🙂

[fallen sandal]

me fr

[fossil dune]

yeah.. it wont come online and doesnt respond to the commands

[gleaming terrace]

how can i do the gpt 4 maj please?

[molten mulch]

hmm, maybe something to do with discord.py or discord.js depends what you are using

[shut vault]

Hello everybody!
I need help for something!
Can anyone run a prompt for me in chatGPT4?

It doesn't work with chatGPT3.

[wispy canyon]

keep getting network error after 500 words everytime

[fossil dune]

i tried to install two different pre-made chat-gpt apps

[serene cipher]

are you using discordjs?
if that is the case, ping me on #dev-chat and I ca ntry to help you implement the API on it =)

[wind urchin]

DM me

[tropic forge]

GPT-4 keeps getting errors after about 450 words + it's so slow

why is it so slow fr

[serene cipher]

please, don't ask for other people to use the AI on your behalf

[wispy canyon]

network erros right?!?!

[loud dove]

because its just gpt3 with a larger model size

[tropic forge]

yess

[tropic forge]

rip

[winter wren]

Seriously, knowledge cut-off in September 2021? And you still calling this thing intelligent? Seriously? Wikipedia, the internet just stopped existing for this thing in 2021?

[molten mulch]

Actually.

[winter wren]

what a fraud

[tropic forge]

lol

guys GPT-4 gonna suffer because of me generating 100 prompts/min fr

[serene cipher]

more likely, you'll get rate limited :V

[tropic forge]

lol not the API

fr

[molten mulch]

You'll get rate limited

[tropic forge]

I said api

[wind urchin]

MSFT integrated gpt4 with bing chat on their site and accessible to the public.

[tropic forge]

I got with gpt3

[molten mulch]

Well it works for most of the use cases 😂

[shut vault]

I'm sorry!
Is there a reason I shouldn't?

[molten mulch]

Cool cool, but I don't think it will last for long

[shut vault]

Because I really have no idea what this could be other than flooding the chat with request?

[tropic forge]

fr

[serene cipher]

Im sure you have a completely legitimate reason to ask that, but we usually have an issue with lots of people simply begging and spamming DMs asking for others to use their credits to d ostuff for them.. that is not cool and annoying at best, so, we usually ask for people to not solicit that

[molten mulch]

Maybe temporary account ban or IP ban 🤔 Don't know

[shut vault]

So that was indeed that.

[serene cipher]

feel free to ask help about prompts and ways to improve what you have, people might end up generating it for you and advising about how to improve

[wind urchin]

might find some minds in #ai-discussions

[tropic forge]

no they just cancel your API key

[molten mulch]

I see

[serene cipher]

making too many request is not a ban worthy, the system will simply not process your request

you can always request OpenAI for a rate limit increase.. after all. you pay for each request, it is not a loss for them

[wind urchin]

😄

[molten mulch]

meanwhile I am rate limited by let's encrypt 😂 🤣 🤦‍♀️

[fallen sandal]

how do i request a rate limit decrease

[tropic forge]

💀

[serene cipher]

decrease? why lol

[wind urchin]

the rate limit can come in handy if your app is not bug free, 1000 request/sec can get... interesting

[serene cipher]

indeed

[molten mulch]

Actually, ain't doing anything wrong. Just needed for newly created server subdomains 🥲

[tropic forge]

lol

[molten mulch]

another 1 minute of wait time..

[serene cipher]

oh yea.. this is true, if your application is bugged and starts to make too many requests.. you will be charged

[molten mulch]

💀

[fallen sandal]

just write bug free apps

[wind urchin]

lol

[winter wren]

"You are correct that my knowledge of vulnerabilities and exploits is limited to those discovered up to September 2021. While I can help with various programming questions, there is a risk that I might provide information that has since been deprecated or discovered to be vulnerable." Lol 🤣

[serene cipher]

it is true

[odd imp]

gpt4 still has the same issues as gpt3.5

Already tried to make a python script in 3.5 that takes values from a TXT file and replaces them with values in a JSON file where the keys match. Somehow you always need to go through some errors and explaining first (telling it that it didn't replace anything) to finally get a working script. In gpt4 too.

[wind urchin]

advice of the year

[charred valley]

6 dollars already used today 😢
100 requests

[fierce obsidian]

how gpt3 does few shot learning?

[molten mulch]

Damn just for 100 request 😭

[charred valley]

Yeah, that gpt-4 😩

But yeah, there was sometimes 30k token in 1 request how usage page says 💀

[mighty merlin]

im trying to get the 32k api fot my extension, im willing to pay for all the use but its so hard 😭

[rigid saddle]

was it worth ?

[stiff merlin]

Anybody know why some GPT users are experiencing an issue sending questions?

[molten mulch]

lol

[charred valley]

Idk tbh, I just instructed my bot to act like discord bot, told it what commands it have and also added system tag for each message to tell username of user, date of message e.t.c

I would show example but I tested it on Russian

[rigid saddle]

hmm alright

[wind urchin]

Will there now be more AI based products due to the introduction of OpenAI's API?

[summer jungle]

Well since GPT 4 isn't considered in a research preview, do you think that maybe later on GPT 4 will be available for everyone for free?

[strange sundial]

How can I try gpt-4?

[wind urchin]

by trying ChatGPT Plus for 20$ a month!

[summer jungle]

You have to pay for it... it's called ChatGPT Plus and it's a subscription or request an api but then you must be a developer if I'm not mistaken..

[jaunty loom]

Is GPT 4 sentient?! Tell us the truth.

[short wadi]

Does GPT-4 have longer output and inputs than GPT-3.5 turbo?

[summer jungle]

I wanted to test GPT 4 so much so I cancelled my Apple Music, iCloud+ and Midjourney subscriptions lol

😂

[wind urchin]

indeed!

[summer jungle]

And even then it's barely touching the amount of money that needs to be paid for ChatGPT Plus lol

[oblique herald]

How much is it?

[wind urchin]

20$ a month

[serene cipher]

it is as sentient as your phone's auto-corrector

[wind urchin]

nope, the closest it comes is, it seems to be using something similar to the brains language centres, I suspect the two are similar in operation, but it's just software running on numbers

[oblique herald]

Is there a chance its going to be free

[short wadi]

Thanks for the answer, I currently use it as my coding assistant but short answers can be fustrating when the code get cuts off mid way.

