gpt-models

[quick wing]

So just saying, if you don't like the current situation, make sure to vote with your wallet

[buoyant canopy]

The principles are good IMO, like when a company releases a product it has to prove it is harmless instead of the US model of "YOLO, FDA will tell us if we should stop selling it".
But all the misuse of the systems leads to paralysis.
I recently saw a report on how much paper some company had to file in order to build a few wind turbines.
It's ridiculous. It takes like 10 years to get permission to start a project and once you can start you're lagging behind the current state of the art. But you have to do it the way it was approved because if you want use the state-of-the-art method you need to start a new approval process which will last for another 10 years.

[tardy heart]

sounds good to me, this stuff is likely pretty expensive to host tbh

i use the API, if you think this is expensive try the API 😅

[worn talon]

how long did it take for u to get access to the API, and how expensive is it lmao

[quick wing]

There is a reason people pay for the subscription instead of using the API, it's for personal use, not to commercialize

[tardy heart]

0.03 vs 0.002 for GPT 3.5

[echo wasp]

I agree, the rack space gets pricy as hell. I just wish they'd pedal to the metal with it. I've got no problems paying more, but the process as they're figuring things out and increasing limits is a bit painful

[tardy heart]

per 1k tokens, and i applied within 5 min of it opening, got in yesterday

[quick wing]

And it's a consumer base either way, so i don't really see the logic still, but I guess we will see what the next cap will be soon

[buoyant canopy]

What politicians say sounds more like "we need rules to assert AI can't be used to harm people in any way and also no jobs are lost".

[worn talon]

🧍

[tardy heart]

i ran a few queries that were like 4000 earlier and i could literally watch the cents rise LOL

[cursive dove]

How many tokens does it cost to send a message?

[tardy heart]

depends on the message

[worn talon]

yup exactly

[tardy heart]

indeed, mine's being used for some fun personal projects

its more then just "chatting" at that point

the API is fairly quick fwiw

[nocturne viper]

“No jobs are lost” is silly. This tech is going to create 1000s of entirely new disciplines we can’t even conceive of yet. Playing Luddite is only going to hurt workers.

[quick wing]

Hope with time the situation improves, a lot of Reddit is coping with the fact that it's a beta anyway and not the product you were buying, but I'm fairly sure the EU would argue against this logic as the first deal i subscribed for was the 100 queries per 4 hours, and so far the first product I bought has been limited while im still paying the same deal, I see a potential lawsuit coming

[lapis bane]

25 messages for every 3 hours now FKKKKKKKK

[lime valve]

How is your project like? Will be cool to connect a NPC in game to Chatgpt

[limpid scarab]

yeah big rip

[wind urchin]

Next time OpenAI don’t false advertise(if ur ever going to release something, make it real, like about that image recognition, people got confused and bought it to see if GPT-4 was actually multimodal, in the end it wasn’t, it wasn’t even complete to be that, why’s that? Why didn’t you guys just release it in a later time? It should have been better if you guys delayed it, not to misinformation about it.
(I know GPT-4 is supposed to be multimodal in its complete state so I don’t want you to say that GPT-4 will have image recognition or something, I know it would.)

[jade cape]

I wish they would just make it so it had a 1 minute cooldown or something

[worn talon]

I mean at least the 3.5 is god speed, but seriously in 1 week gpt-4 is gonna be 1 message per day

[quick wing]

True, a lot of false advertising going on in regards to the image recognition too

[jade cape]

that would be soooo easy. like just have a cooldown

[quick wing]

Not a good look for OpenAI as such and it's subscribers

[tardy heart]

I'm doing something as a fun experiment to see if it can be done.

I'm building a reddit bot to scan a subreddit I run for rule breaking posts, using multiple queries to determine the post type and a ruleset against that post, and if it breaks the rule to report it.

I don't plan on using it live, but its more of a "what could be" question

i got it working already

[worn talon]

Yeah true, for me 90% of the name it was just making up stuff

[blazing shard]

GPT-4 currently has a cap of 25 messages every 3 hours. Expect lower cap next week, as we adjust for demand.
😭

[quick wing]

Well i'll just wait till the 14th day of the refund period to see what happens to the prompt amount, if it only goes lower then i see no reason not to take my money back and others should follow to protect their consumer rights.

[wind urchin]

Costumers are always right.

[blazing shard]

give me ap pleaseeee

[quick wing]

Ah, sorry US fellows

[tardy heart]

you realize 3.5 works amazing still right

[wind urchin]

Fun fact. LLaMA is more optimized than GPT-3.5 and even Bing AI.

[tardy heart]

nice and quick on plus

[wind urchin]

Talking about the leaked one

[blazing shard]

yea just needs a little bit more talking too but yea it works fine

[wind urchin]

Optimization = less compute power required

[quick wing]

Yeah of course it does, but when it comes to complex code solutions GPT 4 still outperforms it by a mile, and i did buy a product that included good access to GPT 4, not to be limited regularily after i've already paid up.

[buoyant canopy]

Just a little anecdote on Germany: We pioneered wind and solar power tech. Really high subsidies created world leaders here. Hundreds of thousands of jobs in that area. But we f'ed up. Hundreds of thousands of jobs lost again. Last German company building wind turbines closed recently. Lobby interests from fossil energy companies were too strong and our politicians just killed the whole renewable stuff. China also heavily subsidized and we couldn't compete their prices.
But on the other hand we keep a Zoo of about 20k people working in coal mines. Also heavily subsidized because mining that stuff here is expensive. But these jobs are sacred. Nobody dares touching them.
It's ridiculous.

[thorny valley]

Push ChatGPT into every area of life before we die.

[quick wing]

Can't really do that with 1 prompt per 4 hours

[wind urchin]

Make GPT-6 Codename: Phoenix comes true!

[thorny valley]

why Phoenix? 😦

[wind urchin]

Phoenixes are cool

Yeah but still, if GPT-4 had that optimization, it wouldn’t change the cap and way way faster respond speeds btw

[plucky willow]

"GPT-4 currently has a cap of 25 messages every 3 hours. Expect lower cap next week, as we adjust for demand." so the feedback loop will be killed sooner or later.. with such limit...

[cerulean robin]

I suspect it’ll be 10-15 messages next week

[wind urchin]

Imagine an AI run on a local pc, like a MacBook / PC / Laptops(stop, bad)

[quick wing]

If you dont like the lower cap, unsubscribe, vote with your wallet.

Not like they're gonna change it if you keep paying up.

[wind urchin]

and Microsoft has a lot of sub servers/servers

[plucky willow]

basically killing GPT-4...

[wind urchin]

It will definitely come in handy that optimization

It would be faster than Bing by a lot

even tho Bing is very fast

[quick wing]

At least that'll be an extra 20 bucks to spend on the Steam sale...

[wind urchin]

On par with GPT 3.5 if not, faster.

[cerulean robin]

They mentioned adding another payed tier if demand was heavy, so they should do that. 25 messages isn’t enough for me personally.

[heavy iron]

Agreed $20 and keep getting less and less, not cool.

[plucky willow]

is there possibility to run gpt-4 on self-hosted hardware somehow? what about training data?

[quick wing]

It's closed source, contrary to the name OpenAI, a bit ironic.

[wind urchin]

not possible, if mainly talking about without internet

[plucky willow]

are there any information about size of training data? Petabytes etc.?

[heavy iron]

But you can get the API and setup your own interface.

[plucky willow]

too expensive, esp. for complex analysis tho :/

[lime valve]

My feeling is that the new bing is a bit stupid and has too many constraints...

[wind urchin]

I’m going to sleep, please note that GPT-4 is going to be vastly improved, GPT-3 was a hazard when it released, everything that is incomplete is bad, when it’s complete it could be good, this situation may vary though.

[wind urchin]

Yeah it uses the internet, that’s why.

[thorny valley]

We need to all master on how to use Office 365 copilot to the fullest..

I'm tired of the constant verbal-only talking where people sometimes don't know what you are talking about 😦

[wind urchin]

I made comparison here

[buoyant canopy]

Seems to me it is finetuned to a "ask -> web search -> generate answer from search results" loop.

[plucky willow]

somehow sometimes it gets wrong data from the internet without having "one source of truth" -

[potent cove]

gpt 4!!!

[heavy iron]

Yes depending on your need it will get pricey@

[wind urchin]

I doubt if it even used GPT-4, it’s way dumber than GPT-4

[cerulean robin]

Microsoft did put a lot of moderation in place, maybe that’s it

[wind urchin]

More like modifications

[ripe hinge]

"Available even when demand is high" chatgpt plus. So not anymore?

[wind urchin]

They probably modified the hell out of GPT-4

[cerulean robin]

Yeah but it’s so heavily restricted now it’s a joke

[cobalt kestrel]

gpt4 is just a... bonus

[plucky willow]

general question, has anyone compared gpt 3.5 vs gpt 4 for general day to day usage for analysis of complex issues within datacenters? if there is such difference - from my POV if you are able to craft precious gpt 3.5 prompt, then it may be comparable to gpt 4 where you dont need such precision

[wind urchin]

True though

[fading walrus]

I'm sure it's been pointed out but the image recognition in 4 is going to be an absolute accessibility game changer when it matures. My uncle is legally blind and I've been thinking all week about how this could enrich the life of someone like him

[deep jetty]

GPT-4 currently has a cap of 25 messages every 3 hours. Expect lower cap next week, as we adjust for demand.

its toooo much no !!

[cerulean kite]

Bing uses GPT-4

[wind urchin]

Must be heavily modified version of GPT-4 then

or it’s a mixed one

[south pulsar]

yikes 25 per three hours is kind of a bummer imo

[wind urchin]

because I also saw that bing said OpenAI GPT-3 before

[deep jetty]

jesus comeon openai next time what 10 message per day

[ripe hinge]

Anything similar to ChatGPT 4.0 for python programming?

[plucky willow]

i would say feedback killer at this moment, prob. we need to wait for gpt4-turbo or something?

