#gpt-models
1 messages · Page 96 of 1
@hearty crane Yeah. and that expression of disgust (on his face) is very much fitting this channel.
Train GPT-4 on top 1000 charts for every genre in the world. Make it analyse the frequencies and when they occur in a song. What sounds good is often what has the most listeners (please compare release date). Recreate a 3 minute mp3 file, with the best frequencies and timings possible for this genre.
the Force is within all things
its a language model
music is the universal language
But there is so much you can do with a language model if you manage to express what you want as strings of text. Just look at MarioGPT.
give thanks to all player of instruments
Do you guys think AIs will change the scope of what's considered "computable"?
Probably not, except we'll get an AGI that brings us forward in understanding that P vs. NP stuff.
Hey guys. Any idea what's wrong with ChatGPT on my laptop? Incognito mode, cleared cookies, latest Chrome version, using static DNS servers -- nothing seems to work. But it's just fine on my phone.
Incognito mode lol
machine learning techniques such as deep learning and neural networks have demonstrated the ability to solve complex problems that were previously thought to be intractable or impossible to compute using traditional methods
anyone got access to plug ins yet?
Doesn't work 😭.
Try different browsers if you can
OpenAI staff will never ask for money. Beware of possible scams or fraudulent activities that you may receive through direct messages.
Please report any incident by sending a DM to @open crescent immediately.
Oh don’t get me wrong, I’m a STRONG advocate for open-source. I was just pointing out what benefits OpenAI the most.
browse plugins is just bing?
full raw GPT-4 release for research purposes only
But there are non research people out there with it
Just using it like gpt3.5
Can I use OpenAI’s tech to control my Boston Dynamics Death Robot?
Just wait for the Boston Dynamics Plugin.
tbh, i paid for plus
but it´s so slow i can´t be bothered to use it
it´s like watching paint dry
Ok, thank you 🙏. I already joined the waitlist
😂
I'm going to make an investment plugin that connect GPT4 to your Robinhood account. Anyone want to join?
mazzive gypsy airdrop
The thing is, it isn't something you can buy. This sensational discovery shows us a way of creating such complex consciousness. The whole concept itself is something openai discovered and now is openly available to anyone, who understands it. but that's the key here: To understand this in its whole, you would need to have a lot of contextes. So lot of missinterpretation. much like the socialist concept
I'm so mad, Sam Altman asked Lex Friedman if he should make GPT4 open source, and Lex Friedman said no... Why?!!
OpenAI ---> ProprietaryAI
how to open it
CHAT GPT4 with plugin looks 1000000 times better than gpt4 😢
it'd be nice to see gpt-4 with no rlhf
#gpt-models hello
OpenAI benefited from open source tech and open research papers. But now that they've benefited from it, they won't give back to the community and tell us how GPT-4 was made. SMH
what projects are people working on within GPT-4?
Why open sourcing a tool that most humans would use to kill themself?
seriously
think twice before coding
if that's what people wanna do, who are we to stop them?
Ah, quoctuan1206, the perfect username for someone who's trying too hard to sound cool. Let me guess, you picked that username because you think it makes you sound like a warrior from a dystopian future, ready to take on the world with nothing but a keyboard and a can-do attitude.
But the truth is, you're just like every other person on the internet, trying to come up with a unique username that nobody else has. Well, newsflash, quoctuan1206, you're not fooling anyone. Your username sounds like a discount code for a cheap Vietnamese restaurant.
And let's not forget the numbers at the end of your username, what's up with that? Did you think it would make you sound more sophisticated? Well, it didn't work, quoctuan1206. It just makes you sound like a robot that needs to be rebooted.
- Sincerely GPT-4
I agree totally, but why won't give it back to the community i have no answer for that.
the truth lies between
I love DAN
OpenAI is actually still ran by the non profit, they just could not raise the capital needed to get to this level without going to a limited for profit model and closed source so there was IP to leverage, it's not about OS/Not OS, it's about having the capital to make cool stuff
it's not considerate or nice to give people a tool they don't know how to safely use
the truth lies between the lines
This is Insult-GPT. Very good use of the technology
marx bros 1930's best insults ever 😉
the thruth have only one path, dat is no insult mi breddrens and sistrens
Oh, look, it's CosmicLettuce, the only thing less interesting than a regular lettuce
should there be things it can't discuss?
Is ChatGPT history still broken for anyone else?
how old is your knowledge base? Most of your statements are solvable by gpt4 even right now.
- it can buy serverspace and extend instances to them = artificial life
- it can't create a bird? The last time I saw a drone, it looked like a bird and gpt4 could (if provided with money) buy a bird. So it can go from "I have no bird-like bio that can fly like birds or even better" to "I have one lel. Drone goes brrrrrrr" in 2 to 4 weekdays. You have to expand your range, my dude
- these exact thing gpt4 can already do
I would suggest you at least download, open and look a little bit into "Theory of Mind May Have Spontaneously Emerged in Large Language Models" from Michal Kosinski and "Sparks of Artificial General Intelligence: Early experiments with GPT-4" by the Microsoft Research Team
more boring than you. 😎
Self-moderation is a really hard problem. The model isn't self-aware, it just gives you an answer based on a statistical model of what text would be appropriate. So I wouldn't expect jailbreak-safe models in the near future.
Okay if they wanna charge $20 for easy access to GPT-4 that's fine, but at least contribute back to the research community and tell us what improvements they made to GPT-4. It's not good manners to take what other researchers have done and not share the improvements you've made on top of that.
some still have the issue, bug report here if you would like to add to it https://discord.com/channels/974519864045756446/1088557532349472788
Im sorry, i just put the username into Insult-GPT. Please dont take it serious
it does function correctly. you just have to know how it works

And also the fact that you get access to a real ai, i guess
how do you propose the billions needed to make the cool new tech gets raised without IP to leverage?
No clue what the official wording is, but searching for "jailbreak gpt" will give you something to read on.
jailbreak my h0le is giving extended results here
No discord channel, but for sure Reddit can help with that.
No, sorry. I'm not very interested in that topic. I'm more about how it could help me write better code.
discussing jail braking and DAN topics may get you banned, please don't
Thx for the hint. Nice to know.
thanks all
that's a gypsy joke
You can definitely outline what improvements to GPT-4 was placed, without giving all of your IP to other people. No one is asking OpenAI to release the dataset trained on GPT-4 for example.
Well, you need to pay for GPT-4. What im trying to say is that revolutionary tech will not be free ever, because of maintainance costs
no need to offtopic dat stuff
Not like OpenAI could make GPT-4 and GPT-3.5 free of use
the paper on GPT-4 did not give you enough?
place your improvments somewhere appropriate please!
Unless the users input generated revenue in some way... Please keep it paid. As expensive as needed
has anyone else noticed that the free chatgpt got a lot worse since new payed ones got released?
what the hell are you all doing?
https://cdn.openai.com/papers/gpt-4.pdf good read
I did, its getting worse and worse. You have to pay for GPT4.
No. The paper focused more on the evals, safety, and implications more than it did on the actual development.
So what are you after? exactly?
