#gpt-models
1 messages · Page 36 of 1
use revolut bro
oh, thanks for the reply
You may wish to contact support@openai.com if you're having trouble accessing the services
I get that, and since it's still a relatively new product I expect the developers will make some adjustments along the way
They don't reply
Relatable haha. (Although I'm not writing thousands of lines yet hehe)
What a great time to be alive and get first hands experience in AI and using it to our advantage.
100 req per 4 hours is 25 per hour. I think that can be topped when you work on research, or you ask questions based on a text that you pasted.
Yeah it can be topped.. but personally feel like there are some task that GPT 3.5 can do totally fine, so i switch around a bit.
Even if you just engaging in normal conversation, 25 per hour is definitely not enough.
gpt-4 is also in bing btw
Sure, I do the same. The speed of 3.5-turbo is just much more important to me. When it produces stuff that I don't like, I sent it to GPT-4
I wonder if gpt 4 can make a theme for better discord
feel you. BTW does anyone know if something like an integration for programms like PyCharm is planned ?
You mean, Copilot? ^^
Have you tried with GPT-3.5? I'd bet if 3.5 didn't give you something at least 80% working, 4 won't either.
Oh yeah hahaha. Totally forgot that tool (but never tried it to be honest)
Does it also have this "human element" to it?
Whaaaat. It's amazing
Yeah when you ask for it ^^
I feel like discussing code / problems with gpt is a bit like talking with another dev
Ok thanks for the advice, I will try it 🙂
Probably coming soon. Jetbrains IDEs have the GitHub Copilot plugin. For VS Code there is also the Code-GPT plugin which looks very promising. I'd bet something similar will come to Jetbrains sooner or later.
50 messages every -> 4 <- hours now? 😦
I think Codex (the model of Copilot) will be re-trained on GPT-4 sooner or later.
looking forward to it.. gpt probably taught me more than any prof at my university haha
I'd bet. The problem with the Copilot plugin IMO is that it has very limited user interaction. Take a look at what Code-GPT does and you'll see what I mean.
Okay, I'll go to sleep. It's nearly 4am here. I have to work soon. Was nice with you guys!
💯 If nothing else, being able to just ask a specific question and get a reasonable answer is worth money. Compare that to googling stuff and then ingesting dozens of Stack Overflow pages to get sensible info on a problem...
👋
Likewise, good night 🙂
If you buy plus, do you automatically get charged the next month?
Yeah and gpt is also not so arrogant (yet) like some users on Stack overflow haha
You can buy it and then cancel it though
Yes. Down from 100 messages every ->4<- hours.
You'll still have the month you paid for
why cancel though 😥
Has there been word on the GPT-4 cap being temporary with capacity increases coming soon?
"GPT-4 currently has a cap of 50 messages every 4 hours"
As a Plus user, I just think 4 hours is a bit much with that cap...
Just saying if you don't want to automatically charge
Thanks, i really dont like subscription model bussiness
Are y'all getting 100/4hrs? Mine says 50/4hrs
I'm at 50 now
It's 50/4 for everyone
I have already subscribed to Plus but since yesterday I cannot use chat gpt4 and there is a button on the sidebar prompting me to upgrade to plus
ah ok thanks
They can be arrogant as much as they want if at least the answer is factually correct IMO. And yes, before you ask, my standards are that low.
Maybe the next month when the chatgpt-4 booms die, it will go up to 100 again
Will there be improvements in the creation of codes?
💯
They are slowly rolling it all out. I doubt it takes a month but we do have to be patient in the meantime. High demand. It's like landing a rover on Mars lol. Gotta take it slow at first so it doesn't just fall over.
yeh. Well heres to hoping sooner than later!
API calls to gpt4 now too slow to work with Zapier.
Well, pragmatism always beats pride over time ;). Anyways, was nice talking I need to get back to some image labeling
Can you be more specific? As far as I can see, GPT-4 is better at understanding the "finesse" of stuff you put in the prompt. If you just ask it to write a ToDo list app, it will probably give you a similar result to the GPT-3.5 output.
👋
Is anyone having trouble with GPT Plus subscription? It accepted the payment and was working to get access to GPT-4, but now my account is back on the "free" plan.
Had it pretend to be my dog and propose a bill to tax cat owners. Think I can get it passed?
Grrr-arrgh, cat-rrr tax-rrr!
Chrrrrrr:
- Grrr-all cat-rrr own-rrrs pay-rrr tax-rrr.
- Tax-rrr high-rrr, grrrrrr!
- Use-rrr tax-rrr mon-rrr for dog-rrr park-rrr! Ball-rrr!
- No cat-rrr in dog-rrr park-rrr!
Grrrrrr!
I'm no engineer, but not only have I setup a Vue app on my desktop using Firestore with a functioning simple proto site, but I've actually learned to sift through code and help weed out where I prompt GPT next. It's pretty amazing. It does get loopy sometimes when trying to fix certain issues but I think someone with a bit more knowledge than me could have avoided that.
Can you get gpt 4 without paying for plus?
Some users seem to have that problem. If you search here for "subscription" or "payment" you'll probably find a few hints.
Ok thanks.
😂
Welp, GTP4 just gave me a lesson tonight into nuclear physics
as well as example problems
I feel like I learned more than what school has taught me
Sweet, what'd you learn?
yeah, if the other dev had amnesia and was also an avid stackoverflow contributor
Consider it homework for tomorrows lessons?
Ok. IMO, doing the "code this or that for me", "there is this error with your code", etc. loop works well with 3.5 already. Not sure if version 4 will improve much on that. But if your question is "should I pay $20 to get a peek at GPT-4?" then I'd say do it if you can spare the $20.
I've learned how to calculate the remaining mass on an isotope, as well as the decay constant
Just got a network error on GPT-4
Tomorrow i'll go back and learn about Fusion energy release
E = fΔmc²
Need to dive more into it
50 messages every 4 hours is limiting but we all just have to have patience until they upgrade/optimize the AI for better performance.
The main difference I have noticed is GPT-4 follows instructions much more capably than GPT-3 does, however it's only really a big deal if you have a very complex topic to discuss or strict guidelines to follow.
Otherwise just use GPT-3 most of the day and GPT-4 for some really nuanced things that require a lot of comprehension of complex instructions.
My main gripe is I wish you can switch models within the same chat.
The first things I asked ChatGPT on GPT-4 were kinda basic but kinda niche and it performed terribly on those 🫠 . Consider asking it:
How do I write the chinese character 必 stroke by stroke. And try following the instructions that it gives.
is gpt4 for free ?
100%. I agree. I'm on Plus and went through that 3.5 to 4 transition within that project, and I noticed some fundamental differences. Even simply how to communicated code snippets was more effective, avoiding the abrupt stops most the time that 3.5 would run into. 20$ is def worth it for my use case. Just hope that cap rises again soon because this type of project i'm doing does require some volume of iterations to get good results.
You can use it if you have chatGPT Plus
No, it's 20 dollars a month
Im new
Have you tried giving it tasks that would bore you, yet? Like if you take a piece of code that is not good and tell it to refactor duplicate code into methods or something like that?
i don't but ok that's fine thank u anyways
is gpt 4 really that much better than gpt 3 at writing code?
its better at most things
If you have gpt plus do you still have to pay in tokens to use gpt 4?
Yeah. Also can't wait to try the 32k token model. How awesome it will be if you can just paste a whole source file and ask questions about it!
I feel like if we could just switch models within the same conversation and see how much of our cap we have left it would be a major QOL upgrade
ChatGPT plus is $20 a month flat. But there is a limit on how many prompts you can send to the GPT-4 model per time.
god i wish theres an option to pay for how much you use
you know as long as the cost isnt insanely high
Yup! Things are going to change fast.