[south pulsar]

yeaaah that's fair

[toxic copper]

Anyone been able to use the Python package after getting accepted from the waitlist? I've tried everything and still getting openai.error.InvalidRequestError: The model: gpt-4 does not exist
gpt-3.5 works fine though

[plucky willow]

those are good news

[worn talon]

How do I try to refund the plus

[south pulsar]

ahhhh nice, thank you so much!

[wind urchin]

Can’t wait for the next subscription become 200$ a month

[silent hawk]

I hope not

[worn talon]

nvm bruh you cant in US 💀

[silent hawk]

Even I can't afford that

[grave hatch]

25 what 💀

[silent hawk]

It changes. I think if you use it at night, the limit is much higher.

[cerulean robin]

I have a suspicion that this has to do with the release of chatgpt plus in India, I bet the demand is through the roof now, which makes me question why they’d even do that in the first place.

[wind urchin]

I wonder what would the next subscription name be

[silent hawk]

Wait, it is in India now? I can already imagine the "Microsoft" and "Amazon" call centers subscribing

[wind urchin]

ChatGPT Demand?

[plucky willow]

hopefully, fingers crossed, the plus price for still-subscribed ppl will remain the same... but for new subscriptions i would expect that the price will rise somehow, or it will be combination of ads within the replies + lower price or something... and then pro subscriptions may be introduced, eg price 50 bucks or more.. at the end all we can do is to speculate, prob. the final price model is still being crafted

[cobalt kestrel]

Chatgpt 24/7

[burnt harness]

everyone's talking about the response limits; it is clear to me why they are hitting capacity. This thing is amazing. Not only verbal acuity off the charts, but meaningfully assessing and taking lines of thought when 3.5 just fell back on NPC like repetition of its training look up.

[forest folio]

$20 a month was already significant... but this... oof.

[mossy cedar]

Lmaoooo 25 every 3 hours what- I'm considering unsubscribing

[echo wasp]

I agree, 20$ was already steep to begin with.

[ripe hinge]

We need competition fast

[plucky willow]

then what will be the "Plus advantage" when compared to free (except sometime the speed of gpt 3.5)

[silent hawk]

tbh, this is better than early chatgpt where the site just conked out at random

[forest folio]

Like it was worth the $20 a month to me... but that was before the limit changes.

[silent hawk]

better to have an actual cap than fail entirely

[vernal pagoda]

25 messages every 3 hours..... im unsubscribing thats it

[wind urchin]

I think they have to make the new subscription and whoever previously purchased plus, will have the new subscription, because that would be highly disadvantageous, due to the fact that you need to pay again just to get proper access to GPT-4 while previous experience was one of the worst experiences ever.

[somber egret]

Indian people are copy writing hella stuff rn on Fiverr. 😂

[mossy cedar]

So the plus version will get what, 10 requests every 4 hours? It's ridiculous.

[vernal pagoda]

I have access to the api now anyways might as well just use that

[somber egret]

How much is the API?

[bronze cairn]

gpt 4 is amazing, i just wish they were a bit more transparent with how they would limit paid users over the coming days

[burnt harness]

they don't know; there is a new roll out every week and demand spikes again on each one.

[wind urchin]

True

[silent hawk]

they probably don't know

[wind urchin]

yes it's all in the prompts

[modern lintel]

Is there a better way to use ChatGPT for coding than the website?

[wind urchin]

copilot and an IDE

[wind urchin]

You mean APIs?

[burnt harness]

re safe, my experience has been that everytime -4 hits any of its late stage human safety prompts, it is transparent that is what happened and its apparent IQ drops about 30 points. This thing is way smarter than the conventional thinking people trying to tie it down.

[hallow peak]

how to find tune that monster

fine*

[modern lintel]

What do I need to do?

[plucky willow]

since i am curious, one question... what do you think: is OpenAI going to combine feedback / tuning between gpt 3.5 and 4.0, or are those training sets, tunables, completely separated between those two models?

[burnt edge]

kinda takes the 'chat' out of chatgpt with 25 every 3 hrs which is like 8 msgs per hour

[wind urchin]

what IDE do you use? its available for vscode

It’s crazy that ChatGPT is taking a lot of time over some string(that’s a joke so don’t take it seriously)

[echo wasp]

Not including regenerating replies too

not ideal for creative writting

[plucky willow]

i am on the same line.. fingers crossed that the resources for gpt4 may improve in upcoming weeks...

[wind urchin]

They need to add that continue button tbh

[feral kiln]

Is there actually a message cap when we use chat gpt??

[wind urchin]

yes

[cerulean robin]

Honestly going back to 3.5 just seems like a huge downgrade now

[burnt harness]

dog.big - I think 3.5 is legacy at this point. -4 has actually been done since last August,they were just testing it and safety-ifying it. I doubt 3.5 optimization will be much of a "thing" for long.

[feral kiln]

Does subscribing change that?

[tired umbra]

hope they will add 2 weeks minimum on our subscription :d

[feral kiln]

Like paying the 20 a month?

[hallow peak]

the german guy ruined the gpt-4 release actually. they were not ready for publish it yet

[charred timber]

25 every 3 hours?????

[velvet crane]

3.5 TURBO is still fantastic for most things.

[magic stag]

Wow, so very disappointed with what they did. If it continues like this, I won't subscribe anymore. If they want to charge more to use gpt 4 chat, I won't pay more

[burnt harness]

xHyper - 3.5 is unlimited ATM, only -4 is rate limited

[plucky willow]

nevertheless after releasing 3.5 for public it has improved a lot, so there is some ongoing feedback loop in place still for "legacy 3.5"

[charred timber]

fr bro I'm getting a refund

[mossy cedar]

Yeah, but for some professional use cases gpt 4 is way better

[silent hawk]

large language models are psychopaths

[magic stag]

Is there a way to ask for a refund?

[plucky willow]

could you elaborate further pls? eg. use case ?

[mossy cedar]

I just hope they plan to get this cap a bit higher because it's a shame rn

[burnt harness]

dog.big- true, but I think that is temporary while they can't handle even premium -4 traffic. Once they can, I like 3.5 usage will drop sharply.

[velvet crane]

For which uses are you seeing a significant decrease in performance?

[wind urchin]

4 is so much better at writing emails my god

[silent hawk]

LLaMA 65B is approximately GPT 3.75

[hallow peak]

coding

[cerulean robin]

In translation tasks, gpt 3.5 falls on its face pretty hard. Especially with things like colloquial Finnish. Gpt 4 however is fantastic

[burnt harness]

dog.big - read a PDF and semantically understand it; predict code; explain and comment code - -4 crushes these and 3.5 wheezes on them.

[wind urchin]

>If you live in the EU, UK, or Turkey, you’re eligible for a refund if you cancel your subscription within 14 days of purchase. To request a refund, contact
lol i guess youre out of luck if you live anywhere else

[velvet crane]

Oh... 3.5 works perfectly for me in coding applications

[long pond]

Yeah GPT-4 performs very well even on poetry in different languages

[velvet crane]

Davinci seems to work better for translations in my experience for whatever reason.

[hallow peak]

4 is 10x better on coding trust me.

[wind urchin]

What is davinci based on

[plucky willow]

@cerulean robin ahh yeah, in those usecase, where it needs to "understand" the text, gpt 3.5 may not be sufficient

[plucky willow]

have you got some example pls?

[velvet crane]

GPT3 Davinci

[feral kiln]

So I’m a bit confused. Chat gpt 4 is avilable for instant access if we subscribe to chatgpt plus

Chat gpt 4 has a 25 message limit
But chatgpt 3 has no message limit and will be available at all times even in high demand?

[burnt harness]

davinci is 3

[charred timber]

Yeah my hope was that it would be over in a few weeks, but since they said next week I didn't expect it they'd change the same day like haha

[velvet crane]

I'm not going to trust you, that's the point of asking for use cases. I code 12 ours a day and 3.5 works the same for me.

[cerulean robin]

I guess the good news is that via Poe app, you can use their 7 day free trial and access gpt 4, 300 messages per month. Not bad

[wind urchin]

Keep lowering the CAP -_-

are they paying you to advertise POE

[cerulean robin]

No there’s a ton of people suggesting it

[burnt harness]

yes, 3.5 and 3.5 turbo are unlimited and free. They can take any traffic thrown at them; they've been ramping Azure with them since the endof November andthe 3.5 turbo is speed optimized to boot. -4 is a much bigger model, thus slower, also bigger prompt size, same, hasn't been ramped up yet except for use cases like Bing and Office 365 for obvious commercial reasons and who funded this stuff. So -4 is currently premium and rate limited,outside those products.

[wind urchin]

How are they able to handle gpt 3.5 easily but struggle with 4 ?

[burnt harness]

see previous post above

[mossy cedar]

I'm curious how much it would be.

[burnt harness]

JFx - exactly. Right now.

[charred timber]

Tbh I think it's a more complicated model, they'd use a lot more resources for it. Still a shame they wouldn't be able to provide a reliable service given that they'd probably expect it making this much traction, this is all really weird

[silent hawk]

You mean there's a ton of shills

[plucky willow]

"their plan is to optimize in the coming months" .. webui devs should add this after the text about the cap under the prompt... to lower the anger

[cerulean robin]

No it’s a legit suggestion since access to it is limited..

[wind urchin]

Wait did they reduce the 50 cap ?

[lime hinge]

how do i refund my sub i got it yesterday

[plucky willow]

now 25 per 3 hours

[lime hinge]

are they trolling

[burnt harness]

sam - they halfed the 50 but also dropped the time window to "per 3 hours"

[minor island]

hey, do you guys have access to gpt-4-32k ?

[burnt harness]

moving target - will change every week or two for the next little bit...

[vale plank]

if it goes down to 10 gpt plus is just a watse

[wind urchin]

Cap is ridiculous

[thick hinge]

Check in help.openai.com

[plucky willow]

hmm wondering if this is intended... to move ppl to use bing chat for gpt4

[fringe pecan]

any ETA for a gpt-4 api?