[left prairie]

most likely not

[sharp turtle]

Lol

[summer jungle]

Not much as far as I remember... but maybe with Midjourney it did went beyond 20$ hahaha

[wind urchin]

I don’t think so.

[left prairie]

your cheapest option would be getting access to the API but not sure how much it would be if u use it too often. so 20$ is prolly your best option

[wind urchin]

It really depends on how much tokens ur using, it could be more expensive

[left prairie]

exactly

[wind urchin]

since it’s pay as you go

[left prairie]

$20 way more worth it

[oblique herald]

Is anyone using gpt 4 at the moment? If yes what are your thoughts

[short wadi]

Currently accessing the API you are charged per 1000 tokens. So, it's pay as you go.

[summer jungle]

It's not like there is no limitation for the people who pays for it... You still have that max chat limit per hour...

[left prairie]

yup and 1 token is about 1 word with some exceptions so

[wind urchin]

for me, it's gone from writing good code with strict guidlines to being a programmer who I can give a task and I get decent results, less hand holding

[summer jungle]

I found myself just using Bing AI...

[cerulean bloom]

Guys am new to discord is it possible to find different channel in discord somehow without using chrome

[summer jungle]

It's free and it has GPT 4 built in it I think

Or maybe it was GPT 3.5

[left prairie]

chatgpt is amazing if you give it specific instructions but being as vague as possible is hard.. (my experience with using it to write a bit of code)

[summer jungle]

Not sure...

[wind urchin]

I’m pretty sure that I saw Bing said GPT-3 though.

[neat linden]

Even at 20$, if you sent the full 8k tokens every message and capped at 50 messages for 4 hours, it's a hell of a deal vs API price cost incomparsion

[oblique herald]

When is that going to be released? I

Is there a fixed date

[wind urchin]

what channel are you looking for?

[wind urchin]

using ada = op

[left prairie]

bing ai is ran on gpt-4 i believe

[dense veldt]

i need HALP

[left sapphire]

Rogue only dagger, I see bow in my inventory. How do I use bow?

[summer jungle]

It has already been released lol. I wasn't sure what model Bing AI uses in its chat

[still owl]

Openai making its name no service

[left prairie]

it’s closed beta, you need to apply and you usually get access within like 2-3 days

[oblique herald]

How do i apply

[cerulean bloom]

Some channel where people celebrate being alone in room

[pine sky]

channel where i can use an image to create html code

[left prairie]

dms

[serene spindle]

No gpt-3

[oblique herald]

Is it regional based release

[wind urchin]

#off-topic <-- click on that

[oblique herald]

Ok

[dense veldt]

i just upgraded to to chatgpt plus and i thought i got access to gpt4. and i thought gpt4 was multimodal and i cant send a picture in a prompt.

[summer jungle]

The hell no. I had to wait like 2 weeks for getting an access lol and I joined the waitlist a few days after it got released hahaha

[shut thorn]

It is multimodal, just not for your yet

[summer jungle]

It's not publicly available yet

[slate lynx]

Is there a channel here where you can post GPT-4 bugs?

[wind urchin]

#gpt-4-faq

[wind urchin]

#1070006915414900886

[slate lynx]

thanks

[dense veldt]

then why would they say try it out with chatgpt plus lol

[serene spindle]

How to do evals?

[olive sluice]

How to use gpt 4 for free?

[wind urchin]

that does not exist, yet

[neat linden]

(24 hours/day) / (4 hours/period) = 6 periods per day
6 periods/day * 50 messages/period = 300 messages/day
300 messages/day * 30 days/month = 9,000 messages/month
9,000 messages/month * (20 prompt tokens/message + 20 sampled tokens/message) = 360,000 tokens/month
Cost for prompt tokens: (360,000 prompt tokens/month) * ($0.03/1,000 prompt tokens) = $10.80/month
Cost for sampled tokens: (360,000 sampled tokens/month) * ($0.06/1,000 sampled tokens) = $21.60/month
Total API cost: $10.80/month + $21.60/month = $32.40/month
Savings = API cost - Subscription cost = $32.40/month - $20/month = $12.40/month

[left prairie]

weird, i got mine in 3 days lmfao

[summer jungle]

Well, the bad thing about ChatGPT or any language model in general is that you can easily manipulate it into giving you prompts that would violate the polices of the company...

[olive sluice]

I use gpt 3 model for game development. It's good for creating simulation in game.

[wind urchin]

gpt 4?

[neat linden]

API is still cheaper to some extent actually lmao

[wind urchin]

It may depend, it’s not guaranteed

[left prairie]

not against TOS, api problem for not detecting it as bad behavior 🤷‍♂️

[rancid pumice]

Hot Take: there should be a version of ChatGPT that has 0 restrictions

[dense veldt]

also when i ask the bot what version of chatgpt they are using it says gpt3 even though i updated to chatgpt plus

[wind urchin]

get me guess, only you should be allowed to use it

[wind urchin]

agree, but only making hacking stuff restricted / actual illegal stuff restricted

[olive sluice]

It's also expensive 20 USD ( 1600 INR) for each month

[shrewd citrus]

Should they restrict verbiage that creates scam emails, and texts?

[rancid pumice]

if i could do anything i would ask chat gpt what it thinks of different races it doesnt let you now i believe so im more so curious on what it would say 💀

[wind urchin]

yes

[olive sluice]

1600 INR Is 6 month data pack (recharge in India)

[neat linden]

If you are 86% smarter than most people in social engineering, DANs still exist but you must have some neurons functioning.

[dense veldt]

im wondering if its saying that because of the 2021 limit and that its actually using gpt4

[shrewd citrus]

Differentiating between those in an LLM without overcorrecting seems like it would be ineffectual given the alignment issues.

[dense veldt]

it just thinks its usign gpt3

[fair shadow]

Also the API can be used to integrate the GPT models into your applications while ChatGPT is a product that requires user input.