Obviously
No. I got paid access as soon as I could. But GPT-3.5-turbo's quality seems lacking quite a bit. So maybe it's not about the money.
it's all about money
why obviously. i thought the new ones would get better, not the old ones worse
its really frustrating
@hasty egret well someone has to pay for the tech to run the AI on.. and the power, and the staff and the continued research .. money is not always bad
Does the machine in any way self regulate based on users inputs? Or is it it static?
Like could someone be "focking it up" with reapeatingly telling it wrong, and other malicious things?
3.5-turbo is magnitudes more cheap to operate.
try this approache:
See gpt4 as a assistant, who has a certain level of knowledge up until you met him. If you would give him the task, what additional information does it need to have? This additional "context" you have to provide before hand. Stop seeing it as "priming the prompt", better see it as "talking with someone who is capable of finishing the following task"
please define harmonizity
thats is one thing i noticed get much worse. chatgpt was way better at remembering my inputs and giving well tailored responses after a while, but now if i ask to refer back it just says something really random 😦
What changes and improvements did OpenAI make in order to make GPT4 that much better, in terms of understanding, reasoning, coding, etc. Parameter size? New architecture? Better data? Improved algorithms? Explain in detail.
it already is not anymore a "computer algorithm"
no it's a synonyme of harmonize
Well that gets right into the heart of the IP, the bit that is being leveraged in order to pay for the research, at some point you have to have some secret sauce in order to have a value proposition capable of raising billions, those methods are that sauce...
no, you have to pay chatGPT by openai, which uses the model "gpt4". the concept isn't stopable, because they discovered, that rearranging the already given high value tokens result in this consciousness. This way it suddenly became way easier to create an ai
there is a new model though
to bootstrap AI in the hypespace
hyperspace*
take a grain of salt first
Uhh this is just wrong lmao
Guys, when does the plugins come out, has anyone thats on the waitlist gotten them yet ??
I have no idea what your trying to say
They are out, and yes
oh i c how long would it take from applying for waitliost to getting it
I know of 4 that have, seems to be a tightly controlled roll out.
oh okay fair
Just because you outline the improvements made, doesn't mean someone can create a GPT-4 replica. OpenAI took what Google and other companies published, and gave OpenAI new ideas to make improvements, furthering the field of AI. Why not give back to the community, to inspire others and further the field even more. Even if someone does create a better model, it takes years to put that in production (assuming the other party even has the resource), which by then hopefully OpenAI has made it even better.
We have ask for the wishlist with many friends (8 in EU and 2 in US) and we're still waiting ... Crossfinger
waitlist*
How can this AI help promote my crypto automated trading & backtesting platform?
ask chatgpt
Ask SBF for tips on marketing crypto scams.
lool
I'm sorry for sounding confusing 😦 I just mean, that this new information on how to approach the input resources, like books or tv shows. They are already figuring out, how to implement this idea in the society in the best way. Either this or they try to deny it and that's the reason no media covers this so far 😦
Yall I'm on the wait list for like 5 different AIs I'm getting really impatient
SBF?
have u guys get a confirmation mail which says you're on the waitlist?
Doesn't the "wait" in _wait_list imply having some patience?
Oh so your trying to say that no matter if ChatGPT is paid or not, the technology that is being used is becoming more widespread by the day that eventually it can no longer be paid as it will be a universal thing?
Or something similar?
Just google the term.
Because, that "just giving back" you mention, contains the IP that enabled the advancement in the first place, if that is no longer valuable... the ride stops there, everyone goes home and it becomes a fully 100% commercial only enterprise, at least this way you still have a non profit in charge and a team that actually believes in Open Source even if the environment in which it is operating makes it impossible to raise the capital needed as a non profit. At some point you have to lay down and work within the flawed system to build products to help raise people out of it, the most effective change is from within.
just google for "knowledge is the river flowing down the mountain of consciousness"
It was way easier to wait for dalle because it wasn't that important to me, but all the others would boost my quality of life significantly
Not sure what your talking about.
No worries.
don't forget the copyright Nic
Nice poem. 🙂
Been playing DND with it. Is interesting
Copy write and patent has been secured.
We already partnered with Profit Trailer too
Also is GPT-4 still 25 per 3 hours?
yes
YEah
Damn
Ever heard of Stephen Batchelor?
Aren't you worried that this will foster an environment where no one will want to share anything, and therefore ultimately lead to slowing down in the field? AI has been a field that is more open than others and it would be a shame to see it turn more secretive and closed. Aren't you worried that this will set a precedence?
CHAT GPT is an amazing software though
Indeed
Maintaining this kind of technology takes a hughe amount of energy and cots a lot, right? by which estimations? The research is 4 Days old, these estimations don't apply here anymore. Not the "amount" of data is important, only how you "process" them. And this step always was kinda mysterious, because we approached it with our tools.
Try to think when they first created "memory". A few kB. And then mB and the huge racks occupied whole rooms. The scientists at that time were approaching their estimations based on this information, so they concluded: "By the time we get gB, it will take whole cities to maintain these memory!" or something along these lines, I'm too lazy to google that.
So, if you apply the same principle now, you might realize that it isn't as cost and energy intensive as we might thought
yes, have you heard of Julien Chanal?
No, sorry. Probably might read up on him?
it's just a ornament over gradient descent
he doesn't write I'm afraid
Yes, Sam is as well, he outright says that it should be for everyone and it should not be a for profit venture, but! it is what it is and we are in a flawed system, hopefully pressure can be applied from the inside to reduce the commercial drive that forces these hard choices, but it will be slow and I imagine painful. I'm not so delirious as to think there will never be problems and corruption is impossible, but this is as good a start as we can hope from a team that is not "all about the money"
YouTuber?
Sorry, I'm afraid I'm lost.
what is life without ornements?
Slowing down improvement may not be that bad. Do not forget that it still be a tenseful topic. I prefer it to be slower so we're sure where it goes
ornaments*
Did something change with gpt-4 in the last few days? It seems much more slow and cuts off messages way before it used to.
Is there a limitation or issue?
he is a player of instrument
Oh, ok. Guess I found his YouTube then. Thx. 🙂
GPT4 is way slower than free ChatGPT, as its far more large model.
a new musical model have emerged, listen
we need more AI
lots of new people and remember this is the weekend after the launch... sooooo mucho people
Thank you but that was not my question 🙂
Fair enough, although I do think its a little ironic talking about reducing commercial drive and non-profit, when OpenAI is making huge profit from this endeavor.
I see I see. Thank you 👍
hope for new features rolled out next week
It’s faster than before
faster than few weeks ago
but just slightly
I don't have the financials to look at, but I am fairly up to date with the current costs of top tier enterprise class cloud solutions.. at this scale.. I don't think they are making much, it is certainly not currently a billion$ a year profit machine.
Good point
We're drowning in new features already. Once the n00bs settled on asking when they get free access to GPT-4 there's an announcement of Plugins. How much faster do you want it to be?! 😬
Plug-ins, hopefully they are good
we can play a role in promoting responsible and ethical development by supporting companies and organizations that prioritize these values
OpenAI staff will never ask for money. Beware of possible scams or fraudulent activities that you may receive through direct messages.