[minor island]

you have to be on waitlist, i have it

[fringe pecan]

ah

[scarlet perch]

Not at the moment no, have you joined the waitlist?

[wind urchin]

just keep in mind you guys paid Plus for unfettered access to chatGPT not gpt4, which is just a bonus currently

[burnt harness]

one co-worker of mine has API access to -4 as of yesterday, others waitlisted

[fringe pecan]

i have not, link?

[forest folio]

Soon we will have GPT-4 currently has a cap of 1 messages every 1 year. Expect lower cap next week, as we adjust for demand.

[lime hinge]

they legit scammed us

[scarlet perch]

https://openai.com/waitlist/gpt-4-api

[wind urchin]

you scammed yourself

[fringe pecan]

thanks

[cerulean robin]

Expect lower cap next year*

[wheat aspen]

scam

[wind urchin]

take some responsibility for your actions

[plucky willow]

OpenAI should pinpoint this somehow... i can now see that the communication is the struggle, not the gpt4 itself...

[slate gulch]

Guys how do i use gpt4??

[cloud dust]

are there plans for gpt-4 to go public? I might go for the purchase if not

[burnt harness]

Yudie - get access or what should you try - which question?

[cerulean robin]

Eventually, not any time soon

[lime hinge]

what a joke

[wheat aspen]

scam

[minor island]

do they have access to -4-32k ?

[vital portal]

why the cap is 25 now, its getting lower and lower

[cerulean robin]

They are definitely not going public if they can’t handle the current load..

[burnt harness]

peac - fair Q, donno, will ask him (might not get an answer until Monday lol)

[wheat aspen]

nice scam

[forest folio]

anyone buying puts on microsoft next week because of this? XD.

[silent hawk]

1 message/day

[plucky willow]

it is currently available for plus users, with 25 messages cap per 3 hours (current status), so my bets are, until the resources are for gpt4, it will be available thru plus... or zou can try bing ai, it is using gpt4

[lime hinge]

dont worry it will be lower next week XDDD

[burnt harness]

Slifer - because more people are signing up and hammering them with programmatic tasks, obviously.

[scarlet perch]

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

[sly wolf]

Is the cap/limit only with ChatGPT or with API as well?

[cerulean robin]

Ha yeah, they need to handle the current demand before they start doing anything else

[plucky willow]

my bets are that the cap is semi-automated? higher load - worse cap? and yeah, they should remove "next week" wording....

[wheat aspen]

nice scam

[burnt harness]

S4 - only with GPT-4, which isn't chatGPT3.5.

[scarlet perch]

I believe the variable message limit only applies to GPT-4 on chat.openai.com (not GPT-3.5)

[shut yew]

Hey, anyone here get the invite to API and go to playground but not find GPT4 available to you? Am I doing something wrong? Is there something you're supposed to do first? The email just sent me to OpenAI.

[sly wolf]

Alright, i have access to GPT-4 API and did not hit any limit in the playground.

[devout night]

If there's a new tier level , what could the potential price be like

[scarlet perch]

ChatGPT Plus subscribers will get GPT-4 access on chat.openai.com with a usage cap. We will adjust the exact usage cap depending on demand and system performance in practice, but we expect to be severely capacity constrained

This message has always been available on https://openai.com/research/gpt-4, for those wondering.

[sly wolf]

You just have to wait, if you have registred yourself. I got mine yesterday

[minor island]

do you have access to gpt-4-32k ?

[mystic girder]

yeahh, not worth the subscription at this point in time, better resub once they increase it back to at least 100/4h, with 25 you can barely even solve a single problem before you are on cooldown

[lime hinge]

lower than 25 is gonna be 5 messages an hour 💀

[strange lagoon]

For GPT-4, are you able to continue chats with GPT with V4?

[shut yew]

I got the email. So you're saying it just takes time for the playground to catch up?

[strange lagoon]

Or do you have to start new chats with it.

[sly wolf]

No it was something else

[minor island]

i checked 30mins after receiving the email and had gpt-4 in the playground models

[charred timber]

Bro OpenAI needs to get a grip not even lying

[dense fable]

?

[mystic girder]

sucks that we cant even see how many messages we've used up out of the 25

[livid idol]

Can we use GPT for testing purpose

[sly wolf]

Not sure about that, received the email and the model: GPT-4 was available in the playgroud. Have you selected chat?

[charred timber]

With all these restrictions bro

[forest folio]

Overall.. some clarification from OpenAI would be great... Like is the planned to be a permanent problem? That you only try to solve by "higher subscriptions"? Or is this just temporary... Cause if its just temporary for a week or two.. I can live with that, its expected... Just sounds like its bad, and it will be worse next week, sucks to be me.

[scarlet perch]

What sort of testing purposes?

[livid idol]

Testing for web applications

[plucky willow]

you can ask in the chat, spending additional messages :/

[shut yew]

Yeah, all the same models are there. No GPT 4

[wet arrow]

wont include other convos tho

[scarlet perch]

I don't see why not, seems like a great use case 🙂
I assume you mean developing web applications.

[mystic girder]

each week they will reduce the cap until its down to 1 message per hour lmao

[chilly dragon]

Make a longe deep meaning poem using only the letters : A, D, E, O, P, Q, R, S, T, U, V, W and Y

[charred timber]

@scarlet perch any updates on further restrictions? Just curious

[sly wolf]

Weird, something is scuffed probably then. I would suggest contacting with the support in this case.

[livid idol]

I mean automation testing and all..

[plucky willow]

on same line there, they should be more communicative about the "cap" .. thats the feedback for them

[wind urchin]

I just know the open ai team is laughing behind their screens knowing that they scammed everyone who bought plus and then limited gpt 4

[burnt harness]

New chats with the new model

[lime hinge]

^

[chilly dragon]

GPT 4 is just another slow copy of GPT-3.5

[scarlet perch]

I don't have that information (unaffiliated), but I'm guessing as further optimizations are made limits will ease. We'll have to wait and see.

[charred timber]

Not really

[lime hinge]

im glad thousands of us just paid $20 for a product that wont work next week

[wind urchin]

Right

[strange lagoon]

Unfortunate. I might have been tempted if I got to keep my old chats.

[dense fable]

Y'all are overreacting, my dudes.

[charred timber]

Thanks, this was my prediction initially, we'll wait and see I guess

[mystic girder]

it is genuinely much better in some use cases

[burnt harness]

you can keep them, and you can copy-paste them, but they don't seemlessly transfer

[gray talon]

You paid $20 for unfettered access to ChatGPT 3.5. You got access to GPT-4 as a bonus.

[chilly dragon]

Like?

[burnt harness]

you can have chats with multiple models in your chat history no problem.

[lime hinge]

no they sold the sub for the gpt 4 in their exbo actually

[proud sundial]

confused...doesn't lower cap mean they won't cap as much?

[wanton mirage]

How is the performance and overall accuracy between bing chat, ChatGPT Plus (GPT4) and the playground GPT4 ?

[lime hinge]

no one cares about the 3.5 who pays money

[mystic girder]

creative writing, for example

[scarlet perch]

I'm a volunteer here, unaffiliated with OpenAI.

[dense fable]

That part.

[strange lagoon]

Yeah, that kinda sucks, I wish there was a way to download and upload chat history

[silent hawk]

You did back when all the moochers slowed ChatGPT to a crawl

[chilly dragon]

It's not that creative in my poems

[wind urchin]

It's vague

Probably vague on purpose

[lime hinge]

how is paying $20 for nothing being a moocher

[plucky willow]

i think that bing ai may help in these

[unkempt plank]

What the actual hell? Now its 25 messages per 3 hours?

[mystic girder]

it is only as creative as your prompts, if you give it something generic the result will be generic

[silent hawk]

It was a problem back when plus came out, where the free users slowed it down. Now not much of an issue.

[wanton mirage]

I am just wondering because I have been experiencing GPT4 being better in math compared to bing chat

[proud sundial]

can go either way it seems lol

[wind urchin]

yep

[silent hawk]

I wish GPT could call users basic. It would be funny.

[unkempt plank]

And they have the gits to tell us "expect lower cap next week, as we adjust for demand"

[mystic girder]

enjoy while it lasts it will be 10 messages tomorrow

[burnt harness]

Agreed I want that feature too. But that's how it works right now.

[dense fable]

I think of that $20 as a small donation to help the company continue to develop something I really, really, really enjoy! I've damn near become a god on every dating site I've joined, thanks to CGPT.

[minor island]

@scarlet perch do you have any info on gpt-4-32k availability ?

[charred timber]

lmao 10 5 2.5 1.25 etc

[unkempt plank]

Bruh

[chilly dragon]

Fr but with huge mistakes and Non-compliance with the input

[forest folio]

At least they warned us I guess.. XD

[wind urchin]

Join the waitlist to send messages

[strange lagoon]

Yeah, If they let ya download the chat history and upload it back into GPT4 so we could keep things like ongoing stories that'd be awesome. That's what I use it for.

[granite thorn]

I tried to upgrade but your system wont let me. When I registered with you I gave you my current number but Stripe for the email for the account is connected to a Stripe account ive not used for years and dont have the phone number for. Its a really weird system. I bet you are losing a fortune because of it.

[unkempt plank]

Wait it says "expect lower cap nezt week"
Does that mean itll be even lower like 5

[lime hinge]

probably

[wheat aspen]

tomorrow there will be 12 messages for 3 hours

[jolly cloud]

You've reached the GPT-4 cap that's currently 25 messages every 3 hours, which gives all Plus users a chance to try the model.

can't wait for the limit to be once per year 💀

[unkempt plank]

I want a refund

[lime hinge]

same

[unkempt plank]

This is so stupid

[charred timber]

No I think this was left out from the previous message, I don't think they meant that really

[unkempt plank]

Can we all write an email in to them

[mystic girder]

GPT4 seems to have a pretty big context size on ChatGPT, I can paste entire chapters of my book and ask for improvements and it will remember the entire thing

[wind urchin]

Open ai has to buy better servers before they become another nortel networks

[chilly dragon]

Waiting for bing to use GPT-4

[unkempt plank]

Lets just write an email in and complain

[charred timber]

They already do.