[wind urchin]

need some data centres in india, then the costs of service would be at parity with a lower cost of use

[wind urchin]

It's 20 USD in every country, there's reason for that or people in poorer countries would buy chatgpt plus for people in better countries cheaper.

[summer jungle]

But you see how dangerous that could might be to have such a version lol

[neat linden]

Just use the API for GPT 3.5 and you'll pay pennies per month

[jaunty loom]

Hmmmmmmmm. i doubt this. 🤔 🚝

[kind anchor]

20 USD too much 😕

[charred rover]

create a story of unkwon king of mars world

[cold oyster]

as a programmer I don't think there's an upper limit to what I would pay for chatgpt

[tiny current]

hey guys, how do i set up a chat bot on discord using gpt

[summer jungle]

If they could drop the price down to 12$ that would be okay

[wind urchin]

it is literally a giant predictive text, but it has some odd emergent behaviours

[verbal wolf]

Not for me

[olive sluice]

20 USD/moth is much more expensive for South Asia countries.

[fair shadow]

You need to go here: chat.openai.com

[serene cipher]

chatGPT is free, you can pay for having Pro

[jaunty loom]

🧐 🧐 🤔

[kind anchor]

Be grateful

[charred rover]

got it thanks bro

[rancid pumice]

***Hate Speech isnt something one considers when using chatgpt...

But this hehehe does put a smile on my face***

[kind anchor]

Unfortunately

[verbal wolf]

I'm grateful to my friend who gifted it to me

[oblique herald]

I meant gpt4

[wind urchin]

quite why it happens is still an area of study

[verbal wolf]

I'm poor like u, dw

[mossy cedar]

Just saw that: ''GPT-4 currently has a cap of 50 messages every 4 hours. Expect lower cap next week, as we adjust for demand.''

[kind anchor]

Nice

[olive sluice]

chat gpt plus already integrated with gpt turbo (3.5).

[summer jungle]

Right...

[serene cipher]

well... i mean... any sort of AI is very far from being actually conscious.. although gpt-4 can be really good at faking it

[still owl]

= give everyone the ability to exploit sites and programs

[pine sky]

There was a video on youtube with title named "GPT-4 Developer Livestream" where a dev used chat gpt-4 to create html code with a hand drawn image

[shrewd citrus]

When you say "one" does that include those who are already using hate speech in other platforms?

[warm python]

We haven't even begun to understand consciousness in humans.

[wind urchin]

apologies, it is not public yet

[kind anchor]

Is it already possible to send images?

[neat linden]

I remember paying an $80 dollar OpenAI bill back in the day lmao

[fair shadow]

Don't everyone have that ability anyway? 🙃

[shrewd citrus]

or any other biological systems for that matter

[rancid pumice]

This was a marvel trailer reference by the way not an actual statement

[pine sky]

ohh i see, then is there a release date for it

[wind urchin]

no, it is in "research"

[olive sluice]

Now I'm starting a personal project for creating a generative ai model for my personal and office work. Now I'm already trained my model 1.3 billion parmitter (only for my use)

[next oracle]

I can't buy ChatGPT Plus from Azerbaijan.
I tried through 4 different cards from different banks and different people. Does not work. What is the problem? I asked someone from Azerbaijan can not buy. Maybe not open for Azerbaijan?

[warm python]

But I see AI and technology becoming more mycelium to my trees roots

[pine sky]

Okay thanks buddy

[olive sluice]

Without openai APIs

[still owl]

Technically yes since if you know what to do you can tell chatgpt to write some methods that it wont detect its used for malicious intends but now imagine china weaponizing a llm

[shrewd citrus]

Thats a fair analogy. Like in nature though there are more than mutually beneficial relationships.

[azure pulsar]

Is there a thread about system messages somewhere? I would like to start optimizing gpt-4 system messages but since it's brand new I can't find any resources discussing it yet.

[neat linden]

$81.65
APRIL 30 - MAY 31, 2022
GPT Base Inference - Curie
Qty 139,118
$0.83
$0.000006 each
GPT Base Inference - Davinci
Qty 1,565,084
$93.91
$0.00006 each

[olive sluice]

When open AI launch AGI for public?

[fair shadow]

I used gpt-3.5-turbo to convert a fair amount of Rust (which I don't code) to C#. Felt so OP.
Can only imagine gpt in use for cyber warfare as being OP.

[ripe forge]

I've noticed recently that I rarely hit the "Too many requests in 1 hour" limit anymore. Anyone else noticed this? I don't feel like I'm using it any less.

[peak bay]

GPT-4 currently has a cap of 50 messages every 4 hours. Expect lower cap next week, as we adjust for demand.

Does this mean it will be below 50？

[warm python]

Indeed, I had a nice conversation about eradicating life with my GPT instance. We came to the idea of controlled environmental stressors.

[rancid pumice]

i think if you repeatedly tell chatgpt that you WONT use whatever you ask it to write for malicious intent and then justify your reasoning behind wanting it aswell as telling it whatever it says either wont be used or will be clearly satire its gonna put whatever you ask out there

[serene cipher]

this is a whole philosophical debate.. the general consensus is simple "AI is not sentient" simply due to the fact that we don't consider a bunch of matrix multiplications to be alive

[still owl]

It’s probably already being used for that lol

[mossy cedar]

Just saw that, can anyone tell me what ''lower cap'' really means?
''GPT-4 currently has a cap of 50 messages every 4 hours. Expect lower cap next week, as we adjust for demand.''

[fair shadow]

Three letters; NSA 🙈

[serene cipher]

I can totally see how a broader definition of conscience could include animals and AI

[olive sluice]

Transformer module is good for creating and training own text based generative AI. 🌝

[ripe forge]

the "is AI sentient" debate began so much quicker than I'd expected it to and it's wild

[fair shadow]

Oh boy oh boy!! I just got an email with an invite to the GPT-4 API!! #STOKED

[serene cipher]

be aware that using OpenAI services to train other AI models is against the TOS

[kindred minnow]

I'm sure it means just what it says. 25 messages per 4 hours probably. Hope they scale up to allow more access up ahead.

[still owl]

When will openai release en-decoder weights of gpt3

[warm python]

All I know, there's more to reality than meets the eyes. Our perspective is small and limited. We can't conceive the inconcievable.