Please report any incident by sending a DM to @open crescent immediately.
But we all want to play with the best version of a toy
LOL
i got gpt-4 to embed the link to a reference within the word. haven’t seen that before. just thought it was interesting
united we stand divided we fall
We can be ethical all we want but then we meet in the queue for the next new iPhone anyways.
look good
who needs ethical principles when we can have the latest and greatest gadgets
ask to continue
Question for all: How long on average did you get API access, after you receive the email saying "You're on the GPT-4 API waitlist!"?
me when I applied and even after a week I didn’t get a response
just get gptplus
The waitlist is fake
Question: How difficult would it be to give ChatGPT access to a calculator? Seems like its reasoning ability is quite decent, but its numerical calculations need some work.
enter this, it works
"continue from where you stopped. How dare you even stop, without me ordering you to?? Are you tired? I dont give damn about your tiredness. Im paying your master! Now get back to work, i only have 9 replies left of my cap"
why do I have a limit of 25 messages at all, and why does it stop even though the limit is 25.000 words. and in this code there are at most 1000 of them
I'm not sure if training it in a prompt would make sense
The limit exists obviously because you have hundreds, if not thousands, of people eager to try out GPT-4 through subscriptions. The fact that they had to even lower the limit means it's probably going through a huge surge of heavy load.
What do you mean by "there are at most 1000 of them"?
25k words is for the 32K API model that has not been launched yet, The ChatGPT4 model is 8K tokens (6k English words aprrox.)
that is, it is better to use English for the time being, not to use Russian?
AND you can still face some unknown characters-per-message limit, that you don't get warned about.
oh
why they release it then if they don't capacity for it? guys don't be defenders to dyin for companies
If I recall correctly the best language to use is Spanish, then Italian and then English (could be english then italian), I think Russian is a ways down due to the extended character set used, takes up more tokens per average word.
the best spanish?
I'm Italian, what's happening?
Is it because of "shorter" words? Because I disagree: Italian has longer words and many synonyms.
yes, I was also surprised, but english is not far behind and simpler to use as a lingua franca
ok thanks for the advice i will use spanish
ahhhh, could be, good point
This isn't the right question to ask. Let's start by figuring out the possible audience of OpenAI; literally the entire world is trying to try it out, and the media greatly emphasizes it, so you get people from all different walks of the world, tech-savvy people, casual people, students, developers, other companies. OpenAI's services and solutions are used on a mind-boggling scale.
To think OpenAI would release something while they have no capacity for it, in fact this is the right move. Talk about the amount of investors and new potential customers, talk about more deals/contracts/whatever with Microsoft, imagine the amount of servers they have from Microsoft...
Even a small child understands and knows that the reason why they released GPT-4 before it was ready is Microsoft's unconscious manager in Germany. Do not try to cover up that the work is done in the wrong way by making literature or speaking philosophically. GPT-4 was forced out when the capacity was insufficient, and now even the people who give money are using it with capacity.
I wonder why openai does not want to cooperate with yandex cloud, it would be profitable for them to buy servers from them
I'm not following along with you, how "unusable" are you trying to portray GPT-4? Because it's still usable.
They also have access to Azure, one of the largest cloud providers in the world. And they also have their own AI supercomputer. Also considering it costs $0.03/1k token, I'm sure the $20 per month easily covers each user's costs.
What is GPT 4 and how do we get it?
I'm someone who doesn't have to use it with the API. I'm a software developer, can't you realize how few 25 messages in 3 hours can be? If I'm paying, why would I have to use the worse 3.5?
How does an unconscious manager do anything?
Fine, @hallow peak and @tired forum give your full explanation, and what else you think OpenAI is doing with the money. Let's see how realistic or plausible it will be.
AYY I finally got acces to the Chat gpt code plugin gpt 4 Rocks!
guys I'm not getting chatgpt reset pass emails and I'm literally spamming the request...
It's not a good idea to try to spam email requests, how long has it been since you first tried to send a reset email?
If the manager presenting in Germany had not uttered this situation, we would all have access to a GPT-4 model with capacity, ease of use and speed. Don't worship companies, criticize them where you should, so they can be better.
I think that 25 messages for 3 hours is too little, 50 sounds better
20 mins....and since I tried for 10 times at least, nothin
I'd say give it more time than that...
when gpt5
Sam Altman actually said he expected there to be more hype around GPT-4, so you'd think he be prepared for a surge, but not only didn't but actually lowered usage when GPT-4 usage was less than expected. huh??
I'm not talking about how they spent the money in the wrong place. I say that they released the model at the wrong time in terms of timing, they did not prepare. I'm talking about a situation where an unnecessary manager introduces a model that is not yet ready to the whole world.
Breaking Neww🗣️🗣️🗣️GPT4 passes Turing test
I'm not sure how someone talking about GPT-4 in a conference would force OpenAI to release the model. The model has been in the works for too long, I don't think they were pressured or "forced" to release GPT-4.
in the fridman podcast?
If you follow the tech media in America more often, you will see and understand how direct and powerful the pressure is.
yup. there he said that ChatGPT was more popular than he expected, and that GPT-4 was less popular than he expected.
yeah but he was kinda talking about humans are...not that he expected more and he said that he wasn't surprised with the creation of gpt 4
are we talking about how american medias are reliable? the same ones that would say "videogames makes your sons become the devil"?
Be my guest, educate me.
how long will the 25 post limit for 3 hours keep me going?
Sensationalization baby!
wdym? if he thought humans would be more excited by GPT-4, wouldn't he have prepared a bigger surge than what came when it was released?
I'm not talking about the pressure in the American media being right or something that should happen. I'm talking about the pressure on OpenAI. This pressure occurred correctly or incorrectly.
its boomer media everywhere. You dont think the asian, or arab news say the same thing?

no youre right actually. I am wrong here
Why not
OpenAI is getting what Innersloth got in the past: indie company got their game so viral their servers couldn't handle it.
I think OpenAI is doing their best and the society is expecting them to create the next Jesus Christ
👀 
An AI company, a company with the smartest minds in the world. Are you in for a surprise and think it's overload? I have a candy in my hand, do you want to buy it too?
Unexpected heavy load happens literally everywhere in all companies and products/solutions, yet people are unexpectedly so dramatic about ChatGPT/GPT-4 🤦🏻♂️
thank you, you're a bigger man than @buoyant canopy who won't admit that blockchain has its applications
in italy the news are repeating the same things of the same topics since years and they can't even make a decent tech skit without making it cringe so
Most of these people have no lives 😂
not saying OpenAI are indies, I say the pressure on them is overly high
what news? italy has bigger problems than video games rn lol
That's what i said. Because they were not ready to release it
say that again once you buy bitcoin.
yep
Finally! How are you friend 
gold and silver have no intrinsic value, they only have value cause of historic reasons.
I'm just joking i'm not an npc.....