[dense fable]

Give me your account and I'll paypal you $20. ^_^ Then I get 50 posts per hour. WIN.

[gray talon]

I don't see that. Can you link it to me? https://openai.com/blog/chatgpt-plus

[lime hinge]

wont do anything they already scammed our money

[wanton mirage]

bing is already using GPT4

[unkempt plank]

Lmao

[chilly dragon]

Nah its myth

[scarlet perch]

Just like you, I only have what's on the research page

gpt-4 has a context length of 8,192 tokens. We are also providing limited access to our 32,768–context (about 50 pages of text) version, gpt-4-32k, which will also be updated automatically over time (current version gpt-4-32k-0314, also supported until June 14). Pricing is $0.06 per 1K prompt tokens and $0.12 per 1k completion tokens. We are still improving model quality for long context and would love feedback on how it performs for your use-case. We are processing requests for the 8K and 32K engines at different rates based on capacity, so you may receive access to them at different times.

[mystic girder]

bing is really bad though, it always ends my chats saying it doesn't want to talk anymore or something, its no fun

[chilly dragon]

Didn't see it, maybe only on paper

[minor island]

i think chatgpt gpt-4 uses the 8,192 tokens context

[charred timber]

I don't think so-their outputs are consistent but imo worse than gpt4 because of their bing sources, otherwise I don't see a reason to not believe that claim

[strange lagoon]

Nah, I think some of ya are getting too caught up on V4. If you got Plus for being allowed to test v4, you brought it for the wrong reasons. GPT-Plus is just meant for access to better response times and prompt speeds.

[lime hinge]

no one bought the sub for faster 3.5

[chilly dragon]

Yes like a kid when you beat him in an argument

[minor island]

ok thank you, i've re-joined the waiting list asking for 32k context

[strange lagoon]

Access to V4 seems like a bonus

[wanton mirage]

Is it worth buying the plus version or not ?

[scarlet perch]

Many people did before GPT-4 was announced.

[dense fable]

I feel like I learn more when using GPT-4. I like the slower speed. I get to read the entire output as it's being written.

[lime hinge]

yes and 4 being released crashed the sub

what am i missing o.o

[jolly cloud]

that's true I did

[plucky willow]

OpenAI should pinpoint this... wondering if there on discord is someone from OpenAI that could get the feedback

[wanton mirage]

damn that sucks

[mystic girder]

if you buy it for GPT4, nope, its very restricted right now, wait until they increase the cap back up

[trim garnet]

no, I think you can blame the OpenAI advertising for that. They had pasted ALL over the place "USE GPT 4 NOW WITH CHAT GPT PLUS"

[wanton mirage]

any words on when they will ?

[unkempt plank]

God dammit, im so annoyed
I bought it for the sole purpose of using it extensively and now theyre limiting it so much

[gray talon]

Thank you for that. I read "ChatGPT Plus subscribers will get GPT-4 access on chat.openai.com with a usage cap. We will adjust the exact usage cap [blah blah]."

[lime hinge]

^^^

[wheat aspen]

scam

[mystic girder]

likely a few months down the road when they scale up GPUs

[wind urchin]

why are you using a language model for math??? why does everyone want to do this??? use wolfram mathematics

[plucky willow]

this

[silent hawk]

because they can't copy and paste a word problem into wolfram

[charred timber]

so real

[wanton mirage]

oh its more about physics than math

[chilly dragon]

I hate when bing does not adhere to the inputs and must be corrected in another 50 messages in order for it to give what you asked for

[wind urchin]

OpenAI limiting messages is such bonker behavior

[gray talon]

can Wolfram Alpha output in LaTeX?

[brittle pike]

Bing may be using GPT-4. but it is infinitely more frustrating to get use out of than ChatGPT

[stiff gorge]

Gpt 4 good for answering my doubts when studying for College entrance

The message limit kinda annoying tho
Will it be lifted in some months?

[trim garnet]

Bing is literally useless

[scarlet perch]

Keep Rule 4 in mind please, ... or making very similar posts sequentially are considered forms of spamming.

[lime hinge]

bing filters your message to something that wont help

[charred timber]

Think the formatting they use is universal enough so you could paste it without much modifications

[earnest vine]

will gpt4 ever come to free users lol

[silent hawk]

now they will complain about censorship of the truth

I bet $25 on that

[earnest vine]

maybe in 2025?

[chilly dragon]

I don't think it's using GPT-4 cuz chatGPT-3.5 gives better outputs

[lime hinge]

it wont even be here for people that pay

[earnest vine]

wym

[trim garnet]

Bing will stop a conversation after like 3 messages and then says, in an extremely passive aggressive tone "I think it's time to change the conversation 🙂 "

[mystic girder]

yeah buddy dont count on it even paying users are suffering

[lime hinge]

we're getting messages <25 / 4 hours next week

[wind urchin]

Of course not. Be for real

[trim garnet]

literally

[earnest vine]

I'll just stick with gpt 3lmao

[wind urchin]

GPT-4 currently has a cap of 10 messages every 6 hours. Expect lower cap next week, as we adjust for demand.

[wanton mirage]

how ?

[lime hinge]

have you tried it

[wanton mirage]

no not enough

[chilly dragon]

Cuz its sucks

[wanton mirage]

would love to hear more about that

[mystic girder]

it will straight up refuse to talk to you a lot of the time

[trim garnet]

Have you tried using it for anything mildly complex that isn't just as complicated as a google search? It will refuse to help you and then ask to switch topics.

[wanton mirage]

damn

[earnest vine]

yeah I uses bing for 4 minutes then stopped

[brittle pike]

It excels at very cut and dry searches and comparisons, but ask of it anything even remotely complex and it tips over

[analog linden]

Man, the capacity keeps getting lower and lower

[chilly dragon]

use this:

Make a longe deep meaning poem using only the letters : A, D, E, O, P, Q, R, S, T, U, V, W and Y

[bitter wagon]

by lower cap do you mean less messages per hour or more

[cobalt umbra]

Tbh - probably in like, 2 to 3 years lmao

[trim garnet]

you're better off just using the search engine for anything that Bing will do okay at

[scarlet perch]

If you're looking for more flexible GPT-4 usage I do recommend joining the API waitlist
https://openai.com/waitlist/gpt-4-api

[silent hawk]

It was neutered because people got it to say Bad Things™️

[trim garnet]

yeah, they went into hyper damage control mode and gave it a lobotomy

[wind urchin]

pov: chatgpt before January 9th

[restive chasm]

so the value of our money has gone down right?

its not worth 100 messages anymore its worth 25

[lime hinge]

immensely

[burnt harness]

Wolfram Alpha returns XML for things like fractional forms and integral signs and the like. Doesn't natively support choosing LaTeX as an output form that I know of.

[lime hinge]

less next week

[silent hawk]

I mean, you originally paid $20 for 3.5 only

[mystic girder]

if you care about GPT-4, yeah

[restive chasm]

thats the only reason why I even bothered to become a plus sub

thanks Microsoft

[chilly dragon]

Already did in 15/3 and waiting for replies

[mystic girder]

yeah I was using 3.5 just fine as a free user, this sub is useless to me without GPT-4

[burnt harness]

Not my experience - at all.

[restive chasm]

Exactly

[kindred girder]

is GPT4 still down

[scarlet perch]

I don't believe it's down

[forest folio]

I'm curious why they can't just spin up more servers in the cloud.

[mystic girder]

you probably ran out of messages

[trim garnet]

I feel like this is an unfair statement. OpenAI was putting in the banners TRY GPT 4 NOW WITH CHAT GPT PLUS. Very sneaky advertising since clearly they were ill prepared for the influx

it feels like we just got robbed

[kindred girder]

anyone tried to jailbreak it successfully?

[steady tiger]

inb4 subscriber tiers come out, pay more for no limits for GPT-4 🙃

[mystic girder]

their hardware is sorta "fixed" right now so they cant scale up

[trim garnet]

^

[wind urchin]

I don't know why they started with such a high cap. It really does not make sense. Why didn't they just start with a low cap and increase it

[steady tiger]

I'm sad to say I'd probably pay regardless hah

[vague grove]

mine isn't working rn, it thinks i haven't upgraded to pro

[scarlet perch]

I imagine they are doing that, and also spinning up a ton of servers already

[cobalt umbra]

They did suggest the idea on the online form to apply for more GPT4 usage

[steady tiger]

Try different browser / clear cache

[trim garnet]

what is this "increase" you speak of? i know not what that word means

[kindred girder]

there should be a 20 message free trial option for GPT4 tbh. I believe i heard the number of paying users has not gone above 500,000 users of the 100 million users for free version

[mystic girder]

man if the 20$ tier gets at least 100 messages back at some point I'll be happy, but no way i'll pay 50$ sub or something

[languid mural]

I ran out so fast

[wind urchin]

I would not be frustrated if they had just started with 25 every 3 hours and have it stay that way

[chilly dragon]

I Think GPT 3 It will also will become paid when user information and contributions to its development are no longer useful

[trim garnet]

yeah they conditioned us with 100

[steady tiger]

Yeah, the 100 message limit was nice while it lasted.

[lime hinge]

💀

[wind urchin]

And good and well i used all 100 messages

[kindred girder]

I believe OPENAI realized the type of prompts users outside their team is using it form costs too much for them so they reduced it

[trim garnet]

☠️

[rustic charm]

I used my 25 messages so fast and already out

[steady tiger]

Lol, yeah I hit the limit a few times before it went away.