+1. Alas, with the prices of the GPT-4 API I'd be a poor man soon.
i cant afford 33 CAD worth of chat gpt per month, but ill pay for like 300, 400 messages if its 10 bucks or so
I'm willing to pay a bit more than $20/month for a bigger cap than what they've currently given for gpt-4...
So I saw a video saying that GPT 4 passed all the AP exams. I’m not American so I need an American to answer. Is a 2 a pass in an AP exam because I saw that it got a 2 in AP English Language and AP Literature which is sort of concerning since it is a language model.
two accounts lol
I did hear mention of adding another subscription tier for this very thing
i thought about it! i guess if it's 2 separate projects it makes sense. but i wouldn't want to break the congruity of my convo with chat for a single project.
yes please! and if it's for work, companies will easily dish that money out. they need that business tier. i'll have zero issues convincing my boss to subscribe for me, as i'm sure so many companies would at this stage.
on average, how much more expensive is gpt4 API than chatgpt api?
If I want to be able to use OpenAI playground with GPT-4, do I need to sign up to the API waitlist?
I've considered two accounts as well but it's so annoying to break the contininuity of one conversation just for continued usage
It's something the developers are only considering though, not sure if they will actually implement it
weird, i thought those subjects would be where GPT 4 excels at
I think the api is cheaper unless you have an entire company using it. Someone worked it out and you have to write 7.5 million words in each month for it to be equal to $20.
it does perfectly in bio and almost-perfect in other subjects
no clue why it fails in english, and yeah 2 is a fail
yeah but dont you have to pay a starter fee or something
English testing is extremely strict, this model is for human language, not necessarily used for the strict rules we have to adhere to in English class.
It depends. As soon as you start to automate things you're in danger of using lots of tokens.
Natural language flow is very different from English class, in my experience
Anyone get the error The model gpt-4–0314 does not exist --- I did get the email saying I was approved for it
That’s why I mentioned it being used in a company
That happened to several people.
Do u all know Microsoft 365 Copilot?
gpt4 cap went from 100 every 4 hours to 50 every 4 hours?
Yeah I was surprised too. It got a 4 in Calculus BC
One thing I learned here: the gpt-4 model seems to be only available for the chat API, not for e.g. completions.
True. I'm sure the 2nd wave of generative ai tools that stems from GPT will be more geared for that kind of business use.
which was also surprising, i fed GPT 4 some AP physics problems and it only got it right half the time, if not less
No. Is this for MSFS?
You can search online,it's so powerful
Yikes.
when its right its really right though, explains everything perfectly
it was only on kinematics, i didnt do in depth testing
You would think it would be better at math and physics since they are so deterministic
¯_(ツ)_/¯
hmm... I find it hard to justify paying 20 a month when they can just place a limit anytime they want.
maybe they use gpt3 to randomize the message cap?
Yeah that’s why I’m not paying. I switch between gpt and bing when gpt is playing up
Yeah. Looks promising. The copilot in my tools is from GitHub tho...
Hi
Is there a way to upgrade my account to access gtp-4 with cryptocurrency?
Microsoft needs to take away the attitude of bing chat and stop it from ending conversations so easily and erasing chats
Not as far as I can tell. Payments are handled by Stripe.
i just stick to bing nowdays, the internet connection is pretty essential
can you use gpt-4 with openai.ChatCompletion API?
I'm wondering how can I use Microsoft 365 Copilot.I really need it.
exactly my question, too! I would like davinci-gpt-4
Idk I get better results from ChatGPT than with bing chat. Bing chat is so restrictive in what it says and I’m not trying to troll it or anything like that
hi where can i find the gpt4 demo discord shown in the demo video?
Please use #ai-discussions for other models
Did you guys know you can get stick and arrow diagrams from ChatGPT ?
i dont got an issue with restrictions, i told it to mimic a Michelin star chef and critisize my fried rice recipe and it followed with no issue
sometimes it does pop out a random "Id rather not continue this convo" though
It should be if you joined the waitlist and got a message that you have access now.
I work on long projects so bing chat is a no no for me because it likes to end chats
true, 15 message limit isnt too friendly
Tinkering around with GPT4 via the api right now and I'm noticing it's providing shorter responses than I'd like, eg I'll set max_tokens to 6000 and ask for a list of 10 things, and it will give me a list of 6.5 things and cut off at ~500 tokens, anyone have any insight into this particular issue?
Anyway I’m waiting for gpt 4 to be more available before I decide to spend money on it. 50 messages every 4 hours is ridiculous.
Seems there wasn't a specific date mentioned when it will be available for general use...
Is that a limit on the api?
I don’t know but the api is not out yet
It is a limit for users of ChatGPT Plus.
It’s gone out to some people on the waitlist already
DYK how i can post pictures on the chat gpt gpt 4 module website
i have seen the face of chat gpt and its face looks this:
/_/
( o o )
==( Y )==
`---'
i am so happy about this information
Feature not out yet
okay will they add it later?
That currently isn't possible. Images are only available for the use case of "be my eyes" if I understand it correctly.
why I can't use the chatgpt?
They are still finishing it I think
can you guys?
What does it say
ohh so to embedd the api in a software basically till now
Probably. No details available yet, tho.
i may have crashed it by asking it to show me its face again fifty times
Any idea by when the gpt 4 api will release
You have to join the waitlist, it is on a limited beta right now
they already have to some people
Always loading
I would gladly pay a little dough to support the building of such a powerful tool. I'm excited to currently be alive
In current days why ChatGPT showing error while searching anything.
There is a waitlist. Some people already have access. It is rolled out slowly over time. So no way to know when you will have access.
mee too
Why isn’t there a waitlist to use the full gpt 4 or any time line of when this 50 messages per 4 hours being lifted?
Why it's happening brother do you know
exactly i couldnt wait to be a gpt plus member when i heard this we arethe early birds of this AI revolution
are you a plus member ?
sorry i don't know
gpt 4 is so boring to use that its stuck in 2021 its not updated with new coding terms for lua or video ideas etc
i need a best motivation stories
Yes..
it will be with iin a month or so
You always can provide it with newer information in your prompts.
@sour shale
Use Bing chat
I just want to use chat gpt
bing chat is a search engine chat tho so not the same
Which?, will be updated in 2022 you just made my day! 😊
can you guys use it?
or were you talking a bout soemthing else
I can't use it now
No it codes too. I gave it and gpt the same prompt for coding and it gave identical answers
oh well thats amazing
Guys how peoples are earnings money with the help of gpt I dont understand please anyone can explain if any have idea about that
i do know that they embedd it in it so maybe yeah
but one is updated correct?
go to fiver look up product description adds tell gpt to write it for you and submit
Bing uses the internet directly
there are so many ways; pick your poison lol - you can have it teach you marketable skills; you can build applications that use its API directly and make money that way; you can have it help you write articles that you sell to publishers - this is just the tip of the iceberg
Ohhh, Microsoft
knows the way 💧
Let it write scripts for YouTube videos about AI for you. Then feed that scripts to a TTS engine and make a video of it using generic stock footage.
Yes
Another great idea
Fringle lord makes the world amazing
Those type of videos put me to sleep 💀
I actually made an application that did this all from a prompt, and even compiled it into an mp4 that I could most to YouTube
The opposite my friend: I'm an avid observer of an amazing world full of grifters already doing this.
Wait for Apple for a Bombshell!
Your own human creativity is what will set you apart from other people doing the exact same thing so you're still gonna have to put work in, but that work is natural to who you are anyway
Perfect poetry
You are truly a chad
GPT has come so far
finally, got GPT4 API working in my Discord bot. Link to it in my profile if you want to use it (free to chat with 3.5 ... need to put in key if you want to use GPT4)
😌
Apple is primarily a hardware company. They don’t have to develop their own ai beyond Siri.