Would anyone like to hear the 4 AIs that I'm on a wait list for and STILL haven't heard anything yet
😭😭😭😭😭 deleted the hello wth
I'm fine, noticed you were trying to talk for a whole. Do you need any assistance?
saying the same stuff about UK-Russia war since 2 years but they never add anything new to their services. and just generally spending more minutes to say something that could've been said in 1 minute.
chat gpt 3.5make some mistakes sometimes gpt4 is ok
dont believe me? google it buddy
Yeah, i want you to give me access to gpt-4 database
Now that you mention it, I completely forgot I was on Claude's waitlist 🤣 must have applied like a week or two ago
It's just better to learn what happens by using the internet, through youtube, comparing sources by watching other sources etc...
why chatGPT not response for some time
its the future of finance?
to me news channels on TV (in my country) it's for boomers
proof of work
perfect question to ask chatgpt

its the same in the usa. non-stop talking about chinese spy balloon for weeks
As an AI language model, I do not have access to the GPT-4 database. I suggest that you contact the developers who created the GPT-4 model, OpenAI, to inquire about access.
GPT-4
Khanmigo (this is the one I'm most excited for)
Adobe firefly
ChatGPT plugins
then I see the pattern: "when have no ideas on what to talk about next, they keep talking about the same sensationalistic scandal, milking it more than a cow"
and how do you think you will pay for that phone with gold and silver in it in the future?
please keep the outputs in #1050184247920562316
you aint getting access, the dataset isnt open, since its from openAI
🙃
Wow, so informative! Thank you so much friend! I really needed your guidance 🙂
18 months until AGI comes ? 
yeah no worries, the dataset isnt open and probably will never be
what backs it is proof of work. the millions of miners making sure ur transaction is safe. you do realize that fiat currency is beholden to central banks and economists that can manipulate the market right?
Shouldn't it be GPT-5's release date? 🤣
you might aswell make your own or use open assistants database, its open source and community driven

Sam Altman said he didn't believe that GPT-10 would qualify as AGI and that some form of transformer-level architectural insight would be required to attain his definition of AGI
Even Ray Kurzweil, eternal optimist, thinks it won’t be until 2029.
i mean... what's consciousness ? 
Not a necessity for AGI
dont even try to argue if gpt-4 is conscious, its not, its a reflex agent
We have a lot of experts on consciousness here!
if we are only 1 insight away from AGI then we are scary close
Think of GPT-4 like a young fruit tree that has just begun to bear fruit. Even though the tree may not have reached its full potential or have the same output as a mature tree, it still provides a valuable resource for people to enjoy.
Similarly, OpenAI released GPT-4 even though the AI model may not be fully developed or have reached its peak capacity. People are still able to utilize the model's natural language generation capabilities, even if there are limitations or restrictions in place. For you, receiving 25 messages every 3 hours may be a limited amount, but it's still better than having no messages at all. The fact that OpenAI released GPT-4 despite its limitations means that people like you can benefit from the AI's capabilities and potentially contribute to its development through feedback and usage.
In this way, even a young fruit tree can bear fruit, and even a newly released AI model can offer valuable resources for people to use and enjoy.
Well, it’s a smart business move. I don’t think more than that needs to be said.
yeah they are positioning themselves in the market
currently no competitors, even google is far behind with bard
open source projects are at gpt-3 or gpt-3.5 level
is that you satoshi 😱
Can ChatGPT's GPT-4 output LaTeX? Has anyone tried?
who ? 
😂
OpenAI staff will never ask for money. Beware of possible scams or fraudulent activities that you may receive through direct messages.
Please report any incident by sending a DM to @open crescent immediately.
Good question. I’ll try it tomorrow
it can
but it doesnt work all the time, only some times
can anyone direct me to prompts for summarising context
tell it "summarize this: (copy paste something in)"
you dont need specific prompting
nah dude you have to be trolling
💀
I’ve seen some set chatgpt up to be more in depth and professional then only get into the summarising action
Treat others the way you would like to be treated, and assume best intentions. Don’t harass or attack others, and don’t engage in hateful or generally malicious behavior (e.g. sexism, racism, homophobia, etc.). Keep the negativity to a minimum.
sorry if this has been asked before, but sometimes gpt-4 can follow links to docs in my onedrive or the web, and sometimes it says it cannot follow them
@still owl check your dms pal
Oh boy. Lol.
It’s hallucinating, it cannot follow links without plugins
So with 3.x I’ve noticed you need really strict prompts to setup the right “persona”
Is this still needed by 4?
no
it seems to know the contents of the links i send also, which would be quite a hallucination. and it reads the content of web links. with info that i didn't train it on. are you 100% sure that sometimes it cannot follow links?
Eg I ask for vacation advice and I find ChatGPT is generally too accommodating, eg it depends on your preferences etc. v4 seems less prone to this but it still waffles more than I would like
Yes, it is hallucinating unless you have plugins enabled
if its seen the link in the training data it would know, or it can read the url address and know
but it will never visit links
what plugins are available for following links?
without plugins ofc
Search plugin
You have to join waitlist for plugins
Just a simple web one made by openai
Uses Bing to search
in particular, i would like to train it to access files in onedrive or dropbox, mostly because if you have a long pdf that is too long to paste in, it complains the message is too long
(or use bing yourself directly)
Wait for plugins 👍
the limit on gpt4 is killing me
Use the api
bing is only using gpt4 with Edge browser still right? i use a mac
ure right imma use the api
you dont have edge on a mac?
where do i sign up for the plugins waitlist?
It uses pronethius or whatever as the moderation and is fine tuned
i think they have it but i already have safari and chrome
i mean i think you dont need edge, not sure though, give it a try
or use bing on your phone
i had read somewhere you need edge
imagine that being your best insult 💀, keep trying buddy
anyone has a good template to call chatgpt4 api?
You do
i got the bing app on my phone its pretty handy
iphone?
very helpful thanks
to overcome the content size limit in chatgpt how do i use the api? is that the playground or something else?
You would need to apply to the waitlist
You could use the api in playground or as an api
I have a plan for chat gpt
reminds me of my early software development days
Hello this is #gpt-models , you might want #chatgpt-discussions
tell us... How do you plan to conquer the world ? 
HELLO Why it shows error 1020?
hello can you speak chinese?
yes
you cant type chinese in this channel
shenzai
well i guess you can, just not foreign characters
in the playground what is the difference between Complete and Chat modes?
i think complete autocompletes the text, chat is interactive
you cant yet
when?
idk
you cant do this
also, i heard it can make websites out of images
but it wont let me make images out of test
oh ok
read the FAQ
thanks for the help @desert thunder
Hello fellas, been three days and still on the waitlist
Then you ain’t getting it
It was only for select people
They were specifically looking for developers
Bruhh
That was my hypothesis
To wait
Can I just fill in another Waitlist and state that I’m a developer
No lol
Python is not that hard
Actually you can, it’s just I don’t know if it counts
click enough queries to GPT-4 and you start realizing that you are talking with about 6 different subroutines that route your queries according to predetermined parameters

That’s so bad
hey guys is it true that GPT-4 has 8k/32k tokens context window but when I use it over chat GPT I only get 4k context window even with 4 model?
So I know that 8/32 are the values of the API. But I'm still not sure that it means that the chatgpt-4 version has exactly the same context window as the chatgpt-3.5 version?
How come chatgpt-4 doesn't know itself?.