[vague grove]

they reduced what they offered by over one order of magnitude, that's a good way of making all early adopters hater you

[kindred girder]

25 messages is not worth it for 25. thats almost 80 cents per message

[scarlet hinge]

bruhhh 25/3 hours is barely usable lol

[trim garnet]

this

[wind urchin]

Right, like they are just digging a hole atp

[mystic girder]

you are paying monthly

[trim garnet]

and they say to EXPECT LOWER NEXT WEEK 😄

[lime hinge]

quartered actually and it's going lower

[chilly dragon]

Waiting for the 1/1 year

[wind urchin]

Sidest eye to OpenAI

[scarlet hinge]

there's a lot more than 3 hours in a month bro

[wind urchin]

Anyone here got alpaca working

[wind urchin]

yes

[mystic girder]

probably still cheaper to sub than pay for the API but not that much cheaper

[kindred girder]

is Alpaca free?

[wind urchin]

Has anyone gotten GPT 4 to pop up in their Playground

[trim garnet]

you're expecting too much

[wind urchin]

How does it perform

[vague grove]

25/3 = 8.3 repeating which is > 10 which is 1 order of magnitude dif from 100. or did i misunderstand the 3 hour thing?

[wanton mirage]

so just to clarify, bings GPT4 is really bad, Playground GPT4 costs a a ton, and ChatGPT4 is having a limit

[scarlet perch]

$20 for ~6000 messages in one month = $0.003 per message

[scarlet hinge]

they probably did not anticipate this much demand

[wind urchin]

not good compared to openAI models, but you can do more with it

look up george hotz youtube streams playing with it

And most of those messages are just gonna be "sorry i made a mistake you are right (proceeds to not give the correct version)"

[wind urchin]

Which version is it the 7b parameters or 13 or 65

[kindred girder]

is there even a playground gpt4?

[wanton mirage]

yes

[wind urchin]

i use 7b

[scarlet hinge]

i'd assume they got a lot more subscriptions because of the plus restriction on GPT-4

[river stone]

did they at least remove the cap for ChatGPT 3.5?

[trim garnet]

to defend gpt 4 a little, I think GPT 4 was actually significantly better than gpt 3.5

[wind urchin]

For plus yes it always has been

[trim garnet]

at least it was giving me better results

[vague grove]

then they are fools i dont respect lol

[wind urchin]

Is 65k version usable ?

[scarlet hinge]

yea it definitely shows a lot better understanding of concepts

[rustic charm]

GPT-4 is great, hopefully the limit increases soon

[lime hinge]

sadly it's going down next week

[river stone]

it has not, I have Plus and I still get the "Too many messages in the last hour" message. And then I literally have to wait one whole hour. And again, I have Plus.

[faint spruce]

they plan on shortening it next week lol

[wind urchin]

no experience with it. just used llama @wind urchin

[stoic lodge]

in the gpt4 api, how long can the left "system" box be? The one where you put "you are a helpful assistant", can i get super in depth or is simple better? When I ask it a question, does that textbox add to the tokens used? ty 🙂

[wind urchin]

Oh you used the website ? Or ran it locally

[scarlet perch]

My point was just to quickly disprove the 80 cents argument, since the subscription is monthly 🙂

You are correct nobody will 100% max out usage

[scarlet hinge]

lmao bing's 15 message per context window is getting close to GPT restrictions 😂

[opal badger]

By this time tomorrow we gunna be at 10 msg per hour. LUL

[wind urchin]

Really?? How many messages do you send

@wind urchin oobabooga/text-generation-webui github

[vague grove]

chatgpt is not realizing my account that i'm logged into is plus subscriber. i tried chrome and clearing cache. any ideas beyond that?

[lime hinge]

that's more than what we get now LOL

[opal badger]

per 3 hours* HAH

[wind urchin]

Is there any way I can get the December 2022 version that had no filter 😩

[forest folio]

Someone write a script to queue up every prompt you can think of.. and have it submit your prompts every three hours.. that way you make use of all your messages for the month, even when your sleeping XD

[wind urchin]

Got it thanks do you see this becoming as good as gpt 3.5 in a few months ?

[native musk]

If you add it up it's still 100 messages per day assuming you only have the 12 hours that you're awake. Maybe even more than 100. If it lowers it is still somehow gonna add up to 100 I think.

[sharp kettle]

Same here

[wind urchin]

probably not

[native musk]

And you can actually see that is moving faster than before, this is just for optimization reasons I think.

[opal badger]

Be one thing if the replies from the AI that are useless didn't count, need a refund button on the replies.

[gray sluice]

im gonna lose me job 😭

[wind urchin]

Wait gpt4 has a 100 messages per day limit ?

[native musk]

Yes

[opal badger]

25 per 3 hours atm and lowering fast.

[stoic lodge]

gpt4 told me that it does use the msg on the left each time so tokens are spent when using a behavior description, and its obviously limited by the max token limit (which will consider both the behavior msg and the regular chat msg, so 4k in each for example if its a max of 8k)

[agile narwhal]

How to fix something went wrong. If the issue persists please contact us through our help center atau help. Open. Com

[vague grove]

they officially have enough money that they don't need to care about alienating people. that's the big tech play nowadays

[native musk]

I think they may half it to 15 per every 2 hours or smt like that eventually.

[wind urchin]

Wait so isn't the monthly limit of gpt4 then 3000 messages ?

[trim garnet]

📉 ⚠️

[native musk]

Actually a half would be 12.5 but you can't do that.

So 15 seems more reasonable

[scarlet perch]

I'm not affiliated with OpenAI, sorry! but I'm sure they're aware and are working on improving things

[mystic girder]

you wont get the max amount of messages unless you set up a script to send messages 24/7

[stoic lodge]

weird how they dont do it by tokens in chatgpt, i feel like it shouldnt count one word replies and full essays as the same, could end up hurting u tho if u use long inputs/responses

[wind urchin]

Still it's 3000 messages compared to the 6000 you get from using the api unless I am thinking wrong

[mystic girder]

yeah you are kinda incentivized to let long rambling outputs finish, wasting tons of tokens even if you just needed a short answer

[scarlet perch]

That was assuming you are awake for 12 hours. It's actually 200 per day and 6000 per month, assuming you max usage.

[jolly vault]

@viscid bone how did you get pro mode to work? I downloaded a few different ones to try and followed all the steps to a T. basically, download and then add api key in the right spot. nothing.

[trim garnet]

It might depend on your use case, but for me with debugging coding or listening to instructions. It seems to understand the context of what you paste in and remember to use it for future prompts a bit better and doesn't just make random stuff up or get dementia as early as the older model does.

[wind urchin]

That's still not better than the api what's the incentive of buying the subscription then

[silent hawk]

what is this "pro mode" thing?

[wanton mirage]

I am honestly still consdiering buying chatgpt plus

[jolly vault]

its a shortcut for chatgpt+siri on iphone

[wanton mirage]

just because the API is so expensive

[wind urchin]

Yes most definitely

[trim garnet]

From what I've tested so far, yes it appears to be better at math. It hasn't gotten anything obviously wrong yet. Not sure about physics/chemistry but before 3.5 would get tripped up on simple math, this hasn't really for me yet.

[vague grove]

Do you have to pay for pro...? or just seti it up

[wanton mirage]

GPT4 is insanely good in math

[grave hatch]

Found this at the veeery bottom of the OpenAI developer page:

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

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.

[wind urchin]

"GPT-4 currently has a cap of 25 messages every 3 hours. Expect lower cap next week, as we adjust for demand."
I'm crying 😭

[mystic girder]

not only edgy content, it will straight up refuse to work with you on most things

[scarlet perch]

ChatGPT Plus was released a month ago and offers

General access to ChatGPT, even during peak times
Faster response times
Priority access to new features and improvements

GPT-4 availability was announced just a few days ago, and is limited to x messages per x hours (variable dependent on server capacity)

The original purchasing incentive was never exclusively GPT-4. 🙂

[wanton mirage]

lol didnt we just say bing is garbage

[lime hinge]

HUH?

[wind urchin]

not anymore 😦

[vague grove]

bro i can't even access chat gpt right now! this has happened every time i've tried to use it but now is the only time it's not fixing itself. just wont remember my acc HAS PLUS ALREADY ugh

[opal badger]

They keep lowering it..

[pale vale]

Wdym most things, I feel like you’re complaining a bit outside of reason

[trim garnet]

if by GPT-4 Plus you mean a horrible message cap then maybe, otherwise it feels like SmarterChild or Cleverbot

[jolly vault]

i think you just have to pay for the API usage

theres tons of videos on youtube

[vague grove]

dang alright

[wind urchin]

But normal GPT-3.5 is extremely fast now at least

[mystic girder]

nah I have tried bing many times and I can't get anything done, things that ChatGPT would get right in a single output

[wind urchin]

Nah it still hallucinates

[finite kayak]

Damn, AI is finding aliens now? /j

[trim garnet]

aint nobody want that though

[pale vale]

Like what???

[jolly vault]

@vague grove if you get it working please lmk lol

[wanton mirage]

what about math stuff ?

[jolly vault]

but as far as i know its super easy to set up

[mystic girder]

random questions, creative worldbuilding stuff, etc

[grave hatch]

gpt4 first got 100msg/4hr (equals 25 msg per HOUR) limit but it now shrinked to 25msg per 3 HOUR

[native musk]

Use them but don't fully trust them.

[trim garnet]

ok I want anyone defending Bing AI to explain how the 5 messages per topic is better than chat gpt?

[wind urchin]

Bing chat is severely limited due to the internet access since it will always base itself off of web rather than the procedure gpt goes through

[silent hawk]

Bing AI was GOAT for the first few days I had access to it.

[trim garnet]

yeah...have you used it since it went full meltdown mode?

[finite kayak]

Not long ago that would have sounded extremely weird

[native musk]

If they lower the time frequency, like every 2 hours is still doable even if the amount gets lowered.

[young ibex]

GPT4 now down to 25 per 3 hours. Is this 200 per day, or will i be cut off after 3 hours?