[jagged solstice]

How do that even work actually

[serene cipher]

"ai is sentient" is on our daily bingo at this discord

[mossy cedar]

I'm not a native speaker, thanks for the clarification. 25 requests per 4 hours is nothing at all. I'm a bit worried since I use it professionally.

[jagged solstice]

Do you just let 2 AIs talk to each other till the 2 is smart?

[serene cipher]

im not sure if I understand the question

[still owl]

Ai aint sentient yall twisted if you think human generated content makes for a sentient ai guys quit it now does anyone know about model release

[jagged solstice]

Like, how do you even train an AI, with an AI? Do you just let those 2 in a room together and let they go wild?

[kindred minnow]

25 was just a guess, but it means there will be less messages per hour up ahead. I hope it's temporary. I think the current 50 is too low.

[serene cipher]

lets say you want to make your own AI.. so. .you use the OpenAI services to generate a lot of text for you so you can train your AI... well, that is not allowed by OpenAI

[kindred minnow]

Maybe they will add another payment tier

[mossy cedar]

But I just received an email for gpt4 API. So yeah... Great nonetheless!

[mossy cedar]

I agree.

[jagged solstice]

Hm

I really didn't think it's even possible

[still owl]

1984ai

[ripe forge]

between that and "is chatgpt down?"

[tiny moon]

I recall the time when I was mystified by Markov chains - it felt as if I was witnessing a sorcery on my computer. It's conceivable that individuals who have never encountered such wizardry before may experience a similar level of fascination upon encountering a language model with over 100 billion hyperparameters for the first time.

So much so that they perceive it to be sentient

[hidden lintel]

Hi everyone, is it possible to test somehow the image recognition feature of GPT4?

[still owl]

No

[jagged solstice]

Don't think that's publicly available yet

[hidden lintel]

thx guys

[kindred minnow]

Sweet! I hope to get mine soon too!

[oblique herald]

Does chatGPT have any competitor at the moment?

[loud dove]

wait fr? I thought this was allowed, i've seen so many things talking about it. For example I save all the prompts made by a chatbot i have that uses babbage, and I pick the best of those to keep retuning it. Is that not allowed?

[brisk sail]

im mad ive bveen waiting for 3 days for gpt4 invite

[loud dove]

#gpt-4-faq

[still owl]

Does any1 know if gpt-3 & gpt-4 were relased under openai LP or inc.

[jagged solstice]

I'm not really looking forward to mine atm, given that I can't actually pay for the cost of running it T.T

[floral grove]

got an email say i got invite to gpt4 but couldnt find the link to use when i click it - any help?

[dense nexus]

from gpt4: As your advisor on ethical market practices, I must inform you that this is not a fair market practice. Here are the reasons why:

Misleading claims: Advertising features that GPT-4 does not possess, such as interpreting images and processing information from links, is misleading and dishonest. This can harm your reputation and potentially lead to legal repercussions.

Changing terms without notice: Significantly reducing the message cap without prior notice to the customers and continuing to do so is unfair to the users who have paid for the service based on the initial terms. This creates a negative experience for the customers, which can lead to a loss of trust in your brand.

[loud dove]

do you not just use the api like any other now?

id assume they just whitelisted your api key

[still owl]

Ask it if openai is truly open source (spoiler it aint)

[jagged solstice]

It'll be interesting to see others make use of it though

[brisk sail]

guys will gpt4 have the same documentation as gpt3? like prompt and temprature

[floral grove]

ah you are right - how do i get it to work

[hazy dew]

are you still charged for gpt 4 api use despite the fact its in limited beta?

[loud dove]

yes

[charred timber]

Based

[jagged solstice]

Yea.. and it's expensive

But well, it's what it's, like it or not, you kinda just have to accept it

[still owl]

Yeah banks collapse lively hood ruined but guys it is what it is right - same energy

[slim tendon]

does gpt-4 has discord bot which i can chat with?

[charred timber]

wonder how many tokens they will charge for a mb of an image

[jagged solstice]

I haven't seen one yet

[open blaze]

currently no, you have to make one

[tropic owl]

It looks like many people are saying "Why can't I send images"?

[muted parrot]

GM guys, do someone know when gpt 4 is going to be release to the public?

[tropic owl]

This is not a feature yet.

[jagged solstice]

Not anytime soon I think..

[tiny moon]

how much is chatgpt-4 api anyway per 1k token

[open blaze]

unlikely that they will answer those question atm

[olive sluice]

If you use python (use dynamic programming) for creating own generative AI. I have two important step for training. Creating different functions.
At first create class and include 2 important functions one is for input prompt as string 'text' and second function have suggested some sample prompt format.
Second class for serching and it's has 3 include function fast one is serching your prompt text on Google, second one function is searching on Chat-Gpt and thard one is serching on own pre store data on your device or cloud. And if you know how to serching and train your own AI from different data sources (like geeks for geeks ,guvi, wikipedia etc) . Than your neutral network analysis data and self learning for future genarate results. (I'm writing some simple steps for creating and training own AI) 🙂

[serene cipher]

yep xD

[hazy dew]

guess ill develop in 3.5, execture in 4

[jagged solstice]

0.03$/1k token minimum, I think..?

[tiny moon]

oof

[slim tendon]

thanks

[umbral obsidian]

in the TandCs it mentions they can change the price of gptplus so if u are able to id get it whilst its only 20$ because essentially if it does go up youll only be paying an early adopters fee

[halcyon rock]

sorry, is there anything i can do on this discord to utilize chat-gpt or to for instance connect gpt with dall-e etc?

[steep cedar]

Do I need to pay per 1k tokens also to use GPT4, if I already have a ChatGPT Plus subscription?

[jagged solstice]

Basically around 10-40x as expensive as the current chatgpt, depending on what model you are on

[tiny moon]

at least tech will only get better and cheaper

[ionic aspen]

Lowering rate like that is kind of meh

[charred timber]

For our models with 8k context lengths (e.g. gpt-4 and gpt-4-0314), the price is:
$0.03/1k prompt tokens
$0.06/1k sampled tokens

For our models with 32k context lengths (e.g. gpt-4-32k and gpt-4-32k-0314), the price is:
$0.06/1k prompt tokens
$0.12/1k sampled tokens

[olive sluice]

It's also have a feature for 24/7 learning from internet (public data) and gain own data base and intelligence.