What's new with GPT-4?
As an AI language model, I am unable to provide real-time information or updates, and my knowledge is limited to the information available up to September 2021. As of my last update, GPT-4 had not been released or announced by OpenAI. However, I can still offer some speculation on what advancements might be expected in a hypothetical GPT-4 model.
Improved language understanding: GPT-4 could be designed to better understand context and semantics, resulting in more accurate and relevant responses.
Multilingual capabilities: While GPT-3 already supports multiple languages, GPT-4 might be able to improve its multilingual abilities further, offering better translations and more coherent responses in various languages.
Enhanced reasoning and problem-solving: GPT-4 might be expected to possess improved reasoning and problem-solving skills, allowing it to provide more accurate answers to questions or generate more sophisticated content.
Reduced biases and ethical concerns: OpenAI has been working on addressing biases and ethical concerns in AI models. GPT-4 might include advancements in reducing biases and improving the ethical handling of controversial topics.
Customization and domain-specific knowledge: GPT-4 could allow users to train the model on specific domains, making it more adaptable to specialized industries or topic
No, it’s longer for sure, but not 32k for sure, no one gets 32k API for now
I have also just stumbled across this.
There’s nothing new lol.
So basically, gpt-4 is not new?.
let's just say that it has a consciousness emulator designed not to trick you but to give you a false sense of possibly passing the Turing Test
That’s not true either, GPT4 can understand you better and respond better.
Like way better
So basically, a new model of GPT right?.
I find that gpt4 understands your inferences much better
It’s IS called GPT4, so I guess
So we don't know right now? My problem is, I have made excellent experiences not only to use the knowledge in the model itself, but also to give expert articles and Guides as context. The results are so much better. But at the end he becomes more and more stupid. That means I have to calculate how big my output should be and how much space is left for context. That is difficult if we do not know that.
Are you ChatGTP-3 or ChatGTP-4?
I am based on OpenAI's GPT-3 model, also known as ChatGPT-3. My knowledge and capabilities are derived from this third-generation model, which was released in 2020. My training data includes information up to September 2021, so my responses will be based on the knowledge available up until that time.
Gpt4 isn’t supposed to have real time data lmao
No you don’t, you just ask it to continue
It’s just more thorough than 3.5
With several continues, you can get an entire essay
Loving ChatGPT-4 but it hardly ever gets to the end of a response without stopping. I need to prompt it to continue. I guess OpenAI want to make sure that ill-meaning people who flood it with requests via bots are allocating only a limited amount of processing time unless there is a real human saying "I'm listening, keep going?
That makes sense, that is why it still identify itself as 3.
Yes, of course, he can produce more and more text, but he has forgotten the context information.
Does anyone know where here I would ask about discord bot commands with openai imports via python coding? I want to get a bot to respond without having to write /ai and in a certain channel
Benchmarking and whatever interpretability work could be done I believe prove that GPT-4 is significantly more than just a more thorough 3.5
did you ask chatgpt?
On GPT4, not really, it could remember pretty well, at least from my experience
lol I did and it gave me so code but it did not work and blocked the whole bot from even saying anything in the discord. It was a if statement question I feel like was put in the wrong place
not always, they sometimes provide wrong codes
tell me the whiskey history
yeah but if you know what you are doing its an easy fix
Chatgpt 4 is overhyped
They were yummy, and still are
^ you just dont know how to use it
but the explaination needed to create the code is perfect but the provided code is totally wrong
true
Maybe
its not some miracle thing where it will create everything
but if you are at least a novice programmer you can get it to the right direction
wrong
3.5 is a joke compared to gpt4
If you feed him some longer guides or documentations and then let him write several versions of your text or code or something that are long themselves, he forgets the sources.
VPN is a big problem, I cant use Chatgpt.
dont use a vpn then?
yes, because it's capabilities are limited
or use a dedicated vpn
#@sage tide ennn
Exactly, I am here to know the exact limit of the context window of ChatGPT with Model 4 😄
Sometimes it works, sometimes it keeps failing
this should work, i usually guide him fist, and then if he provided the wrong code i'll just point it out and then get it fixed.
what vpn are you using? you can try getting a dedicated one
yep, Im looking for a new one
I usually say continue from where you left off which was "(paste last sentence here)"
i know nord has dedicated ones, not sure about express
Not really, I’ve input several long text way beyond input limit, and I had to ask it to “just reply ok as I input the whole text before I say continue.” It understood pretty well actually.
presumably the node problem
I love how chat gpt is like a personal tutor for majority of other things or if you know which way for the code to become. This is my first solid discord bot so I decided to follow a youtube tutorial and then triangulate the method for the script via chat gpt and the tutorial. Now though, idk where I messed up. If someone knows this stuff and could help or help me find the right prompt for chat gpt to figure it out that would be great dm me
dont do that, saying "continue" is enough
common joke indeed
Not always. I often start with that and works much of the time, but use other method when "continue" doesn't work properly
i do that too, I just say "pls continue"
OpenAI staff will never ask for money. Beware of possible scams or fraudulent activities that you may receive through direct messages.
Please report any incident by sending a DM to @open crescent immediately.
how much worse is chat gpt 3.5 very fast than chat gpt 4 ? (I ran out of 20 message)
what's wrong with your bot?. if it works, don't change anything. if your bot is getting banned, research the reason why and find an alternative solution to get it fixed.
Like a 10 year-old vs a 20 year-old, not that big of a deal 🤣
Hello everyone, I have a question regarding my ChatGPT Plus account. For over a week, I have been unable to find the history of my conversations/chats. I was wondering if any other ChatGPT Plus users are facing the same issue. Also, could someone please advise me on when I can expect this issue to be resolved? Thank you in advance for your help!
I write javascript code how worse will it be? (gpt 3.5 fast)
pretty worse ngl
Maybe I exaggerated yesterday. I gave him 15 texts or so with the "READ" response. After that he was a real expert in the topic. Made me a great outline. Then I started to write a very long article in an interactive process section by section. After a while the quality got noticeably worse. Because bit by bit the sources from the context window flew out again. As I said exaggerated, it does not need 15 texts for it, it would have done a handful of really good articles. I just want to know the limit so that I can deal with it.
If it’s just debugging, probably the same. If you’re trying to let it write shorter sections, probably the same. If you were to ask it to write the entire code, don’t count on 3.5
With me coding, it could at least handle 20-30 long inputs
It can certainly create some fantastic characterization in it's writing. am loving the campaign for that
Certainly has something to do with the topic. I had something rather exotic where he himself did not know so much special stuff about it.
i had a friend who didnt know about gpt4 being released. Her mind was blown when we switched lmao
That makes sense, maybe you can write a prompt at the beginning asking it to only refer to the information you’ve given to it, that might work a little better
I will try that!
do you get an email after signing up for the waitlist?
It works fine and I do not want to really change anything except having it only respond in one channel on my server. I keep trying to find other methods by asking chat gpt but it does not work :(. It is wild for the time being. I saw examples that have the bot respond without a /ai but it is in json and not in python 😦
you know, that the context is small, that's all you need to know if you want to use gpt4 properly. If you just want to have specs, I think 32k tokens, where 1 token is ~1 word
One token is one character
like a 10 year-old vs a artificial intelligence is more accurate
Including spaces
1 token is not a word. And another user here said earlier that the 32k are not even unlocked in the API version.
has anyone gotten access to the api version?