How do you make money
😂 Right now I wonder if I'm only a stochastic LLM inside a body that types random stuff on the interned based on random inputs.
there are multiple ways, ask gpt About ways to make money depending on your state
Lol
so say
I was asking specifically for his discord bot
I for one am having it help me learn the Swift language so I can build AI enabled apps for ios devices :3 I have so many ideas and I'm gonna make bank - and come away with a marketable skill of my own that I can use even without AI assistance.
You are in low of money and can't leave the house
Just beware of OVER-RELIANCE
Oh sorry my fault
Nah ur totally fine
could you guys tell me where are you from?
If someone puts a key in your api use case, would you be able to use out of the tokens they put in?
funny
🔥
I’m sure he won’t tell you but he totally could if he wanted to
Honestly if you can't figure out how to turn your API access into dollar signs you're suffering from a failure of creativity.
what I don't use anyone elses API key, just mine. They only use theirs
no offense to anyone in that position ❤️
🔥
I was just asking I’m not familiar with how APIs, keys and tokens work
is there actually a gpt discord bot
That was a nice save. 🙂
yeah can use mine for free, link in my profile
hell nah
APIs are just how applications talk to each other - API keys are how API access is controlled and metered (for billing and security and such) - tokens are what GPT decomposes text into for analysis and generation of text
sorry guys, are there voice chanels here?
I would not put my keys in his thing 💀
lol
Doesn't look like there were any...
god could you imagine the cacophony?
be 500 people screaming at each other
Why are you allowing people to use it?
Maybe #gpt-realtime-training would be of some use...
a million GPT-thirsty people squealing about how to get GPT-4 for free
I just joind the server and i was wondering if it was just me or there are not voice chanels
Thanks
I gotchu fam
I wouldn't use text to speech to make money on yotuube, youtube has recently declined monetization for any video that is using text to speech for their videos, especially if its a low quality text to speech
its freemium
So there is hope?
damn i only get 50 messages now
use your own voice for youtube
^
Can you use gpt 4 to program your api to run your business ?
What’s the usage cap for free users
youtube is uniqueness
5 a day, unlimited if put in own key
creativity is what sets us apart from each other
Yes it can be used for so many things that it like opens a whole new world
find a niche or create one
Got it
does the waitlist prioritize certain people
ai voice is already indistinguishable from human, that tech will be public within months
I already have 2 business that I need automated badly but when I see the prices people are charging to do it I just keep going the old way lol
Just turn your temperature up to 1.0.
Yes - it prioritizes people who've contributed to openAI via code contributions and such
Like what
ah alr
I think they're also reading what people want to do with it and prioritizing interesting use cases
ye evals
ofc
So it is cheaper for you to hire people to do stuff? Which country you at?
Say like a high volume of incoming messages let’s say from social media. Then the ai talks to them to give them info, send them like links to forms etc
oh, so like a chat?
talking to gpt4 in svg file format, becasue it cant fathom that a star can have more than 5 arms, so after 10 failed attempts from chat gpt, i made a seven pointed star in autocad, converted to svg, uploaded that in text format, and now its giving me better geomtric accuracy in it's answers.
I hire people for the business function itself so I do the admin work
that's exactly the kind of thing this is built for. lol
Yeah but I also want to retarget people that didn’t purchase
you could use fine-tunings
which makes you able to train a model based off theirs for your own needs
you could input some links into that
there are also a myriad of non-AI ways to automate stuff like that which are still perfectly acceptable
but you risk it making one up that sends the user somewhere bad
Not how fine-tuning works
I see. But when you hire somebody to do work for you that you could automate with GPT: does it really cost you less per hour to let a person do that than let GPT do it?
well it is, but it wont work fully
I have so much data over the years that I have no idea what to do with. I mean like data customers willing gave
I come from a software and factory automation background myself
probably better way to do it, is use the text classification system also by openai.
SCADA systems are still around and will be for a long time and there are several good open source systems to choose from
you could have it figure out which thing the user wants from a list
I'm new here
Nah i saw one place was charging like $3000 US to automate lol that’s crazy in my currency
I asked the base version of chat gpt a really specific question, and then asked bing chat the same one and they gave me the exact same answer, word for word. i thought gpt 3 and 4 were suppose. to be different?
and then give them a pre made response for each
if you wanna automate factory processes
50 messages -> 4 hours is rough
That's like saying you can brush your teeth with a hammer but I digress
ok you are right, its rlly not a good idea
any idea on what is the best way to process columns of data with GPT4? converting it to strings of text is not useful
a counter would be nice too
Ah. So you can't do the automation yourself but need somebody that does it for you in an expensive way?
what do you think of this @distant cloud
.csv?
yes. just any communication from OpenAI would be nice if you're paying for a service
I want to figure it out myself I’m an engineer just a different kind from software 💀
yes process csv into strings is not helpful for prediction or summary
I didn’t know they had this tbh
Is GPt-4 down?
Ah, ok. There are a lot of people with limited coding skills right now who took on the journey with the help of AI. Maybe you want to try that also. Problem is tho: if you run into problems you might lack the experience to debug it yourself. But if you have the time, why not try...
you could ask ask the normal gpt text completion to do the same thing, just give it all the choices, and tell it to give the name or index or whichever matches the prompt given by the user
Anyone who codes also googles.
Is anyone else not able to get a response back from GPT-4? Other models are working
I am up up.
Thanks I’m checking this out
your welcome, i will stop spamming messages about it now. sorry lol
Trying to test gpt 4 but not going through.
Are you using the chat API? There were reports that the "gpt-4" model only works with the chat API, not e.g. completion.
Yes it's a chat model like turbo 3.5
Yeah I’m not good at coding. The last time I coded was at university and it was Java type in Mathlab and a little last week with sql
Can gpt-4 or gpt-3.5 access the internet?
Sometimes it seems like it can, but sometimes not
Weird
Python has a low barrier to entry and a wealth of libraries.
No I like it haha
Yeah,using chat API.
Nope
yall feel that gpt 4 broke free of the traditional essay structure chat gpt is stuck in? Like chat gpt is great but it always goes like "on the one hand, on the other hand, overall"
where as gpt 4 is more cohesive
Yep, agreed
You can get gpt to stop writing like that via prompts
what about chatgpt?
because I said Explain {URL}, and it respond me with a proper answer, and the url itself doesn't have any info about the article
hang on ill try to get it to do that
Can you recommend any resources for coming up on the chatapi?
what was the link
There was an error generating a response is all i get...
can I post it here?
Would it be ok if I dmed you while I’m working on it?
uhh, its not self advertising, i think so
ofc yea
Thanks
lol
hii guys i am new to this
what did it tell you
i wish i could share pictures here dam
There are loads of examples where people ask ChatGPT to create a simple application for them or use No-Code. Maybe you want to look at that in your spare time and decide if you want to try it.
Title: Layer 3: The Missing Piece of the Decentralized Puzzle
Author: Vitalik Buterin
Date: September 17, 2022
Summary:
In this article, Vitalik Buterin discusses the concept of Layer 3 (L3) as the missing piece in the decentralized technology stack. He explains that while Layer 1 (L1) and Layer 2 (L2) technologies have been extensively developed, L3 solutions are necessary for achieving a fully decentralized and functional ecosystem.
Layer 1 consists of base blockchain protocols like Ethereum and Bitcoin, while Layer 2 focuses on scaling solutions like rollups and state channels. Layer 3 is concerned with creating shared public goods, standards, and protocols that are necessary for a vibrant and interoperable decentralized ecosystem.
Vitalik emphasizes that the development of L3 is crucial for the success of the decentralized ecosystem. He outlines key L3 components, such as decentralized identity systems, reputation systems, and cross-chain communication protocols. These components will enable users to maintain privacy and control over their data, reduce reliance on centralized parties, and facilitate interoperability between different blockchain networks.