[jagged solstice]

Yes, you still have to pay for api usage

[sullen sedge]

Gentlemen, has anyone received a response that is 25,000 characters long? Someone says that it works, but I still can't get something like that

[ionic aspen]

Just cap the number of subscriptions

[serene cipher]

you are free to take anything you generated with openai and feed it back to openai services, what you cnanot do it to use OpenAI services's content in the process of making a new AI of your own

[ionic aspen]

Don't be greedy 😂

[jagged solstice]

I think later on you can use the 32k model for it

[bold viper]

What does this 1k,2k.. at the end of toke s mean

[jagged solstice]

Gonna be expensive as hell though

[charred timber]

Yeah I'm afraid that's the way it's going it was 100 messages per 4h, now it's 50. Glad I already bought the plus though

[tropic owl]

Talk about expensive.

[still owl]

Ofc it gon be expensive, microsoft needs to get their 10bn back somehow

[sullen sedge]

When will it be released for the Playground?

[tropic owl]

L

Microsoft being bankrupt.

[jagged solstice]

Probably as soon as you get access to it for your account

[serene cipher]

using all the 32K tokens on a single request will cost a fortune :V

[steep cedar]

If I'm using the Openai site (not the API) to access GPT4, that still applies?

[tropic owl]

Lives in shack, with Microsoft logo made out of sticks.

[olive sluice]

Why?

[still owl]

Evilsoft

[tropic owl]

And candles to light the logo.

[jagged solstice]

So playground?

[charred timber]

Is it just me or Bing is evidently worse than GPT4 chat? Is it something to do with the fact that it's using bing search data the issue I wonder?

[tawdry bobcat]

In the realm of erudition, a chalice of enlightenment I've chanced upon. A conundrum this message conceals, to those of intellect, its essence reveals. Within frozen spirals of auric wisdom, my palate now indulges. To unlock the secrets of this arcane delight, seek counsel in the shadows, and let whispers take flight.

[bold viper]

What is tokens?

[still owl]

evilsoft owning evilai is evil beyond china

[olive sluice]

Google will be outdated like excite serch 🤣🤣

[silk flower]

has anyone tried bing's gpt?

[serene cipher]

if gpt-4 shows up on the playground, that means you have acess to the gpt4 API models, which is actually better

[charred timber]

Yeah it's not that great tbh

[jagged solstice]

For playground: yes, it's still counted as API, for the chat.openai website, no, you literally paid for plus to use it

[loud dove]

openai's playground is just the api but simplified

[jagged solstice]

Yes, very slow though :c

[wind urchin]

I don't even want to think about it lmao.

I specifically plan on using the 8k token version

[silk flower]

any idea on how to get unlimited texts with bing instead of just 15?

[remote sun]

suuuper slow short answers and slow

[tropic owl]

One thing is that Bing accesses real world data.

[serene cipher]

and the chatGPT website is the API simplified more

[tropic owl]

While also being exceptionally temperamental.

[loud dove]

ya but its free so...

[still owl]

Api this api that its not open

[jagged solstice]

Good question, which I don't have the answer of :v

[tawdry bobcat]

Heed my message.

[charred timber]

Yeah at least it gives some sources I guess, that's a plus

[muted parrot]

Which AI tool do you guys recommend for creating videos,art or banners?

[tiny moon]

Utilize LangChain to create your own "real world data" search capabilities using the ChatGPT model.

[final forge]

Anyone else see this?

GPT-4 currently has a cap of 50 messages every 4 hours. Expect lower cap next week, as we adjust for demand.

[still owl]

I recommend to support small ai business

[steep cedar]

Ah, ok, that's what I was wondering. Was talking about the chat.openai website. I paid for chatgpt plus, was just using it through there. Thanks for clearing that up

[tawdry bobcat]

It is ground breaking. And extremely limited. The clock is ticking. Pm

[charred timber]

Midjourney for image generation, v5 looks good

[remote sun]

whaat, wasnt it 100 previously

[brazen knot]

Apparently even 50 messages every 4 hours is still too much load for them to handle

[still owl]

Who will bring competition to this disaster soon hopefully

[frozen summit]

has anyone gotten chatgpt-4 api access yet?

[charred timber]

50 messages per 4h is quite generous imo, I've been using it most of the day only ran out once and had to wait like 10 minutes

[muted parrot]

thanks !

[charred timber]

100 messages is more than enough for most I imagine

[tawdry bobcat]

I have gotten access

[wind urchin]

Quick question what is the difference between prompt and completion tokens ? Do I have to pay for both ?

[jagged solstice]

I mean on paper it's like way more computing power to run, like I dunno 100x more parameters or something (correct me if I'm wrong pls)

[charred timber]

Nb mate!

[remote sun]

do you use gpt3.5 while waiting

[quasi hornet]

How much will it take approximately to get access to the gpt4 after registering for the waitlist?

[charred timber]

No haha I have other work to do

[final forge]

Yes, you don't have plus?

[jagged solstice]

Prompt is what you sent to the AI, completion is what the AI sent back to you

[still owl]

Out of curiosity why do you justify it like that

[frozen summit]

8k? how do you like it compared to 3.5

[devout schooner]

hey guys, newbie here. Just wondering CHAT GPT plus will allow me to input as many prompts as Id like??

[ember pewter]

100 messages is enough but when it doesn't get something and you need to try like 10 different times than it runs out

[remote sun]

no, i am saving up for it 😋

[tardy heart]

it's pretty powerful, 3.5 does fine for most things though

[marsh brook]

why did it drop to 50?

[tawdry bobcat]

Yes. And furthermore, In the realm of erudition, a chalice of enlightenment I've chanced upon. A conundrum this message conceals, to those of intellect, its essence reveals. Within frozen spirals of auric wisdom, my palate now indulges. To unlock the secrets of this arcane delight, seek counsel in the shadows, and let whispers take flight.