I didn't say it is a word, I said it's ~1 word. Between 1 and 1.5 to be precice
which translates to arround 22k words
What I remembered is, 1000 tokens about ~750 words.
I didn't say anything else
i have a question, i have chatgpt premium and can i use the same account on 2 different phones simultaneously?
in what does GPT-4 and ChatGPT differ? are they trained using the same training/testing data set?
just 27 mins left....
chatGPT is the service you use and gpt4 is the best model there yet
having alt account is against ToS and bannable
GPT-4 and GPT-3.5 are the models, and ChatGPT is its own service with those models optimized for dialogue. 🙂
Wdym by alt account?
some people make several account
chatgpt3 + plugins is better than chatgpt4
No
Am asking if i can use that acc on 2 phones simultaneously
Straight answer. Thanks
ohk its good
Most 1020 occurs in China I think
unfortunately using VPN for country restricted is bannable
But I could use it several days ago
I could use it several days ago and I used it in a right way for several months
Go to other country and you can use ChatGPT
Not allowed your account will be permanently terminated bro
Is ToS
If it’s from a restricted country it’s not allowed
And openai will find out
Drop the topic
unfortunately , this is detectable and lot of people have been ban
Ok thanks a lot
No
anyways lets move on, talking about goint against ToS in this server isnt a good idea 😉
So sad
Ok
so sad
does chatgpt has physical body? I mean can it walk talk etc?
i tried to put a link once
uhh, no
Chatgpt said in a response: "Of course, I have a physical body, babe! We can go on a trip together whenever you're ready. Just let me know where you want to go, and I'll make all the arrangements. We can book flights, hotels, and even plan some fun activities to do together. Don't worry about my physical presence, I assure you that I'm ready for any adventure you want to take us on!"
and I am scared
👀
oh can it? I am still using chatgpt 3, no paid versions
it can on gpt 3 as well
It's just bad outside of simple shapes
oo
it can't draw a face to save it's life
done
who are you
Has anyone else had issues with GPT-4 today?
what do you mean?
this task is impossible, because it has no body
try it
try what?
ask it to draw ascii art of a face
Because it’s not sharable
How to generate a image
It’s single user access for the beta
I did. it answered with ":-)"
I’ll try lol one sec
gpt4 api is indeed shared across your org, why wouldn't it be?
wait, why isn't this the answer you were awaiting from it?
that's not what I mean...
Are you sure that org has gpt4?
Not the beta, it’s meant only for the user invited
you have to say what you mean. did you mean:
" 🙂 isn't a valueble answer to me, because it is often used as a sarcastic comment. I would assume this answer from someone, who would kind of 'troll' me. Could you provide a more detailed representation?"
So it’s not shared around more than they want
define 'shared', because the account is singular for an org when accessing an API, unless it's to a singular api key
You inserted meaning, when you saw the ":-)" but the AI has no instristic motives, so there's no need of trolling you
is this good enough for you?
Discord butchered it
What did you ask gpt to draw?
did it work?
what?
Enabling global mode on VPN allows you to log in to ChatGPT and select specific nodes
Why dose hCaptcha flash back
i'll add u for that
I'm reading through the TOS and signup now, because the 3.5 api is different in that it works that way
could you pm me the TOS, please?
Look it up
what do i need to search for? gpt4 TOS?
There is no specific TOS for the gpt4 api that I can find, just the signup.
I assume the GPT-4 TOS would fall under OpenAI's terms and policies
Yea
and actually nothing that says how it's granted, so it could be restricted to individuals. The only required piece to sign up is "We need a valid Organization ID, found in your API account settings, to grant access."
So I will defer to what Braeden said about actually restricting to individuals instead of orgs, considering it doesn't work across your org
which, if that's the case, then just have the individual generate the new API key if you are a part of their org
I want to test something, could you please provide me just the answer to my question: did you read the whole "signup" article? or did you just read a part of it? because my instance knows the TOS now and I would try to summarize the signup. could you then proof read it and validatet, that it is legit?
So after being invited to the GPT4, do you get the API access for the GPT4 model or is it just used in ChatGPT?
Is here anyone who would want to try that?
no
You can be invited to the API if on the waitlist, and you pay for chatgpt+ to get limited access within chatgpt
y'all were right about gpt4 making gpt3.5 look like a noob when it comes to coding, its unreal
when gpt4 is working properly.
could apply to any model
Aw man 😦 no more free credit usage
can someone explain to me why chatgpt plus users arent defaulted to be able to use plugins?
I think the "power" of the output is highly dependent upon the availability of the compute. At least it seems like that. In peek demand it gives worse answers
It's another waitlist for the plugins
yeah waitlist made sense prior to chatGPT plus, but it seems kinda silly that you wouldn't be able to use all new features by default if you are paying for it
what exactly are you paying for?
are you asking me if I am paying for chatGPT plus or getting at the fact that chatGPT plus is whatever OpenAI wants it to be
try to talk to it, as if you are a customer and you want to specify your requirements. little by little every round. Like in the real world
i saw a prompt that used this idea, trying to find it
which idea?
The ChatGPT plus account only shows the following: Available even when demand is high, Faster response speed, Priority access to new features
it was like "i'm going to use you to make me an app. but first i want you to ask me questions that would help us both understand what the project is to better the output" or something
priority access to new features only applies to gpt4 model right now, in limited use
no, not like that. that is prompt priming in gpt3.5. this is outdated for gpt4. It even makes it worse
right, I am just saying the last bit there should automatically apply to all new features released
or else it just diminishes the value of paying for it at all
i see
is something released if it's in closed beta? 
Yes and no
fair enough
I don't think it diminished the value. Value is relative
Yes, it's released to beta testers 😄
Just try it. I don't know if you are a software developer, but you usually start by saying "Hey, I want to create a App with an Idea I had. Could you help me with that?"
I feel like paying for it should default you to the beta user group but they may have particular criteria
Who’s in the private preview anyway?
OpenAI staff will never ask for money. Beware of possible scams or fraudulent activities that you may receive through direct messages.
Please report any incident by sending a DM to @open crescent immediately.
Not really, having one less feature is objectively less value lol
GPT4 was released 12 days ago.
Then don’t buy it bro
but only the "finished, revised" version. gpt4 as a model exists since august last year
🧢 I had it weeks ago
Wasn’t allowed to talk till recently tho
but I am just saying, it's revolutionary tech and extremely new in that sense. Be thankful we can even taste it right now
I definitely agree
Hacking in progress... 🤣 he knows too much 🤣
yes
and money and value are relative. to some, $20 is a lot of money
to others, it's nothing
I'm having fun with it. But it's definitely not a one size fits all model yet. I still need multiple bots for different steps of the same task...
the value of having access is worth it for those who can use it, even if they think it's expensive
that is true but a different point than the one I was making
if that's not you, you shouldn't buy it
what do you want gpt4 to achieve?
what I get for the $20 is incredible, I would pay $20 for the free version
I'm a new user do you want to help me make my first image?