[vague grove]

i'm officially a stubborn idiot. logged into the wrong email. sigh

[obsidian heron]

I have a discord bot with GPT-4 and it randomly starts speaking like a pirate???

[pale vale]

I would say bing is better at things that don’t require a huge block of text to be generated, like essays, portfolios, write-ups, etc.

[grave hatch]

200 per day

[trim garnet]

try using bing ai now, it's actually useless

[wanton mirage]

what about math stuff ?

[jolly vault]

@vague grove its cool i feel like an idiot with everything i do 😂

[finite kayak]

Ye scurvy landlubber

[young ibex]

cool, thanks.

[pale vale]

Really? I don’t think you’ve actually used it recently

[thorny valley]

Spread ChatGPT to every area of the world.

[bleak bloom]

gonna spread something

[trim garnet]

If I try to have it output a table, or code, it refuses to give me any more than a sample. Even with the longer formal prompt modes. I ask it for more information and then it tells me it's time to switch topics.

[mystic girder]

use wolfram alpha

[wanton mirage]

what about ChatGPT plus (GPT4) vs Bing chat for math

[wind urchin]

You can actually set wolfram alpha up for bing chat to use it's much better

[lime hinge]

sorry there's hundreds of countries and i only have 25 messages every 3 hours. that'd take me like 3 weeks

[silent hawk]

I'd take 25/2 hr, or 40/3 hr

[pale vale]

Why not? Khan academy has built an AI tutor using gpt 4

[trim garnet]

yeah I am using the Edge Dev build and I've tried all of it

[wind urchin]

BTW all of you here you do guys do know you can use the same token over and over again right ?

[wind urchin]

This should be basic stuff

[covert stone]

m t DQ FJxD

[grave hatch]

there will be something like chatgpt ultra or gpt4 plus subscription

[scarlet hinge]

its still far from perfect, and something as technical as math, it may hallucinate somewhere and it'll be really difficult for you to know

[grave hatch]

with unlimited access to gpt4 8k, limited 32k and maybe its $42 (like the cancelled "chatgpt professional" thing)

[peak bridge]

there was a chatgpt professional?

[scarlet hinge]

maybe they introduce a 50$ superfast gpt-4 version with unlimited access

[grave hatch]

yea and it was $42

[peak bridge]

why was it removed?

[scarlet perch]

There was a limited "release" of an early version of ChatGPT Plus that was retracted after a few hours. The $42 may have been a placeholder price.

[scarlet hinge]

yea that was mostly placeholder

[silent hawk]

42 is a meme number, obvious placeholder

[grave hatch]

maybe its too expensive but idk

[scarlet hinge]

cuz 42 is used as a placeholder number in many cases

[silent hawk]

I use 42 as a placeholder for a lot of things

[wanton mirage]

I feel like GPT4 in the API was still better overall tho

[grave hatch]

true

[peak bridge]

how long did you have to wait for for the API?

[lime hinge]

dunno. subs can still barely use it

[polar light]

What are you finding is better for coding? I’m working my way through Ansible and I have the logic but the playbooks are a pain to get step by step.

[mystic girder]

there won't be an "unlimited access" version, even the most expensive tier will likely have a cap

[vague grove]

My mouth literally just dropped at how expertly this dang AI just wrote a literary scene for me. It's so specific and it basically nailed it. Just, wow.

[wanton mirage]

a day

[grave hatch]

i guess next is 15 msg per 2 hours

[bleak bloom]

1 per day

[vague grove]

The difference between 3.5 and 4 was HUGE as well. I specifically told it the opening lines and 3.5 freaking put them at the end! It overwrote me, arrogant AI

[wind urchin]

0 per day

[grave hatch]

theres a site that you can access gpt4 for free but its 1 per day

[scarlet hinge]

i mean unlimited = too much to use for one person

[mystic girder]

yeah it's super good for writing, I even played an entire pen & paper RPG session with it back when the cap was more forgiving

[gritty path]

Doesn't it feel like we're being scammed here ?

[vague grove]

I literally am inspired to finally write my book. This is what I wished I had last year omg

lol "write"

[lime hinge]

yup and there's no option for a refund when they just sold this for gpt4

[scarlet hinge]

pretty sure Google is rate limited too and it's not "unlimited" in the absolute sense

[formal kraken]

I asked it invent something using a led light, condom, and a pack of pasta - nailed it!

[gritty path]

I mean I subscribed two days ago and it went from 100, to 50 to 25 prompt per 3 hours and it says it will keep going, what am I paying for exactly ?

[mystic girder]

unlimited fast gpt3.5

[gritty path]

Not interested in that.

[grave hatch]

bruh why openai why

[shell gust]

I don't really see how GPT-4 is better

[grave hatch]

no the plus version still has that 1 hour limit

[gritty path]

I used both for coding and let me tell you that the difference is massive.

[polar light]

15 tokens is like, not even a compose file.

[mystic girder]

really? huh, I tried to hit some kind of limit by asking tons of short inane questions but I didn't run into any caps on 3.5

[scarlet hinge]

they did say (even at the beginning) that they'll revise the cap depending on demand

[shell gust]

Haven't used it for coding yet, I know what I'm doing most of the time and don't need chatgpt

[lime hinge]

it does do more detailed review and it's able to store more in memory. the only issue is the severe message restrictions

[tawny wave]

Bro first it said 35 messages and I reloaded and it changed to 25. WTH

[grave hatch]

💀

[peak bridge]

ye lmao

[shell gust]

is it good for competitive programming

[lime hinge]

WICKED

[silent hawk]

It's dynamic

[shell gust]

ok

[gritty path]

My point is that I've purchased a subscription for a month and two days into it the main reason why I subscribed is spiraling down. Feels like a scam to me.

[silent hawk]

@shell gust are you a large language model?

[grave hatch]

so they may expanded the 1hr limit on plus, but it isnt unlimited

[lime hinge]

PREACH

[shell gust]

what do you mean by that

[trim garnet]

My concern arises from the overwhelming demand for GPT-4, which has led to a reduction in the usage cap not once, but twice. With a $20 fee multiplied by the countless individuals who have subscribed to ChatGPT Plus, I can't help but wonder about the revenue generated from this endeavor. Given the server constraints and the necessity for such significant adjustments to user access, it's hard not to suspect that a substantial profit has been made, while our access has been considerably limited.

[scarlet hinge]

it has a limit it's just very high

[true ferry]

"GPT-4 Currently has a cap of 25 messages every 3 hours"

What the hell! When I first got it I got 100 every 4 hours. This feels ridiculous that it is going down and down

[silent hawk]

I mean, you're llama, the large language model released by facebook/meta, aren't you?

[mystic girder]

it went from 100 to 50 to 35 and now 25

[short vault]

They did the same thing with 3.5. They are still improving the thing. the demand is high as it is, considering how fast they went from 3.5 to 4. i think its fair to give them some leeway to put things in place. i dont see them keeping a limit for the plus subscribers forever

[gritty path]

Yep, and will keep going down even more.

[true ferry]

Exactly. What a joke.

[trim garnet]

also true

[true ferry]

Might have to refund it really. This is just not what I am paying for.

[umbral urchin]

anyone have any thoughts when itll start going up and not down?

[mystic girder]

They are just panicking because too many people bought into the subscription, just gotta wait until they can scale things like they did for 3.5

[wind urchin]

Ernie Bot CCP 5 can't come soon enough... OpenAI will be out of business

[shell gust]

no

[gritty path]

Probably never, pretty sure we just gave money to OpenAI so they can develop their product to sell it to the upcoming developpers.

[trim garnet]

Not when you pay 20 bucks. How many people does it take using gpt for them to enforce the cap reduction? Because they made 20 bucks off each one of those people.

[short vault]

Will likely be a month before its considerablly better. the traffic will unlikely get better in the short term

[silent hawk]

just checking, you can never be sure nowadays

[wind urchin]

Why are these nuts releasing GPT 4 at such a fast scale without being able to handle it??? I'm seeing them release API access to random people when they're clearly not able to keep up.

[trim garnet]

The big difference here is that these aren't the free loaders like before

[short vault]

Have you seen how much the API costs? 20 dollars is a DEAL

[true ferry]

Right. They should have known this would happen! They saw how many users they were pulling. They are now finally making money (and a lot of it!). Surely they can scale the servers now they are making absolute bank

[lime hinge]

yup

[wind urchin]

I've seen how fast the API is though, they're putting priority of random lottery winners ahead of people who PAY

[wanton mirage]

very true which is why I am considering to buy it - especially since bing chat is garbage

[gritty path]

If it keeps going down, we will get something like 10-15 messages per 3 hours then what ? 5 ? 3 ? 20$ for 5 prompts doesn't seem like a good deal to me even if it's the best thing around.

[trim garnet]

yep

[scarlet hinge]

scaling is hard but yes they should have had the foresight

[grave hatch]

5 msg per hour 💀

[true ferry]

For real. Honestly. They should have planned better for this many users.

[trim garnet]

^

[scarlet perch]

They are currently scaling, as described on their website

[wind urchin]

They're going to have us doing a message a hour like Poe soon 😭😭

[grave hatch]

true

[wind urchin]

I'd be ok with it if we got the 32k version tbh but it's probably not gonna happen

[gritty path]

They are currently scamming millions of users you mean.

[scarlet perch]

I don't believe so no

[wanton mirage]

guys chill - they might have it figured out by the end of the weekend

[wind urchin]

More like downscaling! 25 now?? Crazy!

[gritty path]

Nah this will be exclusive to the big corps.

[true ferry]

Scaling would be hard if they were on their own servers. It's literally hosted on Azure isn't it? They have pretty much unlimited funding. They can easily increase it.

[short vault]

Im not terribly concerned. its very much a beta item yet. they have bugs to work out and everyone is slamming the thing. they are likely looking to take it wide so they need to limit the beating they are taken somewhat

[wind urchin]

This!!

[gritty path]

It literally says it will keep going down by next week.