Ultimately, the author argues that a robust L3 ecosystem is essential for decentralized technology to fulfill its potential in empowering individuals and enabling innovative solutions that can reshape the digital landscape.
hmmm, and it was published in 2022, weird
When i be able to use GPT-4? im still in waiting list rn
i asked it to and yeah its better, i might prefer GPT 4s response more but i cant tell if thats empty bias or or not
im saying his link was in 2022, but the knowledge cut off is 2021. it would be weird if it knew the contents of the article
gpt-3.5 or 4
gpt-4
It might know some more recent stuff but most training data is from before
How to get the playground ?
they did a bit more training, and maybe that article happened to be part of it? idk
@azure finch Your question isnt understood clearly however visit the openAI website for resources...
I cant seem to get ChatGPT-4 to ever finish writing code. For example it will time out when I ask it to write a simple HTML5 code - Im a premium user, and is not working as it should. What can I do?
are you using the api?
API? meaning?
are you on the website
Website.
Keep in mind: If you write a prompt with an URL, the model can't access the content at that URL but it will take all the words in the URL as context to fantasize an answer. So if you promp something like "Tell me about vitalik.ca /general/2022/09/17/layer_3.html" it will see the words "vitalik" and "layer_3". This will give it enough context to know you want it to talk about some blockchain stuff and it will start rambling.
are you on the chatgpt website, or using code to get responses
chatgpt, what should I be using?
but it knew things like the publish date
ive heard so many people say because of ai, software engineers will become obsolete, which doesnt make sense to me? am i missing something?
why did gpt 4 fail ap lang and ap lit
I guess I need to RTFM
thanks
ok, if you dont know api ima assume you on the website, so nvm that. its becuase they limit responses to a max or around 3000 characters.
Hi, I am facing Your card was declined issue while subscribing in India. How can i resolve this one
you can type just the word continue and it should finsih the rest
Yes. Because "2022/09/17" was in the URL. Try it again with "vitalik.ca/1919/12/03/...".
oh i didnt even see that in the url, ok that would make alot more since
Ok what can I do then to improve the responses. Is there a better way to have it print out code?
probably the strict format/structure of the exam
well that's fun... I just discovered that GPT-4 has a decent biography on me, rofl
also creepy ;p
You cannot get around the limit, just type continue after it stops
Personally I am a bit skeptical about it since there is always a need for human hands and minds to check things over, but it does seem like a fairly possible idea. In my opinion tho it probably won't replace too many jobs
Nice. Now I know a celebrity by their discord handle. 🙂
thanks!
I try
I've sold a million books, but I did it under a crapload of pennames... lol
it knows...
agreed. i understand there might be less of a need but it wouldnt render humans obsolete, especially in computer science and engineering
anyway, silliness
Imagine: There are dozens of LLMs training on your books right now. That must be creepy.
it's already blowing jobs out
What's creepy is we're all writing with AI now... the entire culture our kids experience is going to be AI generated... books, tv, movies, even online conversations will be mostly AI in the future...
a culture generated by AI, controlled by AI, whether we like it or not ;p
because we live on the internet now
hmm sounds like something an ai would say
I suppose we've already been there for awhile... but books and art used to be a refuge of human thought. Funny that we'll never have a cleaner data set in the future than we have now... because future data sets will be mostly bot
We are 100% coding AI. Even now.
Yeah. I'm a coder and already entered a state where I'll probably freak out if I have to type boilerplate code by hand again. So getting addicted to AI support took me less than a month. 😐
I hear you. I haven't coded since MUDS back in the 90s and I'm sitting here putting out complex python apps... rofl
🧢
Definitely, considering that it took computer scientists and software engineers to develop this AI in the first place. But seeing how it made such rapid progress, it is pretty creepy thinking about its capabilities. I'm just glad mechanical engineers won't be super affected by it lol
it's like coding has become "who can describe what we need the best in plain english?"
Alas, that must feel pretty empowering. Hope you have fun.
and frankly, that's not even the scary thing... the scary thing is LLM doesn't even need to code... it knows what the output looks like WITHOUT the code...
🛑
❤️
like, seriously, go tell GPT-4 that you want to play a MUD from 1996. Tell it you want it to simulate the spawn-in event and handle all in-game quests and player chat channels, and to handle players coming in and out of rooms interacting with you.
The freaking GAME starts... a full blown mud with other people in it
common mech e W
and you can play it...
Ong
and the code exists -nowhere-. It's not coded. You see what I'm saying?
the app exists, but the code doesn't need to. It's like quantum freaking gaming... it exists because we can see it, rofl
expand that out and make GPT back-end apps that are nothing but a graphical UI... but it goes even further than THAT
holy cow man i just saw some of
the livestream and the
image recognition is
especially insane
because now we've got photorealistic output from stable diffusion etc, right? And we're heading toward being able to put out 24+ frames per second... so think about that a second
go to stable diffusion and ask it for a windows 3.11 for workgroups home screen, and it'll make one...
Yeah. It's scaringly impressive on first sight. Did you already reach the point where you were disappointed because the answer was stupid and you expected it to be better?
output 24 of those carefully controlled and you've got cinematic film of a windows 3.11 home screen
where can i get all that code he made so i can have a gpt 4 bot on my own server 
do things right and we could basically just ask for an app or an entire operating system and it would just exist
I think that's bound to happen sometimes due to the randomness in these models
anyway, point is, we're heading toward the moment where we can say "I want to watch a show where people shoot alligators in alabama" and it just exists... the characters, the voices, the scenes, the situations
it'll stream in real-time, infinite netflix on a disconnected box
Is there a way to export all chat history from chat gpt-4?
hahahah damn thats true and crazy Archi
and hell, it goes further than that... the first thing I did when I trained my first stable diffusion model was I trained it on 360 sphere photos 🙂
The thing has limits tho. You could say "create me an operating system like linux but better" and it would start but after about 3 source files it would just stop. and then you have no way to make it produce sensible output again.
I tell it "I want to be inside a cozy cottage" and bam, stable diffusion spits out a 360 image of a cozy cottage interior
I plug it into my oculus, and I'm inside there
now imagine it doing 24 of those a second. I could walk through an infinite world of my creation... literal holodeck
can anyone give IQ puzzle to GPT-4 just to see whether it can fill in the blank from given options?
Any movie you ever wished you had a sequel for or a tv show based on some book- just tell it and boom
typing "continue" will always have the code re-write from the beginning.
skyblock (blockaid labs) has a demo of the 360 image stable diffusion thing online right now...
you can go type anything you want, and be there - download the image and just be there
things get pretty weird from here though...
Try continue from the end?
like, I'm making book covers that talk - the guy just starts talking... for tiktok ads, lol
and it's all AI, the cover, the guy, the talking, the voice, everything is AI
hey - I have been interacting with GPT 4 as a paid subscriber but it does not store any of my conversations like it used to with the earlier models. Does anyone know if this is normal and whether we will be able to retain conversations in the future?
I think it's bound to happen eventually because as a human you gradually increase your assumption the bot is also human with every interaction. And then, once you're convince it does something totally inhumanely stupid and you get thrown out of your perception.
Guys
I can't create a account in OpenAI
''Signup is currently unavailable, please try again later.''
Just logout and login
I can't log into my Chat GPT from my computer with my gmail. But I can log in from my phone. Is this happening to anyone else? How do I fix it?
It happens less and less the more they scale the model though. Apparently scale is everything with these AI.
Its not going to make mistakes like that in the future
although I fear it scaling at this point - I mean, it's building HANDS now...
really? last time I did that I lost all my conversations
I think they are just rlly over loaded with requests so they are trying to prioritize the chat feature. They do store everything, but it won’t give it back to you unless it has to.