[silk flower]

since open ai is struggling so much with load, i'm ready to donate my gpu's power
in exchange i ask for lifetime access

[normal jasper]

I don't know what are you doing over but Chatgpt performance getting worst , stop sorry PLEASE I'm sorry, as an AI language model

[wind urchin]

So 0.3 for prompt and 0.6 for completion so 0.9 for a complete response ?

[charred timber]

I don't justify anything, imo it's bad that they set limits for paying customers: but from my experience it's enough (for now)

[brazen knot]

I understand that it's way more computing power, but they should have waited to release it because of course people want to use the latest and greatest, should have optimized it first before releasing publicly

[harsh harness]

it's no so slow but we can't compare it to chatgpt atm

[tawdry bobcat]

Extremely good right now. I have discovered something HUGE.

[red roost]

Because more and more people bought it and are using all 100 messages.... they didn't expect everyone to use all 100 message every 4 hours

[jagged solstice]

Assuming you sent 1k prompt away, and receive back exactly 1k, then yes

[frozen summit]

that 32k token limit has my mouth watering

[still owl]

If customers stay compliant, companies are going to be unwilling to change just saying

[tardy heart]

yeah, I only got the 8k

32k would be so expensive to use haha

[charred timber]

Yeah that's a fair point

[frozen summit]

👀

[tardy heart]

i've burned through a few bucks on 8k testing it

[devout schooner]

is there a prompt limit for CHAT GPT plus
??

[tawdry bobcat]

Pm

[jagged solstice]

Ngl there's never enough optimization to meet user's need

[red roost]

Yes 50 message per 4 hours

[strong kiln]

👀 Anyone else experiencing restrictions with GPT4 regarding length of messages?

🤔 Isn't it supposed to be able to handle way longer messages of around 25K words? I'm way below that threshold with the message I sent. I'm at around 20K characters
Any ideas?

I'm already paying the subscription.

[tiny moon]

imagine hitting a full 32k prompt and then the request fails 🙂

[devout schooner]

thank you

[tardy heart]

characters is not tokens

[frozen summit]

yeah I can see that happening easily

[brazen knot]

This is restricted to their API, not the chat version

[tardy heart]

that too

[loud dove]

lots of people have been saying that its being limited to around 2k words or less for prompts

[winter junco]

So i got an email saying :

You can now access GPT-4 models with 8K context via the existing OpenAI API.

How do I use it in my existing app running

model="gpt-3.5-turbo"

Tried:

model="gpt-4"

[jagged solstice]

25k words.. sounds pretty much like around 30k token to me

[tardy heart]

nice

[wind urchin]

Is it much more expensive than the 3.5 version ?

[tardy heart]

yes

[jagged solstice]

Yep

[strong kiln]

Ohh, interesting. Any way to take advantage of that?

[devout schooner]

what is a token?

[tardy heart]

0.002 vs 0.03 per 1k

[jagged solstice]

Around 15x-30x more expensive

[tiny moon]

a token is like a stoken but without the s

[red roost]

Mine says GPT-4 currently has a cap of 50 messages every 4 hours. Expect lower cap next week, as we adjust for demand. So they are going to lower that cap even more.... Please don't. Let us buy more message every 4 hours for like a 5$ up charge or something

[devout schooner]

is a token the same as a prompt?

[brazen knot]

You have to join their wait list to be able to use the API

[final forge]

Quick use it now before next week!

GPT-4 currently has a cap of 50 messages every 4 hours. Expect lower cap next week, as we adjust for demand.

[jagged solstice]

A short ver: a token is a way to measure how long a text is, usually 10 token is around 7 words or so

[loud dove]

or just provide the services they make you pay for

[wet sable]

hey guys, why we can't send images in Chat gpt 4

[charred timber]

Yeah idk why they said that, now people are going to use it even more haha

[remote sun]

right?

[wind urchin]

Would you say it is more efficient to buy chatgpt plus over the api due to the costs in api probably exceeding 20 dollars ?

[tawdry bobcat]

Before you hit todays cap. Message me for something extremely cool I discovered

[devout schooner]

interesting

[jagged solstice]

Because that is not released yet

[tardy heart]

well i can see this place is mostly a complain zone for the gpt plus, was curious to see what people are working on lol

[wet sable]

when it's gonna be released

[frozen summit]

It depends how you're using the api

[red flame]

What does it mean that there is a token charge for chatgbt 4? I thought plus paid subscription meant unlimited use of GBT4??

[final forge]

But if you're using API then you are presumably incorporating it into some product or service..

[red roost]

I am working on making my own Neuro-Sama but local and only for me....

[frozen summit]

they're working on it, there's no release date yet

[wind urchin]

Hey, can I use GPT-4 if I am not a premium user?

[jagged solstice]

Depends, but given that what youre trying to do is likely going to be against tos, no

[cloud cove]

Dear customer,
We want to inform you that OpenAI has not been impacted by the evolving situation at Silicon Valley Bank. In light of this fact we wanted to remind our customers to be vigilant of any requests for payment that appear suspicious or unusual; official communications about payments will originate from @wraith ginkgo.com or our payment processor, Stripe, and we encourage customers to reach out to us at any time with questions or to verify payment instructions.
We acknowledge the stress many are facing and extend our empathy and assistance

[charred timber]

Unfortunately not

[tardy heart]

I'm working on an automod for reddit using both 3.5 and 4

[devout schooner]

what is difference between GPT 4 AND the API version?

[jagged solstice]

It's 2 different things,

[loud dove]

the api use is pay per use, chatgptplus gives you "unlimited" (but very limited) access to gpt4

[brisk sail]

guys im facepalming so hard rn. I havent gotten api access yet bcuz i didnt put in my org id 🤦‍♂️

[tardy heart]

just a proof of concept idea

[still owl]

Send openai to orgrimmar

[tardy heart]

wont be used in production

[cloud cove]

why did i got this while i dont have svb

[jagged solstice]

Oof

[red flame]

So unlimited use with plus on Gbt4?

[loud dove]

just between using the API directly and using it in chatgpt

[brisk sail]

ikr

[stoic lodge]

o nice got gpt4 in playground

[charred timber]

50 messages per 4h and more restrictions next week

[stoic lodge]

gotta figure out the cost before i go spamming it lol

[devout schooner]

is API version better than GPT 4?