[trim garnet]

yes this....where the 10 billion at? 🔍

[true ferry]

It's not scamming no. But this is pretty awful what they are doing, making us pay such high amounts for this poor product.

[wanton mirage]

wait can you copy paste it in here what it exactly says

[short vault]

I dont see that being likely. mabye by the end of april. im good with waiting a bit.

[grave hatch]

even the api of 32k was too expensive, $0.12 per 1k for completion. so they prob release a higher tier subscription

[gritty path]

GPT-4 currently has a cap of 25 messages every 3 hours. Expect lower cap next week, as we adjust for demand.

[scarlet perch]

Server scaling and further optimizations will likely raise that number soon, we'll have to wait and see 🙂

[wind urchin]

Where's the 5 billion and change they had? I'm sure they could even work out a discount with Microsoft with the partnership they have!

[true ferry]

But at the end of the day it's the entirety of Microsoft. If Microsoft really can't spare a little bit of server host for it, then they should move elsewhere. That's a con!

[wanton mirage]

damn that sucks

nvm I wont buy it then

[true ferry]

They told us to prepare for a lower cap next week? What the hell!

[grave hatch]

lmao

[pearl scaffold]

are yall bots

[wind urchin]

I can see that they were absolutely not ready for this release though. Next we'll see a increase in price with no increase in quality!

[gritty path]

Don't, at 100 per 4 hours it was totally worth it, but now, I expect something like 10 prompts per 3-4 hours by next week.

[true ferry]

Yep. This is not the way to keep retaining customers. Especially with all eyes on them right now. Even if it had rubbish generating times, what they are doing here is not it.

[short vault]

I think everyone forgets you still have 3.5 and they haven't asked for more money.... yet

[formal kraken]

I think they may be trying to monetise whilst they can! Google will shock the world.

[wind urchin]

it might take time to scale up, build new clusters, but in the short-term it goes down. it's friday

[true ferry]

Nah mate. We're all idiots giving them 20 quid.

[gritty path]

Like I said, having tried both pretty extensively, it's not even close, I wouldn't have paid for gpt-3.5

[true ferry]

They won't keep them for long at this rate!

[wind urchin]

Ready for Google and other companies to drop the bomb of their AIs on OpenAI. Then they'll try and keep us, but I'll be gone ✌️

[grave hatch]

i guess no one here paid $20 just for gpt-3.5 in higher speeds

[true ferry]

Yes mate, It's Friday but this is a company valued at 29 billion dollars unable to handle people using their services. Users paying a lot of money

[wind urchin]

If I wanted 3.5 I'd just use the API! I only want 4 😭

[gritty path]

I will definitely try LLaMa

[mystic girder]

Getting only 1 month of subscriptions that get instantly cancelled would be very short-sighted of them, I am confident that they'll fix the cap by the end of the subscription renewal date

[gritty path]

I don't want them to lower their prices, I want the number of prompts to go back to what it was when I bought my subscription 2 days ago.

[wind urchin]

Watch them make a ChatGPT Plus Plus 😭😭✌️

[formal kraken]

as in the end-game that that alien said to hulk, and thor, wait - let them have their fun!

[true ferry]

They are offering GPT-3 for free. What I am saying, Is me (and many others) are just going to go back to GPT with the limited access if they keep this up. Really we're paying 20 dollars a month for full time access. GPT-4 is a privilege we get to look at! The "alternative" here is their free version!

[grave hatch]

same. its enough for me

[trim garnet]

I think you're delusional if you think the only reason people paid was for 3.5 speed....

[wary iris]

Llama is text generator?

[grave hatch]

more like chatgpt ultra or gpt4 plus

[gritty path]

Yeah, open source I think.

[wind urchin]

If they do what sparknotes did and make originally free features paid im gonna be so mad

[wind urchin]

GPT-Platnium! Wait list only and $65 a month only!

[gritty path]

Tell me a fun story about a company that scams people.

[wind urchin]

think of it more like a brain. it needs training first.

[grave hatch]

see, when gpt4 releases ig the plus user amount blasted

[wind urchin]

its CAPABLE of being a text generator

[kindred girder]

Looks like LLAMA is the future. it can run as good as GPT 3.5 on a Raspberry Pi. its way more efficient compared to GPT3.5. GPT4 might be really good but ina. few months LLAMA might run GPT4 level on a Laptop. Where do you think businesses will go once they see who can run it locally?

[grave hatch]

🤣

[true ferry]

For real, that's what we're moving too this rate.

[trim garnet]

exactly, there is a reason why they reduced the cap and we're all having this conversation

[true ferry]

Lmfao.

[silent hawk]

I still wish they added the "pay to ban other users" feature I requested. I pay $10, get 10 random users banned.

[gritty path]

Yeah, and let me guess, rince and repeat ?

[true ferry]

It's a joke man. I am so done with Open AI.

[gritty path]

Because random people are overrated.

[autumn plaza]

Shut the hell up bro.

[scarlet perch]

Rule 1 please

[grave hatch]

i aint complaining. i understand the openai server cant handle this big demand, but cant they just finish adjusting their server size before launching the gpt4? i can wait.

[autumn plaza]

Sorry.

[silent hawk]

because we don't have lists of the users, so random is the only way

[gritty path]

They needed that juicy money mate.

[grave hatch]

they even released it two days early

[true ferry]

That's literally the entire point of this complaining. They released something they were unprepared to launch, and they are now feeling the result of it.

[silent hawk]

while they're at it, pay to read other people's chat messages

[gritty path]

I'm sorry, but as a small language model these are the only words I know.

[grave hatch]

true

[autumn plaza]

My Nan died and I need therapy. 😭

[true ferry]

Rule 6.9 please.

[scarlet perch]

Thank you 🙂

[autumn plaza]

OMG that is so offensive!

🤬

[true ferry]

Rule 5.3X please.

[silent hawk]

now that's just mturk

[gritty path]

So is it always that busy here or is the 25 message cap got you all fired up ?

[autumn plaza]

Can you refer me to a specialist please?

[silent hawk]

yes

[obsidian heron]

gpt-4 is the future of cyberbullying

it just gave me an insult i can never recover from

[scarlet perch]

This channel is usually busy

[gritty path]

It could also be the other way around, cyberbullying is the future of gpt-4

[obsidian heron]

true

[left sleet]

what was it?

[autumn plaza]

Well if you look at Rule 65-B, it is clear you're breaching community guidelines.

[true ferry]

Always this busy. But today I am really rowdy because of this cap!

[trim garnet]

the scam has gotten us fired up, yes

[scarlet perch]

None, we don't allow self-bots

[wind urchin]

+1

[obsidian heron]

i can dm you the entire thing if you want

[true ferry]

I pay over 20 dollars a month for this! And I feel conned.

[grave hatch]

same

[lime hinge]

+2

[autumn plaza]

Damn AI!!

[true ferry]

+1!

[silent hawk]

Discord doesn't either, but you never know

[left sleet]

no thanks but that's interesting

[obsidian heron]

aight

[autumn plaza]

+1 My dude.

Mum said I am not allowed.

[gritty path]

I'm thinking about selling my 25 prompts per 3 hours now, here's another good YouTube video idea.

[grave hatch]

+1

[true ferry]

If only Bing Chat was actually half decent.

[gritty path]

''How to make money selling you 3 prompts per year that you pay 20$ a month for with GPT-4''

[obsidian heron]

luckily there's not limit on the api usage

[autumn plaza]

The Government are watching!

[gritty path]

The api is not even out yet

[grave hatch]

but its about 20x expensive than gpt3.5

and you must wait

[obsidian heron]

yeah it is i got it today

i've built a discord bot with it

[left sleet]

that's because you pay to play

[short vault]

20x? try 60x

[gritty path]

Oh that explains everything then

[scarlet perch]

This channel is for discussing GPT-4 itself remember 🙂

[obsidian heron]

it's $0.03/1k tokens

[short vault]

Model Prompt Completion
8K context $0.03 / 1K tokens $0.06 / 1K tokens
32K context $0.06 / 1K tokens $0.12 / 1K tokens
Chat

[obsidian heron]

it's cost me about $4

[autumn plaza]

Here is a joke guys:
How did the tree gain access to his PC?

[short vault]

3 per 1k prompts + 6 per 1k replys

[true ferry]

The government are inside of Open AI. They know about me asking ChatGPT how to build a Molotov that is able to destroy a medium to small sized car.

[left sleet]

i got into the root?

[grave hatch]

gpt3.5turbo is 0.002 i remember

[autumn plaza]

It logged on!

[wind urchin]

rooted it

[silent hawk]

i realize my free credits expire in less than a month

so I will hurry up and use those

[autumn plaza]

😂

[wind urchin]

lol

[obsidian skiff]

RIght, I"VE JUST GOT HERE AND ALREADY CAN'T STAND IT.......I don't know what is going on in this world these days but I can't just be a bystander for all of it. We have to do something. We have to change. There is no way we possibly can keep living like this as human beings of this earth. We have to fight together to make a change in this situation and GET WHAT WE DESERVE. WE WANT MORE MESSAGES FOR OUR MONEY!!!!! We need to get this GPT owner kid BANNED. I don't care what it takes, get him the frick out of here. I can't take it anymore I'm spiralling down an endless hole of money debt and lack of GPT messages. Please I ask of you to help me and rise up against this. We can do it guys. We have to be stronger as a team. AS A UNIT!!!!!

[kindred girder]

GPT 3.5 140 Billion Words ( December 2022)
GPT 4 1.0 Trillion Words (March 2023)
Lambda 1.75 Trillion Words (June 2022)
Palm 540 Billion Words (April 2022)
LLAMA 65 Billion Words (March 2023)

[autumn plaza]

^ Average Garry's Mod player.

[grave hatch]

same

[kindred girder]

known numbers

[true ferry]

Preach!

[grave hatch]

mine expires on april fools

[thorny valley]

The whole world needs ChatGPT.