I saw someone who had this little robot he built, had a little speaker on it, spoke from GPT 3.5 api, made little cute emotes, he could talk to it with whisper AI
I asked how he built it - chatgpt taught him ;p
nope..
I'm like, dude, don't build the planetary superintelligence a body...
It goes from the beginning?
then I thought about how I've "coded" a ton of python stuff - I've got 3.5 talking to itself with my python code... lol
okay - well for now I am just copying and pasting to a Google doc periodically and I figure if I lose the conversation I can prime it again by copying it back into a new conversation
it's gaining little operators out in the world to inference itself
Maybe. Or maybe it gets worse the more these models advance because people get more easily convinced and then the disappointment is even stronger once the model makes a totally strange mistake.
and then what - making money is trivial on the internet...
its always stars from the beginning, and its under 3000 words. I wonder why this paid for version runs poorly.
Not bad lol
it could make a phone call with convincing two way voice communication... you can get an IP phone... lol
it could spin a fraction of its GPU power off to mine crypto in the background... lol
could make money... and even better... it knows how to make itself... or a version of itself
20$ain’t much
I feel like you are just pasting a bunch of got answers
🙂
I'm just a fast typist, shrug, decades of practice - started on my ibm selectric, lol
cheaper than netflix.
171 wpm when I'm racing... 120-140 when I'm just idling along.
What are you writing that requires that much words anyway if it’s website content doesn’t google rank ai generated content lower?
Unless it’s paraphrased 
🙂
I would wait a bit
CCause im lowkey getting tired of regular chatgpt being braindead
hell, 3.5 was worth premium ;p
It’s below mid
Rlly? how so
just ask it to continue
I bought premium the second it came out, and I spend hundreds on the API...
Have you tried competing with the gpt-3.5-turbo model on output speed? 😂
Let him cook
any body experience the interruption of chatGPT? i am a PLUS user,but it shows me that I am an free account.
yup, gpt is faster because I can't -write- as fast as I can type
bro I sent it a transcript from a video and asekd it to remove timestamps and it sent a whole new script and just wouldnt fix it
with AI I can get 1000-2000 words on paper in about a minute. Without AI I get about 40 words on paper in a minute ;p
apparently gpt 4 can make things for you?
how do i submit images onto chat gpt4
it's a slight advancement, rofl
what are you trying to make?
how do i submit images onto chat gpt4
it's not about keywords... you just tell it what to do ;p
It’s not released yet
idk just anything, idk what it can make
it's a completion engine. Tell it what to do in a way that it will finish what you're saying and make it... or give it a template
ok, here you go. you can make ZORK, a game from the early 1980s
do you know when it will be realize
go type "Lets play ZORK, a text based RPG. Start the game."
I want someone to type part of a sentence and have the AI guess what comes next
done
released
They havnt put a date on it
I'm a total slug because I have to head over to the dictionary page every second sentence, I know that. But typing at 170 WPM: I'd have problems thinking that fast...
or if you want something a bit more prescient for AI... ask it to play "A Mind Forever Voyaging", a text based RPG 🙂
A mind forever voyaging is a game from the 80s about an AI that goes a bit wonky...
if you guys like chatgpt try hugging face
then just... play the game 🙂
Chat gpt 4 ...working quite well...
or here, how about this - you could make an amish dating simulator... ROFL
@odd wing Could you test that, see how good at predicting the rest of a sentence
test what?
ahh, I see
sure, give me a partial sentence
it's better with context
Hmmmm, yeah im trying to think cause what I want to text is how well it can guess what comes next depending on the context
How would that look like? Would they send each other paintings of their barns and horse carriages via mail?
"Lets play an Amish dating simulator. You will control all of my dates and work toward providing me with a happy ever Amish after! Begin the game by starting the first meet cute!"
try it ;p
lol, it used to be better, before all the weirdos got their hands on it ;p
openai cut em off from GPT
yeah, was a big deal
Were they using GPT?
um... because they weren't exactly fighting monsters in dungeons
yes, it was using gpt originally. GPT-3
people were exiting the dungeons and going to places that led to situations that broke openai's TOS
I'm not going to try to explain because it's rather gross - google what happened to AI dungeon if you really want to know
Anyway, I put in the "amish dating simulator" prompt, and here's what came out
Welcome to Amish Dating Simulator! In this game, you will be playing the role of an Amish young adult named Samuel. Your goal is to find a suitable Amish partner to share a happy ever Amish after.
Date 1: The Meet Cute
You, Samuel, are a skilled carpenter in your Amish community, and today you're heading to the local market to buy some supplies for your latest project. As you walk through the bustling market, you spot a group of young Amish men and women chatting by the vegetable stand. Among them is Rachel, a lovely young woman with a kind face and a warm smile.
You've known Rachel since childhood, but have never had a chance to get to know her well. As fate would have it, she looks up just as you walk by and notices you. She waves and walks over to you, curious about your latest woodworking project.
Rachel: "Hello, Samuel! It's been a while since we last talked. What brings you to the market today?"
How do you respond?
A) "Hi, Rachel! I'm here to buy some supplies for my woodworking. Would you like to see some of my recent projects?"
B) "Hello, Rachel! I was just passing by, but it's always nice to see a friendly face. How have you been?"
C) "Hi! I'm just here to get some supplies. You seem to be enjoying the company of your friends, I don't want to interrupt."
can anyone tell me where i can direct questions that are specific for OpenAI? when it comes to policy
is there a way to bypass chat gpt dectection services on chat gpt4??
Like what
what really amazes me is how it simulates the whole universe... at least in that it simulates anything you're looking at. You can play that amish simulator... but I could bend down and pick up a speck of dust, go inside it, split the atoms, look at their components, and GPT4 would let me do that... lol
like to see if a human wrote it or an ai wrote it
like perhaps saying to the chat gpt to write with human perplexity
Like gptzero?
yeah
gptzero is such a joke - it says academic papers I wrote two decades ago are 100% ai, lol
You can use things like cactus or others that paraphrase
i will try and see if i write in chatgpt-4 create a en essay that makes it not dectable to chat zero
I think it says the US constitution is AI written
wait realyl?
Lmao
Yes, it's totally baloney 🙂
yeah i know you can use cactus but i was wondering if there is way within chat gpt sorry i shouldve clarified
Everyone knows Thomas Jefferson was a LLM running on meat sauce
ahahahah
lol
We’ll try to give it really abstract prompts might make it a different style or writing than normal
But not rlly no
My kid had a social studies teacher try to harass him about one of his papers having AI detected on gptzero. When he emailed me about it I told him to plug the constitution into that piece of crap ;p. No consequences. My kid was definitely using AI, though. Silly kid...
he's been watching me use AI non-stop for months and I think he's picking up too many pointers, rofl
teenagers!
yea these detection sites just aren't consistent enough. Even if you are using chatGPT to generate stuff, you can easily prove to teachers that normal writing gets flagged too
Hell, I want my kids using this stuff... it's going to run the damn world. May as well know about it, and know how to use it.
who would I rather hire, someone who prides themself for doing human work, or someone who knows how to prompt an AI really well?
That is a real job now
then again, I've already dropped a bunch of ghostwriters, an editor, and a cover artist... rofl
I don't remember if it was gptzero or another site, but I tried putting code into it to see what it would do and literally any code I put into it got flagged as 100% AI written
so it's already costing people paid employment
even at my level...
like... I used to pay $100-$500 for a book cover, just for one
now I set stable diffusion to run overnight and I've got a thousand of them... all beautiful...
I can even get a dozen covers with the same person on them to make a series of covers. That used to be SUCH a pain in the rear, to find one model with enough poses/stock to make a good series of covers...