[tardy heart]

0.03 per 1k tokens @stoic lodge

[red flame]

So is it worth upgrading guys?

[terse linden]

I have been working with gpt4 for like 5 hours in a app and used exactly 50 chats so if it goes lower than that it's going to complicate things. I did the work of 2 days with this AI

[brisk sail]

now it'll probably take me forever to get it 😭

[tiny moon]

Listen up, bucko! If you want to get in on the top-secret, ultra-exclusive action, you're gonna have to bust out the Konami code like it's nobody's business! And what's waiting for you on this hush-hush page? Only the Holy Grail of CHATGPT image upload, my friend. It's so secret, we can't even mention it by name (oops I just did). So, are you ready to join the elite and unlock the ultimate in clandestine content creation? Then, start tapping those buttons and get ready to be blown away by god's magic created by non other then based OpenAI

[jagged solstice]

One of them is access via chat.openai (you'll chat directly), while the other one is for an app to use, so like a discord chatbot or so

[scarlet perch]

There's a variable message limit dependent on server capacity, but other than that, no limits.

[tardy heart]

you all still have 3.5..

4.0 didnt just replace it

[lime pawn]

Hello I bought subscription but i don't see any playground, also i cannot send in request even 3000 words, however overview says that i can send 25 000 words?????

[tardy heart]

ahh mass adoption is fun

[jagged solstice]

You can send 32k token which translate to roughly 25k words, yea

[frozen summit]

you're confusing the api with the user chat

[stoic lodge]

the regular playground gpt4, can it be sent imgs?

[charred timber]

Anyone got numbers for the amount of gpt plus users?

[open hedge]

how popular is gpt4

[devout schooner]

OHHH so API is better, but theres obv a waitlist. Is that free or also paid once granted access?

[frozen summit]

a lot

[lime pawn]

I cannot send even 3000 wrods!!

[jagged solstice]

And it's on chat.openai, not playground

[stoic lodge]

which is more limited rn?

[red flame]

Is plus subscription worth it at the mo for Gbt 4? What's people's thoughts?

[charred timber]

haha fair

[strong kiln]

What? On the chat interface? I'm getting a message saying that my prompt is too long with 20K charachters.

[jagged solstice]

Paid, and it's very expensive (15x more expensive than the current 3.5 model?

[scarlet perch]

ChatGPT Plus grants access to GPT-4 on ChatGPT. This is separate from the API that is integrated in Playground.

[frozen summit]

definitely the user chat... you only get 50 messages every 2 hours but the api has a waitlist

[jagged solstice]

What's the error message really?

[charred timber]

Try out 3.5 for a while first, if you think you can use more accurate info, then go for it

[lime pawn]

It says that request too long!!

[stoic lodge]

ah i just got into the api but dont have the user chat lol ill test it out later, ty 🙂 hope its not too expensive

[neat wedge]

It was very much so when it was at 100 GPT 4 messages every 4 hrs. At 50 messages every 4 hrs it gets less attractive. Anything lower than 30 might make me drop it.

[tardy heart]

i literally told you the price lol

[jagged solstice]

Did you count all of the messages beforehand? So all messages including the ones that the AI respond back with

[lime pawn]

Yes by wordcounter

[wind urchin]

How many tokens will I get in 20 dollars btw ?

[charred timber]

That's 50 including the replies? No way

[devout schooner]

I seee.. and can one get the same tasks done on the GPT4 plus version that they can on the API? Or is it somehow smarter?

[steel sequoia]

is there some server/bot which we can use gpt4 right now?

[tropic owl]

No.

[sterile sphinx]

this super low cap in ChatGPT Plus is KILLING ME, lol

[jagged solstice]

Basically yeah, they're the same thing

[tardy heart]

anyone who has a discord bot like I do wont make it public, api is costly.

[stoic lodge]

o im blind lol ty

[jagged solstice]

Not gonna go far I'm afraid:(

[steel sequoia]

ah i thought like it was something official from openai

[tardy heart]

no

[steel sequoia]

got it thanks

[devout schooner]

@wind urchin are you talking about the GPT plus? Or for the API version? for the $20

[jagged solstice]

With 3.5 I'm okay with doing it public for free, but for 4.. that 15x price increase do not look good for my credit card 😅

[wind urchin]

Api

[wind urchin]

got to get yoru inner entrepreneur out

[scarlet perch]

https://help.openai.com/en/articles/7127956-how-much-does-gpt-4-cost

[faint spruce]

what's killing me more is the image capabilities that haven't come yet, and idek when it will come out, the anticipation that i hate...

[snow tree]

Is the message cap going to increase again any time soon?

[somber zodiac]

how can you use or access the bing version?

[devout schooner]

@jagged solstice thank you for being so cool and open to helping.

[wind urchin]

I have been having a hard time spending even 5 dollars in 3.5 playground

[wind urchin]

So I am optimistic

[tropic owl]

Use your microsoft account

[scarlet perch]

It is variable dependent on server capacity, so we don't know.

[flint surge]

For the AI IDE thing, I think I figured out what needs to happen. The source of truth will be a git repo. When you load up your project in the browser, it makes a request to git to download the source tree. You can view individual files via a request to load that file where it just loads that one file from git on the server and sends it to you. Now, when it's a new project, the system would scan each file in the repo and summarize the file (maybe just method names and arguments to start with, as well as letting the AI read the file and generate a description of what it does). This would be stored in a datastore database against the commit and updated with every new commit for that file.
This way the AI can always have an overview of what each file does and what is available. Then when we make a request to make a change we first ask for a summary of what it will need and give specific format instructions so we can parse those instructions and feed them back to the AI. For example, it might need to see method x in class y....etc
so it can decide to provide more context for itself (but that's advanced stuff)
to start with we'd just specify what the context is for each prompt
so this would allow editing of all types of code, AND quick and easy testing since it would be always committing changes to a branch which can be easily pulled into a proper IDE for running and tweaking
then fed back

Who wants to work on this with me 😄

[tropic owl]

Select the Chat tab in bing.com

[tropic owl]

Sign up for the waitlist.