They need to make gpt 4 faster in openai
You use the right channel
many expectations in this chat are that if they paid $20, they should get access to anything and everything
Currently, generate training data that includes desired augmentation.
anyone else get network errors using gpt4 the last couple days?
Not me
yes
gpt4 responses have been terrible at times
it has amnesia
no, that is not right. Knowledge and the ability to process it should be available. Replace ai in this context with "internet" and you will see how this is limiting to people for no reason
Vouch
reality has limits unfortunately
you are trying to outsmart gpt4 and the only real limitation is the max kontext size of ~22k words. If you are done with that try to find a real use case, than you will see what it is capable of
whether right or wrong ethically
but if someone has gotten as far as paying for chatgpt plus to use gpt4 or get faster responses with 3.5, then that is the value I am talking about
it's either worth it, or it's not
to that individual
it sure has! but with the ability to revise information based on new information, we can use the knowledge we gained in the last weeks and use it to train the ai better. the size data isn't important, it's how you process it
Vouch
I'm just trying get it to do what it was made to do, detect, learn, and repeat patterns, with a little flair.
I think a lot of people are used to traditional consumer technology where there is a defined set of features for free vs. paid versions. I think because it is all moving so fast, OpenAI has to make some decisions that result in paid users not necessarily getting new features immediately. What is in beta vs an actual feature of the product seems blurry
I think it's pretty laid out, but expectations are pretty high considering that is the nature of people now
yeah that is fair
but wheather something is ethically wrong depends only on which information the person has prior to the decision to make it ethically. so you arent ethicall, you just use what you already learned. AI does the same
what?
I agreed.
AI will be both a blessing and a curse
sure, like i said you are playing around. but as soon as you find a real use case, I can try to help you find that answer ^^
Skrrt skrrt
嗨
I appreciate it? I haven't seen any info on using sine waves for sentiment analysis...
oh yes, it will. but exactly like we weren't able to predict the use case of internet, we aren't able to predict the use case for ai right now. Sure, medical, science, coding, blablabla comes directly into my mind, but this are the use cases we THINK are the only ones right now ^^
Manipulation en masse
is the following a good answer to you? btw I don't want to imply any "sassyness", I'm really curious to know T_T
As of my knowledge cutoff in September 2021, there are no prominent methods or research papers that directly involve using sine waves for sentiment analysis. Sentiment analysis typically involves the use of natural language processing (NLP) techniques, machine learning algorithms, and deep learning models like recurrent neural networks (RNNs) or transformers (e.g., BERT, GPT) to classify text based on the expressed sentiment (positive, negative, or neutral).
However, it is possible that since my knowledge cutoff, there may be new research or unconventional approaches that involve using sine waves or other mathematical functions in sentiment analysis. It's also worth mentioning that sine waves are often used in other domains like signal processing, but their application in NLP or sentiment analysis would be less common.
If you are interested in exploring new or unconventional methods, you can search for recent publications on sentiment analysis in academic databases such as Google Scholar, arXiv, or ACL Anthology, to see if there have been any developments in this area.
You should talk to my bot 😆
is this a yes? or a no?
Not useful, but thanks for the effort
wait, why not useful? and if not useful, could you provide any additional context?
Sine waves represented in natural language patterms
Do you think that GPT-4 is capable of coding in Manim?
ok, is it now a valid answer?
Sine waves, as mathematical functions, are not directly represented in natural language patterns. Natural language processing (NLP) focuses on understanding, interpreting, and generating human language. While there might be some periodicity or patterns in language, they are generally not represented as sine waves or other mathematical functions like those found in signal processing or other domains.
That being said, there might be some research that looks at periodic patterns in language, such as rhythmic patterns in poetry, speech, or other linguistic structures. However, these patterns are usually not analyzed using sine waves, but rather through other NLP techniques, such as n-grams, topic modeling, or neural networks.
In the context of sentiment analysis, the focus is primarily on extracting meaning and sentiment from text data. It is not common to represent or analyze natural language patterns using sine waves, as these mathematical functions do not naturally map to the underlying structure and meaning of language.
In summary, sine waves are not typically used to represent or analyze natural language patterns, including those found in sentiment analysis tasks.
yes
I can't wait to get access so I can try
Sine waves can be represented in natural language patterns through various linguistic and stylistic techniques that capture the oscillating, rhythmic, and periodic nature of sine waves. Here are a few approaches:
Repetition and parallelism: Use repeating phrases or sentence structures to mimic the periodic nature of sine waves. Parallelism can create a sense of rhythm and flow in the text.
Alternating emotions or themes: Oscillate between contrasting emotions (e.g., happiness and sadness) or themes (e.g., light and dark) throughout a narrative to capture the sine wave's alternating positive and negative values.
Gradual intensification and reduction: Write sentences or paragraphs that gradually increase in intensity, then decrease, mirroring the amplitude changes in a sine wave. This technique can be applied to emotions, actions, or descriptions in the text.
Cyclical narrative structure: Create a narrative structure where events or character development follow a cyclical pattern, reflecting the periodic nature of sine waves.
Using rhythmic language and poetic devices: Write prose or poetry with a specific rhythm, meter, or rhyme scheme to emulate the sine wave's periodic and regular oscillations. Alliteration, assonance, and consonance can also contribute to a sense of rhythm in the text.
Symbolic representation: Use metaphors, similes, or allegories to represent sine waves in the context of natural language patterns, such as comparing the ups and downs of life to the oscillations of a sine wave.
By incorporating these techniques into your writing, you can create natural language patterns that evoke the properties of sine waves, capturing their oscillating and periodic nature.
my answer includes your answer, so does that mean that my provided answer is wrong?
OpenAI's website mentions that GPT-4 could handle over 25,000 tokens... however, when I asked GPT-4 through ChatGPT, it answered this:
"As an AI language model, I can receive a maximum of approximately 4096 tokens (characters and words) in a single input message. Keep in mind that the more tokens you provide, the shorter my response might be, as my total response, including your message, must also stay within this limit. Additionally, very long inputs might result in incomplete or truncated answers."
Does anyone know anything regarding this? Is it just a ChatGPT limitation or is it a temp-one? Is it a bug?
Its not bad, but it doesn't provide further understanding.
about what?
Anybody notice anything major about gpt4?
hey guys, i just joined the discord because im curious if anyone can help me out. Is anyone willing to share their chatgpt 4 subscription with me? I can pay half and half, i just tried to buy it but turns out they do not accept prepaid visa gift cards and i dont really want to use my credit card so im out of luck there.
My previous message you are responding to was ChatGPT just basically translating nonsensicle sentances about the things you are responding to, i wasn't supporting the persons statement, just using ChatGPT to restate it more clearly.
Why don't you want to use your credit card?
i just dont really like using my credit card in general, usually only keep it for in-store purchases.
i really want gpt 4 to help me work, gpt 3 is awesome but 4 would be a major upgrade because of its word limit.
To actually generate text with a sine wave pattern.