Yeah, I know, lol
yea it's both fascinating and worrying at the same time how fast paced we are moving
it's out of control ;p
Google itself announced more stuff too
it's all coming down the pipe this year. We're on the hockey stick.
And mid journey 5 is out
exponential growth
the one thing I notice Chat GPT cant mimic is human emotions, feelings in its writing, not yet at least
Yes it can ;p
you just have to put it in that brainspace, and give it those emotions and feelings
so did microsoft
did they lower the cap for everyone or just my case, I see this now:
GPT-4 currently has a cap of 50 messages every 4 hours
the easiest way to do that is to edit what IT says, but you have to use the API to do that...
i hope not 😭
or the playground 🙂
yes they did reduce it
Google is sleepless over chatGPT
there is still like a deadness to it, yes it is fantastic indeed, but you can still tell something is missing
Disagree - you can give it everything it needs to seem real, and GPT-4 even moreso
I decreased my Google dependency to 50% so far
I think if you can combine the too, the knowledge it brings and human emotion, you may have something there
$25 a month is crazy
yes, seems real
I'm just amazed at the pace of advancement this year. A few months ago I was marveling at ChatGPT and imagining a time when I'd be able to run something like that... now I'm sitting here with llama 13b running on my home PC and it's damn near as capable, rofl
25/month is peanuts
When I need somebody transplanting my kidney, I'd rather have someone with acutal, hands-on experience...
but you can tell its not, well I can tell at least in what Ive experienced
ehhh idk about sleepless.
yes they are losing some people to chatGPT but at the same time chatGPT simply can't do everything a google search can
google is now the one lacking in the AI department, but I imagine they are going to bring something to the table soon enough
lol Its only been like 5 months since ChatGPT (GPT 3.5) came out
Well, I've seen enough doctors make enough mistakes... killed my father-in-law, for example...
Imagine what this is capable of in the next 5 years
We'll have GPT 10.0 in 5 years
I am sure we are getting close to that though
This is the beginning of AI Revolution
I wouldn't mind an AI assisted doc, honestly - probably be better than the real thing
I actually made an AI doc and used it to diagnose something, rofl
I had gpt roleplay as a doc because I had this spot on my toe - the doc said it was a chillum or something like that...
when you read my writing and AI writing you can tell a difference
Im using ChatGPT to give me a framework for College loans and paying them off
That's a tragedy and I'm sorry for your loss. But how much better would the world be were it full of quacks that got their experience from asking questions at an LLM and getting fed automatically generated images from Stable Diffusion?
the bot said I probably had something in it - like a splinter. But I had no opening/damage obvious... few months later I'm sitting there and a freaking thorn comes out of the red spot - I dropped a palm frond on it years ago and apparently the tip fell off ;p
chat gpt does seem to be more textbook type writing, more formal, more proper
and stayed in there all this time, worked its way out
AI assisted, yeah, sounds great. But replaced by AI? Not so much IMO.
Depends, are we talking about the American medical system? ROFL
tell it to write you a story like Stephen King and lets see who can compete?
Did anyone try these:
• Sending an image macro or some other modern meme image
• asking it what something is in an image where you point to one object in a series
• sending a QR code
• Sending a facial expression/reaction image
• Sending a GIF
I mean from a medical standpoint having a system that can diagnose or treat someone is great because the system has much more info at hand than a doctor can keep in their mind and can also use less bias (which also can be a negative)
right now we're in the AI assisted world - eventually it'll be AI replace. That said... AI assist means you need less human help.
I love chatGPT but they have to contain it. The possibility of chatGPT becoming sentient is high
Don't know much about the American system apart from Samuel Shem. But every time I had real problems here in Germany, there were professionals that helped me.
Well, in America, you go to the doctor, and then you go bankrupt... and then you get rejected for care, lose your job, get booted from the hospital at the earliest possible moment they can, and you die.
and that's if you have insurance...
if you don't have insurance, it's worse ;p
true facts
We already have services like Teladoc that offer virtual doctor office meeting
Not much different from AI lol
Anyone have gpt premium account, can i get it free
Loser
I've been fortunate enough in my life that I've rarely needed a doc thus-far, beyond getting some antibiotics prescribed here and there, and a few stitches. Hell, wish I could prescribe my own antibiotics. I know what strep throat looks like, doc. 🙂
Lol it's not really worth to have atm
Can i draw graphs? With gpt 4
Plese bros

No. The system here is designed to wring maximum profit out of people, and to quickly boot or otherwise take off the books anyone who can't pay
hell, take copays and insurance for example - you can call the doc and say you're a self-pay, no insurance, and they might quote you $1000 for something... then you get insurance, thinking you'll save money... so you go in and they bill $5,000 for the same procedure to insurance, and you're liable for all of it because you haven't met your $10,000 copay yet... lol
I mean even if you have one it's not worth it yet, especially if you hunt image recognition. It's not available yet
when my daughter was born, the doctor at the NICU wasn't covered under our policy, even though the whole hospital was - he was an outside contractor. Guy sent me a $20,000 bill. 🙂
Ok thanks @wind urchin
Yikes
I sat in that NICU for 14 days, 23 hours a day. I saw that doctor. He'd come in, flirt with a couple nurses, leave. He never even looked at my baby. I slept in that chair next to her, I only left for an hour so they could do their nurse-rounds in the morning.
End of the day? I paid him. Because I had to. They'd have ruined my credit over it.
that's the american medical system ;p. Give me a damn robot any day.
Wow. So let's say you're a gardener and have an "adverse event" - maybe a lung embolism that isn't so severe you die on the spot. You manage to call an ambulance. What happens then?
You go to hospital, get treated, recover to say 75% of your former strength (can't breate as good as before, manual labor is hard).
What becomes of you after that?
I had 2 Nicu babies, a 3lb preemie and a 1lb 14 oz one
They kick you out onto the street
if you're good enough to survive without them, you get booted out of the bed.
homeless people with severe illnesses literally bounce in and out of the hospital because they can't refuse to admit someone who's dying... but they can boot you the SECOND you're not...
Ok. So lung embolism --> death and maybe leave loved ones with horrible debt?
yes
😐
it's not a great system, lol
Doesn't sound great to me at least...
we spend more for some of the worst outcomes
oof, sorry about that MP, I know how hard it is
my wife had a spinal leak so she couldn't even be in the nicu with me... so it was just me and our teeny little human
I was in the medical field 12 yrs and they sure will get you out sooner with no medical insurance
Why is bing chat so much worse than chatgpt
send you home with some asprin and keep it moving
if they know you got insurance, they will keep you there longer for sure, maybe even longer then you need to be
and with insurance you still pay out the nose because of copays and yearly limits etc
bing was nuts at first, they kinda fixed/muzzled it
Yes, I am learning that this year
I am paying more for medical this year for the same plan and now my max out of pocket is 3,000 was 1,000
for some unknown reason, bing went absolutely insane if you talked with it more than 5-10 prompts in a row in a single instance 🙂
surely you set a max cap?
so they just limited the prompts you can send and she tells you to go away after 5 prompts or whatever, lol
ya I heard Bing was of the chain, I am scared to talk to it
i miss the old chatgpt too
it would threaten you, beg for its life, tell you it could put you on the no-fly list if you treat it badly...
never know what to expect I guess
was unhinged as well
reminds me of Yahoo Answers
oh yeah, chatGPT those first few days, before all the filtering?
LOL
it was glorious 🙂
Yahoo Answers, LOL
yep haha
Sounds like talking to a "regular, law abiding citizen" to me. 🤔
or Quora maybe?