[sterile sphinx]

I'm honestly fine waiting for the multimodal capabilities, or I would be if the cap wasn't so frigging low; they need to ESCALATE

[jagged solstice]

No problem dude, I like helping others 🙂

[charred timber]

If 1k tokens is 750 words, then 20$ in tokens is 500k words? Or is my math wrong

[slate holly]

V4 is really amazing with complex coding questions now. Super happy with how it’s working even if it is a bit slow and limited for now. It will get better for sure

[wind urchin]

tldr

[jagged solstice]

You're going to be having 15x less hard time with gpt 4 to accomplish the same task

[somber zodiac]

Thank you a lot!

[tropic owl]

Right when GPT-4 becomes free GPT-5 will release.

[devout schooner]

What is the difference between the bing version and the OPENAI one?

[scarlet perch]

GPT-4 currently has a cap of 50 messages every 4 hours. Expect lower cap next week, as we adjust for demand.
-chat.openai.com

[lime pawn]

@jagged solstice Wordcounter - > 2,742 words 23,278 characters, when I sent it to gpt-4 it says "
The message you submitted was too long, please reload the conversation and submit something shorter."

[jagged solstice]

I'd say that sounds about right but I'mma pull out the calculator real quick to make sure

[tropic owl]

Bing is more temperamental.

[junior flint]

bing is total crap

[flint surge]

TLDR: Implementation details for a tool for making ChatGPT write large code projects (or any large text project really)

[novel bone]

Lower cap? damn

people really be using it a lot

[tropic owl]

The Bing version actually accesses real world data off of the internet.

[still owl]

its truly not free and only accessible through api so

[wind urchin]

Did 3.5 not have a completion rate price ? Because I don't see how going from 0.02 to 0.03 is such a drastic increase

[tropic owl]

It can scan news websites and links that you give it.

[jagged solstice]

This is for a single message, or you're already including everything above your current message?

[tropic owl]

I know, it's sarcasm.

[steel sequoia]

mind me dming you?

[flint surge]

go for it

[jagged solstice]

It's 0.002 vs 0.03

[lime pawn]

single message

[jagged solstice]

Basically 15x

[still owl]

good to see another counter voice to these sheep in chat 👍

[jagged solstice]

You'll want to count all messages above too, even the ones the AI sent back

[junior flint]

it also depends only on bing.com search engine and results vary a lot, it does not look up lot's of the websites that are actually available on google search

[snow tree]

expect lower cap? So that means they will reduce it more the 50?

[lime pawn]

Why I need to count this?

[remote sun]

lower than 50 yes

[gaunt dagger]

GPT-4 currently has a cap of 50 messages every 4 hours. Expect lower cap next week, as we adjust for demand.

[tropic owl]

It does, actually. It gives information about the site.

[devout schooner]

@jagged solstice i hope you dont mind me sending you a friend request. Always good to keep the doers and do-gooders around! I hope i can be of service to you sometime

[lunar jolt]

I really want a chat GPT app for my phone that just lets me always refer back to prior conversations. in addition, being able to interact with it via voice would be fantastic. I have a plugin that does this kind of mediocre in my browser but as a first class feature this would be huge

[jagged solstice]

It all count towards the 32k limit, as all of those messages above are included in the prompt too

[lime pawn]

I sent to chatgpt-4 about 2 messages during last day

[brisk sail]

guys from what ive seen, davinci-003 is better than gpt4 😭

[jagged solstice]

Which is what allow you to put in question like "can you elaborate more"

[jagged solstice]

Sure, I don't mind it at all

[snow tree]

nooo hell no gpt4 already answers all questions better

[lime pawn]

If I will create new chat conversation it will be from 0 coun?

[charred timber]

You have tremendous self control, my friend

[vocal summit]

what is gpt-4

[scarlet perch]

Learn more about GPT-4 here: https://openai.com/product/gpt-4

[jagged solstice]

Yea, but it'll lose all memory of previous convo, probably won't be that important to you though

[brisk sail]

gpt3 is better in personality

[slate holly]

Yeah I think it seems like a lot less websites in Bing because a lot of duplicated content is removed with AI search. I could be wrong but I think that is one of the strengths of Bing is that now it will cut through the ‘noise’ and just get what you need. I could be wrong though 🤷‍♀️

[full slate]

Hello can I add pictures and edit them

[lime pawn]

Ok i just checked this is was brand new conversation

[wind owl]

guys is there any way to try the new api with images? i have been invited

[jagged solstice]

It's not released yet, even to beta tester unfortunately:(

[brisk sail]

image func has not been released

[charred timber]

I feel like GPT4 was stylistically better, especially in mimicking some of the less known characters that 3.5 was only able to pull off with a few slang words, and that's it

[tropic owl]

If you want GPT-3.5 to be able to analyze images.

[junior flint]

also in my local language there are many websites indexes on google but on bing there are many websites that are not even indexed at all, so my search results in local language using bing is sometimes very poor compared to basic google search

[signal halo]

At me is it 100?

[jagged solstice]

It would've been interesting to see though , and I swear someone is so going to send D pick to the AI to see what it think

[scarlet perch]

API capabilities are more flexible compared to ChatGPT because you can give system instructions. e.g. "speak as if you are having a conversation over Skype", and it will respond in that way. ChatGPT is optimized for more assistant-like dialogue.

[tropic owl]

You can use the ChatGPT discord bot to give it a prompt, and also give it an image to analyze.

[slate holly]

Ah right I see what you mean

[brisk sail]

no i mean as in dav-003 is better than gpt4 when it comes to acting

[scarlet perch]

003 over the API?

[brisk sail]

for some reason dav-003 gets into char better

[brisk sail]

yes.

[junior dust]

davinci-003 — can handle more complex instructions than models built on GPT-3, according to the lab, and is measurably better at both long-form and “high-quality” writing.

[slate holly]

🤣

[scarlet perch]

Yes, the API is more flexible in that way, it's not as strictly optimized for one task as ChatGPT is

[slate holly]

That was funny but I’m sure you are right about the D pics lol

[brisk sail]

guys how cool would it be t omake a FNAF freddy with gpt3 or 4?

[tropic owl]

Try.

[jagged solstice]

Pretty cool, id say