[grave hatch]

💀

[scarlet perch]

Let's keep this channel for discussion please, not spammy caps messages. Thanks.

[silent hawk]

you mean parameters, not words? LLaMA has seen >1 trillion tokens

[left sleet]

i agree

[true ferry]

Not in its current sate! We need something new!

[obsidian skiff]

Sorry sir, my bad

[rugged dawn]

Bruh when is gpt 4 gonna be free I’m too broke to buy it 😵

[obsidian skiff]

I didn't mean to

[thorny valley]

New ChatGPT upgrades

[true ferry]

Operations Lead eh. You had something to do with this didn't you.

[obsidian skiff]

please forgive me sir laze!!!!

[left sleet]

to an end it could

[silent hawk]

We're far past discussion now. Welcome to, the Twilight Zone.

[obsidian skiff]

I am forever in debt to you

[scarlet perch]

I'm not affiliated with OpenAI.

[true ferry]

Not even paid users are getting it at this rate mate.

[kindred girder]

Google uses "words" as term so i used that

[true ferry]

What?

[autumn plaza]

@true ferry Why are you leaking my IP????

[grave hatch]

imagine 1 msg per hour or even day

[silent hawk]

Google is wrong, just like Bard

[scarlet perch]

Check the staff descriptions in the #welcome channel.

[left sleet]

actually i'm curious, how do you contact openAI?

[autumn plaza]

Google is always right.

[lime hinge]

gl

[silent hawk]

G-Zombies

[scarlet perch]

You can contact support at [e-mail]

[trim garnet]

LOL

[brisk sail]

who hear waiting to get gpt4 access?

[autumn plaza]

Lol.

[gritty path]

You just don't

[true ferry]

Yeah well, you have insider knowledge. Write in the channel we saw them use in the video, asking them to change it!

[brisk sail]

good luck with that

[trim garnet]

👁️

[left sleet]

awesome thanks. I don't think the feedback on the site worked

[silent hawk]

Ah yes, the chaos is here. I'll pop some popcorn

[grave hatch]

lol

[gritty path]

gimme some

[brisk sail]

i probs have more a chance of a girl saying they liking me than openai getting back to you 💀

[autumn plaza]

Done, it showed me a Nissan GTR with a Banana in the exhaust.

[gritty path]

that's pretty much it.

[rugged oxide]

I have found that when you ask chatgpd to make a code it stops.

How do I fix this

[brisk sail]

token limit

[autumn plaza]

You don't.

[gritty path]

tell it to continue where it stopped

[brisk sail]

i say continue where you left off and then copy paste the code back in

[autumn plaza]

It is unfixable. Just deal with it.

[rugged oxide]

When I do it just restarts the code.

Oh oh thanks for the help everyone!

[autumn plaza]

😎

[gritty path]

I normally copy paste the last bit and tell it ''you stopped, please continue from (then I paste it here).

[scarlet perch]

I sometimes inform GPT that "you stopped your response before it was finished" first, so it definitely knows

[wind urchin]

is gpt4 going to be free (srry if this is a dumb question)

[twin yew]

Don't be lazy, guys... cmon

[autumn plaza]

Yeah, I usually just refresh my browser.

[silent hawk]

Well, there's only one OpenAI but you can ask as many girls as you need before you get the response you want.

[oblique merlin]

all hail the gpt 4

[gritty path]

Too soon

[thorny valley]

Wordpress is the best way to create Any website.

[gritty path]

We're all mad here mate

[thorny valley]

I was using Editor X - Wix when I was ultra new but with Elementor Pro and ChatGPT helping with PHP.. lmao

[autumn plaza]

I am going absolutely mental!! 🤬 🤬

[thorny valley]

DigitalOcean is great to host the Wordpress on

[heavy palm]

guys did anyone bought the subscription and then out of a sudden it was gone after 1 hour?

[true ferry]

I would use Gpt 4, but apparently you can't get that anymore! 😂

[thorny valley]

ChatGPT and PHP Wordpress 🙂

[thorny valley]

I still use GPT3 as it's faster

[gritty path]

Let's all prompt GPT-4 to self desctruct at the same time

[true ferry]

Try to sign in with the account you brought it with

[crude dew]

100 -> 50 -> 25 -> "Expect lower cap next week"

For $20 a month?

As far as I know this is an official release, so this is a bit unreasonable.

[true ferry]

AVG Gpt 3 User

[zinc timber]

Open Ai may as well just remove gpt-4 and step up their game and release it when paying users can utilize properly.

[autumn plaza]

Yeah it is because ChatGPT 5 is coming out.

[true ferry]

Yes mate. I know. We're all a bit fuming! We should protest.

[true ferry]

Yes mate.

[thorny valley]

my backend recipes... Unity C#, Wordpress, Android Kotlin, Supabase

[stoic surge]

it was slow

[autumn plaza]

💯

[crude dew]

I mean paying customers... on a non-beta service.

Not cool

[autumn plaza]

😱

[lofty surge]

[light crane]

Was about to subscribe to have a play. Is this not worth it?

[gritty path]

GPT-4 is a beta male

[heavy palm]

i did still same issue

[gritty path]

If you want to have a limit of 0.5 message a year by next week please go for it.

[craggy frigate]

Yeah, I just got cut off and when I counted back through my prompts I have only used 11 and now I'm cut off, it's saying to try again in about 20 minutes.

[autumn plaza]

Yeah, you could delete system 32 though.

[lofty surge]

Chat got 4

[silent hawk]

sqrt(1) messages/e^10 years

[crude dew]

Not worth it until they scale the systems and stop gouging paying customer's. It'll take some time I guess.

[gritty path]

Is that the distance between the sun and my bathroom ?

[autumn plaza]

Agreed.

[stoic surge]

1 character per 100 billion year

[silent hawk]

probably

[gritty path]

Then it's worth it.

[crude dew]

I won't be renewing my subscription unless I see some improvements in availability by the renewal date next month

If it's improved, then I will

[gritty path]

Then you won't be renewing your subscription ever.

[crude dew]

You really think so? I think they'd want to retain paying customers

[light crane]

Ahh ok. I’ll hang fire then. Seems some distress in here! Is it different to Bing Chat Bot? If that used GPT 4?

[crude dew]

I think a lot of others probably feel that way

[gritty path]

They already made the money they needed, now they launched the api today. My guess is that the big guy will have it from now on and we will be left with whatever they feel like giving us.

[supple harbor]

They said they expect to be "extremely" capacity restrained for the next while. Eventually it will improve, but who knows maybe we'll be on to GPT 5 by that point and no one will want to use GPT 4.

[scarlet perch]

I expect we will personally, but that's speculation, so we'll have to wait and see

[gritty path]

Unless they go bankrupt

[sharp elm]

How close are we to the image feature being released to the public. Anybody?

[gritty path]

Can you link it ?

[supple harbor]

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

[crude dew]

That's sad to hear. I'll hope for the best, and start getting involved with the open source projects in the meantime. Actual "open" AI haha

[gritty path]

I think you'll be able to use that feature using the api and developping something like they did in the preview with the discord server.

[sharp elm]

Hm okay

[potent oasis]

what is exactly replit? and what do you mean when you say 'people'?? you mean for free?

[wind urchin]

...25 messages? what??

[gritty path]

Come, we're all mad about it here.

[fleet forge]

if i want unlimited messages, unlimited length...how do i use the api to do that? not a programmer, is there someone i can pay to just use this without limits? am pro but 25 msg not gonna work for me

[gritty path]

You pay for tokens, if you have unlimited money then you have unlimited prompts.

[fleet forge]

oh ill pay . how do i use it?

i need to do dev work?

im a simple dumb writer

[gritty path]

Subscribe to the waitlist for the api and give them your credit card, might as well post it to them at this point.

[supple harbor]

Can't do unlimited length. The 32K version is $0.06 (prompt) or $0.12 (completion) per 1000 tokens: https://openai.com/pricing

[scarlet perch]

You can read up on OpenAI's API documentation here
https://platform.openai.com/docs/

[gritty path]

No but you can break it down

Yeah just give it to them, you don't need that, they do.

[quiet pendant]

GPT should cost more

[gritty path]

GPT should offer me the option of paying more but with a guarantee...

[drowsy cobalt]

me w $2 on my card

[crude dew]

It should cost more and have more restrictions! 5 per 3 hours, why not? ;D

[gritty path]

Should cover the transaction fee

[high sky]

i just barely adapt 50 yesterday, 25 is insane

[quiet pendant]

Its way too cheap your gonna put the foreign telemarketers out of business someday

[drowsy cobalt]

100 for 4 hours was the best ;-;

[supple harbor]

If you use completion you get the "chat" experience. For reference, GPT-3.5 is $0.002/1000 tokens, so less than a tenth the price. 1000 tokens is about 750 words, so every 24,000 words will cost you $3.84 with GPT 4.

Note as well the Chat version of GPT4 is only 8K, so you might not need 32.

[gritty path]

50 was understandable, 25 is COMPLETELY INSANE I SAY

[crude dew]

Yup..

[open bay]

So, will gpt - 4 be able to answer with generated images, videos and audios? When will this be launched? Plus users will be able to access it?

[drowsy cobalt]

yeah especially if u asked short questions

[supple harbor]

They did say they might come out with higher pricing tiers for more GPT4 chat prompts.

[drowsy cobalt]

also a lot of the time it is just terrible at writing references and i have to ask them to rewrite it so that is more messages gone

[supple harbor]

If you're only asking short questions you may be better off using GPT 3.5 for a lot of them.

[drowsy cobalt]

i dont, i ask it to help my essays

[gritty path]

I get it but it's sort of a scam you know, bringing the plus user down to 25 prompts and then offering more prompts to those that pays more is sketchy.

[open bay]

Is it 25 per 4 hours now? 🤯

[peak igloo]

Stop asking GPT a one sentence Q at a time and stack your inquiries in one prompt. It will do everything you've asked in turn