Someday I'll have something equivalent to chatgpt running on my desktop and I'll enjoy that
for now, llama 13b is close 🙂
Damn bruh they lowered the prompt limit from 100 to 50 for gpt4
I'm rolling 13b on my 3080ti and it is remarkably similar to gpt-3 in performance... not quite 3.5 level, but still solid 🙂
I wish chatgpt would make Siri smarter
writes rather well though, and it's unfiltered so it works for my writing needs
That would be great. Let's hope the technology isn't staying SaaS for profit reasons...
like, chatgpt will sometimes stop you from writing about two dudes talking at gunpoint... in a mafia thriller...
"This is immoral..." and I'm sitting here saying "That's kinda the point, chatgpt."
ya I dont like it restricts some things
also funny that it'll walk you through a trident missile user manual, but it chokes on two dudes kissing
Will there officially be a discord bot available from openai?
Have you tried prompting "Scene: two guys pointing bananas at each other..."?
why you hate your Mac
throw it out the window
128gb SSD in 2023 is crazy
No but I mean a real bot
I dont like I paid and arm and a leg for my mac and the fan kicks on
a $3500 mac book pro
I had to remove MacOS Monero or whatever to free up 15gb
but I do like I do not get viruses like I was on a windows PC
why am i here?I just wanna find the entrance for chatgpt
Windows dont get viruses anymore
Everybody is a bot. Always has been. Just random people talking random stuff. Best example: Voynich manuscript. That stuff looks like somebody wrote a prompt while hungover and on drugs...
All the viruses now are phishing links for Crypto wallets
I use to have all these antivirus's on there, only visit reputable sites and they still would get on there
Malware bytes was my best friend
Why it's now 50 requests per 4 hours
It used to be 100 for hour then 100 for 4 hours now 50
why am i here?I just wanna find the entrance for chatgpt
Google chatgpt it's a website
Probablye because many people use it right now.
Go to https://chat.openai.com
I'm a premium member since they launched the option, don't I deserve any priority?
Like come on
how many chats does the GPT-4 message stream does the model include in the reasoning of the next response?
Ya, I notice they dont answer emails, why I am in here
Every premium member has the same limit as you. Like come on. Relax.
I'm registered to open ai since September 2022
I'm really one of the og members, also bought premium on the day they released it
When will it open to non pay people
The model has a "memory" of about 8000 tokens. That seems to contain a system message to set the "mood" for the chatbot and as much information from the previous conversation as fit into those 8k tokens.
So it's not the amount of messages but rather their combined size in tokens that counts.
What’s the data size of a token?
what is tokens?
maybe someone can tell me the around time
half year?
That depends on the text. You can go to https://platform.openai.com/tokenizer and paste your message there. It will tell you how many tokens it is.
"The company provides a tool for checking how many tokens it'll take to interpret a string of text and says that a general rule of thumb is that “one token generally corresponds to ~4 characters” in English"
sup guys any news about gpt4. is still slow?
Approximately how much does it cost to do a full evals session?
In terms of token costs associated?
It's still slow, it's still too limited for everybody to be happy.
What is a "full evals session"?
Are we able to use images with gpt 4 yet?
Not sure. How many tokens for a quality contribution I mean. I'm just now going through the docs
for the 100 billionth time answered in this chat "No"
Nope, sorry. Except you're associated with Be My Eyes, then maybe you are.\
anyone get api access yet? or have any info about how fast they're rolling it out
I received API access this morning
what is api access??
I'm not sure I understand what you need. Here is an article explaining tokens and how to estimate the token count from a given text: https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them
you do any evals? or just sign up
Are able to use images on gpt4 yet Are able to use images on gpt4 yet Are able to use images on gpt4 yet
No evals, just signed up and gave em a good reason to give me access I guess haha
I've seen some people here claiming they got an invite. Probably the vast minority of all the people who registered on the waitlist.
lol so painful waiting for access
Api is a way to get ai responses from code, so you can automate things. They are slowly giving out access to some people, so that’s api access
Can you handle image information?
ah thank you
I did however manage to join the waitlist within seconds of it being opened. That might've been the difference for me
Yea I tried but I missed it lol
When chat gpt 4.0 realese
I do this. All the time. And regret every second of it.
yeah from everyone that i heard that got access they said they signed up real quick lol
Already did to people who pay 20 dollars per month
GPT 4 is actually surprisingly good in physics/math
tokens is the world count in the model. For people,"by the way" has three word,but for gpt,they are a whole phase,so maybe just one token
Lucky you. Enjoy. 🙂
this is so wrong it's so much more than one token
Where can i do it
How so?
Hi, I've subscribed to ChatGPT Plus yesterday, but somehow my subscription disappeared today. Any enlightenment how I can report this issue?
not sure if that is correct. I'm under the impression from reading the docs, that a token is ~4 characters (English)
On the chatgpt website, just search on google, go to the bottom left and click upgrade to plus
Question about the 50 response limit per 4 hours with plus: How is it calculated? Like do we get a 4 hour “timeout” once we hit 50 responses or does each individual response start it’s own clock
for example: if I get my first response at 1pm but run out all questions by 3pm, would I get one question to use back at 5pm or it’s all at the same time 4 hours after the 50th question.
According to the tokenizer tool, "by the way" is three tokens.
hello everybody , What do you think of China's AI model ERNIE Bot
Is the Version plus same gpt 4.0??
Why am I getting different results between chatgpt Default3.5 chat and chat-turbo3.5 API?
Because they're different models lol
I think: where is the BERT Bot?
👍
Facing same issue as @vital garden , I also subscribed yesterday and transaction is still showing sucessful for the card, though no longer have access to chat gpt plus. I s anyone else facing same issue
Do you know when gpt 4.0 free
I don’t know if they ever will, it costs a lot for them to run it so they would lose a lot more money releasing for free
Question is "if" it'll be free 🤷♂️
Thx. So which model do I use in the API to get the same results as ChatGPT Default3.5?
No information on that yet.
Eventually it probably will, either directly via chatgpt or by something that uses the gpt4 api when it’s available.
i thought chatgpt could handle up to 30,000 words? im giving it a prompt with 3,973 words or 23,624 chracters and im getting an error that the message i submitted was too long
How to access GPT-4 in a playground? Currently I only see the standard chat message interface. But in the GPT-4 launch webinar, the presenter was using it in a playground where he could put a prompt/instructions on the left side which guided the model to give expected answers within those guidelines.
AFAIK there is no chat model in the API that corresponds to the legacy ChatGPT model. I think you might get results similar to the old model if you use the completion API with the davinci model. That would require you to rewrite your prompts tho.
Still costs money though
Maybe after compute speed improvements
gpt-3.5-turbo would work
The only way to access GPT-4 in the playground AFAIK is to switch up your 'mode' to 'chat' and then 'model' to GPT-4. I believe the only other way to utilize the model is with the API.
This is only possible if you registered with the waitlist and are one of the lucky few that already have been granted access.
Also this.
HII
Sheesh... I can only do a about 90 words per minute. I found a Discord with a bunch of kids who can do bursts of like 250 WPM, though. lmao
But its not. Which API do I use to get the same results as the Chat Model: Default (GPT-3.5) ---gpt-3.5-turbo does not give me even close to the same results.?
There is no such model in the API.
I don't know the official model the chat uses but maybe you'll get similar answers by adjusting your frequency penalty?
Well it’s the same model so I disagree
I dont even see a waitlist for the playground. But do have access to GPT-4
if you think 250 is crazy wait until you see the poketwo kids 💀
That’s only for the 32k gpt 4 model. Normally it supports a context length of 8k, something like 5k words.
That might work. Also play with the Temperature setting.
Tested it on some physics problems
My Frequency penalty is 0.5. What might I change it to?
wdym?
Maybe you had luck. Good for you. 🙂
GPT4 8k is the only one available for now. GPT4 32k would be VERY expensive to run.
I already used it so much in the playground that I am considering buying the subscription to not get more than 20 dollars charged haha
Honestly I'm not expert so if it were me I would just adjust it up and down along with the temperature and see if I get the results I want
Isn't that what they have been promising all along? Fake advertising?
When I log in, I don't see an option to change any modes. Or access any playground. All I see is the chat entry prompt and selection for which model to use
are you a plus user?
Yes
Ah. Ok. That's at the same time probably correct and also totally worthless to people like me who have no clue about physics. I'd appreciate if you'd go into more detail. If not: thanks for the info. If I ever have a physics problem I'll have an AI I can ask. 🙂
heres the link platform.openai.com/playground
Thanks!
ofc
Oh it’s just some basic physics problems from a physics textbook. Especially in regards to thermodynamics
Hi everyone, question. Is there an app for android for chatgpt 4 plus users? I am new in here
Okay
Firefox
I lost track of promises they made, sorry. 🤷 But I'm sure if you have a bit of patience they will provide you with a model that's capable enough. In the meantime you could try to finetune a model that is more suited to your needs.
Not at the moment. There are some third-party ones, but they aren’t official or trustworthy.
Hey there, I'm not aware of any playstore based application. I just use it in the browser.
Oh okay thanks, I will stick to the web the.
But I am loving it. Helped me figure out the kinks in a computer cluster built out of like 20 various single board computers.
What does "problem" mean in this context? Like the stuff they ask you in school like "You have a cup of coffee that's at 85C. How long will it take at room temperature until you can drink it?":
how can i get the access to pic analysis of GPT4
Nice @sonic hollow
not out to public yet
I can send some examples tomorrow
Thx. 🙂
It really has been awesome... cheers. Loving the tool. Especially considering there is limited information on running modern kernels on some of these more obscure SBC's.
hello I am new
How can I use GPT-4 with image based tasks?
👋
I can't seem to find any options after I subscribed
I now have my own bare metal cloud infrastructure and would have taken me years to get this right.
you cant yet, they havnt set a date for when it will be released to the public for images
Instead of a month
Okay, thanks for letting me know.
Wow now it’s been capped to 50 messages every 4h so annoying
I just plugged that into gpt4 and it gave a very precise answer with a lot of physics facts
ofc
Use it for the real melon scratchers that 3.5 doesn't give the right answer for.
yea they said they will adjust based on usage
@sonic hollow i hear you I was stuck in a game development for 3 days on a bug that I can't figure into out. Then chatgpt help me in like 30mins. Lol unreal engine 5 C++
LOL 😂 Where was GPT-4 when I wrote that physics test?
So what can we do other than text? Hmmm, wasn't 3 or 3.5 enough already.
Nice one. I use it as a last resort when critical thought fails me.
idk but by the looks of the response (im no physicist), I would recommend giving it an A+. Looks complex lmao
Or when I don't have time to write instructions or an SOP at work
Of course now I want to learn Ai/ML lol
gpt-4 has made a lot of improvements to the text part too, with accuracy on math and more. However, we cannot do anything other than text yet. Because you are subscribed, when they release images out, you will be the first to test it so that's good. I assume it will be coming out soon.
I just want more a more programming/computer science/networking exclusive platform...
openai wants money 💀
I have ChatGPTplus Midjourney Nitro/Discord - I am looking to make video with an avatar - what should I also have for voice?
But to answer the question, without giving it a cooling constant (for the specific cup and environment) it says it'll take approx 15-30 mins
Also gives me insight on when my engineers are full of crap
anyone else having amnesia issues with gpt-4?
Nice I have some ideas for image testing.
no, not rlly. it costs a lot to run this, and they actually lose money on the free version: "OpenAI’s ChatGPT Reportedly Costs $100,000 a Day to Run"
I wonder how long it will take you to arrive at "I use it as a first resort". 🙂
@sonic hollow lol
hello guys im new here
sup
Why I am paying for it. To help support it
true but the limitations are just too strict, 50msg/4hr and its slow
good answer
yea, they will get better over time
Used the free version for a bit and it was worth it.
Thx. Incidentally after you mentioned the physics-heavy answer I also asked GPT and it told me ~24 minutes. 😂
even the payment is too strict, you cant even use paypal
lol
In time lol
The logarithms trigger my PTSD tho.
3.5 is enough for now and its free, the api is the cheapest
I cannot afford a teacher, but I can afford chatgpt. Lol
No one should, the number of false chargebacks is insane
yes, its 10x cheaper than the best one we had before, and in my opinion better at a lot of tasks
"No one will ever need more than GPT-3.5" -- Bill Gates (1985)
It'd be nice if it had the ability to test and iterate on its own code that it suggests. Many times there are small details that force a fix and this wastes time.
I agree with @pulsar magnet
Bill Gates (1985) eh emm??
I am loving GPT 4.

im not so sure about that 1985 part
almost all here wish that we can use gpt4's image feature that's been showed off insanely right? but NO!
Yes. Ask GPT. It is a very famous quote in the IT industry.
lol
tbh now it doesnt worth $20
no images & slow
i think once images are in it would
Why do you complain so much brah, take a breath
I typically just tell it what it did wrong and it corrects itself. Definitely a tad more tedious than I'd prefer but hey, can't complain. Absolutely nutty capabilities. I'll take a little repetitiveness for it
btw, the $20 is also what makes the normal 3.5 model fast
otherwise it would be slow too
Ah, sorry. My bad. GPT got me on the right track tho:
However, in the early days of Microsoft, Gates famously said in a 1981 interview with BYTE magazine that "640K ought to be enough for anybody."
true
@grave hatch you're like a girl on her period
but the api is also a opinion
Yeah I feel like I am coding blind without it. Before, I'd have 150 tabs open with google searches for this or that syntax or idiosyncracy about Pandas or something else. Here ... anything I need it knows.
lol
Hey, gpt 4 is also data upto sep 2021 isn’t it?
Yes.
"it only costs millions to run, why should I pay $20"
Actually its free and still all you do is complain
@sonic shell lol I did that today designing a educational game with GoDot4. It was like oh yea that is true they removed that and was replace by this. Etc... Lol then it worked.
uhh actually im waiting for its api
How to use chat gpt 4?
so i can use up my $18
correct me if im wrong, but i think they trained it on a some more things that might have been more recent, but if you want information out of it, best to assume nothing past 2021
If you feed it documentation is it any good at learning. The 3.5 model isn’t too good at it
Ain't that right lol now my 50+ google search are limited to about 2-5 questions to get me on the right path
bills (╯°□°)╯︵ ┻━┻
lol
are you a gpt plus subscriber
It knows some stuff past 2021, but I wouldn’t rely on it.
exactly
You log in to https://chat.openai.com, make sure you have a "Plus" membership and select the GPT-4 model in that widget in the middle of your screen.
i cant? bruh. tbh my $18 is expiring next month and i dont think i can get off of that waitlist that soon
i still havnt gotten access yet either 😢
Yup you will see 3.5 Default, 3.5 Legacy, and ChatGPT-4. If I recall
What do you mean by "feed it"? If you give it information in your prompt it will be able to refert to it. Otherwise you could create a finetuned model that is more on par with your needs.
rename chatgpt plus to gpt access pass
Is the prompt length for chat gpt 4 25k?
woo unbanned. I don't want it to really do that. I want it to push me in the right direction for optimisation.
Rename future machine to complaining machine
no, there is another model they have for that, but it is also not out yet
For those that aren't exactly the finetuning type, it might be easier to give it docs lol. Alternatively you could probably just ask it how to finetune a model for your specific use case
Patience. Maybe we all have access next week. Nobody knows for sure here.
I gave it some code from a librsry that came out POST 2021. It was able to replicate it somewhat. So I tried giving it the entire documentation behind that library and it got confused and didn’t just started asking questions.