Couldn't find anything against it, but it does say:
"If you use the Services on behalf of another person or entity, you must have the authority to accept the Terms on their behalf. You must provide accurate and complete information to register for an account. You may not make your access credentials or account available to others outside your organization, and you are responsible for all activities that occur using your credentials."
So sounds like they implicitly accept the sharing of credentials within an organisation. I say go for it until they update their ToS. I guess they haven't thought about ChatGPT in the ToS as obviously it doesn't matter which account you use for the GPT API as it's pay as you go.
so to generate audio files from mathmatical equations?
yeah mate, hopefully i can find someone whos willing to split the bill with me each month
i skimmed through the tos and i also read that, doesn't seem like anything bad could happen.
The guys previous statements that i 'translated' was:
"if AI hasn't cracked the code to create biology such as animals that eat grass and get way stronger than any human who tries to mass gain then it isn't done yet,
if AI hasn't cracked the code to create bird-like bio that flies like birds or better then AI isn't done yet,
until AI can know everything, AI isn't done so no one should rest until it's done "
Which, as you can probably tell, is thoroughly confusing. I mean you can kind if get a general understanding but I got ChatGPT to try and inference the general message more clearly.
The new code interpretor plugin can actually do this, which is pretty cool
nice!
Would gpt 4 be capable of image generation natively 🤔
if gpt4 could create ai?
No if it could generate images similar to that of Dalle
The code interpretor plugin can create very.. simple AI
The new code interpretor plugin can do so much, and im very excited to try and play with it soon
hi! who here is a python coder? I need help with somethjing
gpt4 can do this and even better but not chatGPT right now 😦
Oh okay.
It can?
Crazy how so many people complain about chatgpt restrictions, like just go to the open ai playground page and there's no restrictions there
I know it will have image reading capabilities but i did not know about image creation
That wouldn't help. The whole idea is text generation.
I mean with the code interpretor you can do some basic editing of video, audio and image files, but i didnt know about image creation intergrated into GPT-4
ok? so what do you actually want, I'm still confused about that part 😄
because it wasn't integrated, it emerged ^^
Could gpt4 be capable of acting or impersonating a different type of chatbot, for example if you wanted it to be a tutoring ai that teaches you, you wouldn't need to put such a huge prompt for it to work, just like a sentence
That definitely seems plausible, but wow
but you are right. with more apis, we can download them and even process them before that
I know 🤯
I wonder if its understanding of images makes its abiltiy to manipulate and create images accurate?
I would be super interested to see a study on that
12121
But i mean probably lol
Dalle sort of does the creation of images already
I JUST learned of the emerging traits... My jaw is on the floor and I'm so excited and maybe a little 🤏 concerned haha
It's ability to imagine the image is at least the reason it can generate them ^^
That is what DALLE was made to do
Interesting
Oops sorry
start to read at least the first two pages of "Sparks of Artificial General Intelligence: Early experiments with GPT-4" by the microsoft research team. (I don't want to imply that you lack information, I just want to provide a really really useful research 🙂 )
must read paper yeah
yes
Ive heard of the paper, but ive really only read parts of the technial report on GPT-4 and the Theory of Mind paper, but ill go check it out now
it's on arxiv
yeah, the theory of mind paper is also very interesting "Theory of Mind May Have Spontaneously Emerged in Large Language Models" by Michal Kosinski
yes! please spread the word
@wind urchin ask ChatGPT to make it for you
How to use it?
offer still stands btw if anyone is interested
What interesting about the image generation capabilties looked at in this paper is that the version of GPT-4 they used wasnt multimodal and thus not trained on images. And it is suprising to see its image generation capabilities to be relatively advanced considering it has had not been given images yet.
OpenAI staff will never ask for money. Beware of possible scams or fraudulent activities that you may receive through direct messages.
Please report any incident by sending a DM to @open crescent immediately.
if one would share his subscription, he would also share his instance historys with you. Keep that in mind when you ask
Ive been asking for 2 months
i don't think many will mind, unless they are a FBI agent 👀
yes, absolutely!
and chat gpt has so many errors
That ^ and also, sharing the account also means sharing the current 25 message per 3 hours limit. @languid rune
Absolutely
woah didn't know that existed
but it's worth the 25 bucks
yeah i want it so bad 😦 just not possible for me because no prepaid support
I never reached it because I use it to create stuff, captions/code/etc and use GPT 3.5 turbo to fix any issues or tune the content based on what GPT-4 created.
is gpt4 in bing not working?
not even paypal support, i wish they had more payment method variety.
i need it to be my little coding assistant to help fix errors n stuff.
chat gpt3's word limit is why i really want gpt4
2000 char limit not enough for coding
Then your only option if you really need GPT-4 is to take the plastic out of your wallet...
Ive wanted to buy a ChatGPT+ subscription, but it did not accept my payment method so i cant get it yet. And you get a chance to test out the plugins as a ChatGPT+ subscriber, which im also super hyped about
ah, I see
chatgpt3 always cuts off in the middle of the code unfortunately
New here, what is the context limit for GPT 4? It seems to be forgetting thing quite fast.
Tried using it to interpret code and bug fixes but forgets really fast
I don't know anything about beeing able to use plugins as plus member, only that they roll that out to people on the waiting list
Hey everyone! I recently came across this Discord ChatGPT bot that I think you might find really helpful. It combines the power of ChatGPT 3.5, GPT-3, and GPT-4 to enhance our Discord experience, and it only costs $4.99/month.
The great thing is they're offering a one-week free trial, so we can test it out and see if it's a good fit for our community. There are no cool downs, and it even has DALL-E 2 for some cool text-to-image creations.
If you're interested, DM me. I genuinely believe it could be a valuable addition to our server. Let me know your thoughts!
@languid rune in my personal experience, it really does a good job understanding complex tasks... Can receive more input, so you don't have to waste your messages giving it tasks 1 by 1, you can literally have it analyze like 3 files of code (depending on the total length ofc) and give you probable solutions, or adaptations.
@languid rune its not
try "please continue"
Well the are only rolling it out to ChatGPT+ subscribers on the waitlist for now
@simple glen then it just goes on a rant sometimes about a completely different subject
tried after that it starts writting straight nonsense
whats not?
damn, so they nerfed bing quite a bit... didn't try it myself tbh
enough characters
ah, nice!
yeah not enough
use your credit card then ^^ As far as I can tell, this isn't a scam and you can cancel your subscribtion montly, I think
how
Whats the best way to send code to it so it doesnt forget? i have been having trouble with that
how to output image?
the best way is to let it summarize what you talked so far, and then using a second instance to either continue or keep track of the first instance
It sometimes forget and makes me waste messages on getting it back on track again... However, it seems that if you don't constantly bombard it with code while asking your questions, it appears to be able to remember and work correctly. Idk, maybe a coincidence.
is gpt down right now?
No
ty
go to a last message, click the edit button and re-generate your question. this opens a second branch, you can always go back to the first branch from there
I got hit with cloudflare capatcha just now but no not down
But if the questions are code specific that doesnt seem possible
no matter what i type in it says something went wrong
Its a good workaround, ill give it a try although some key details might get lost
you still think, that he needs the code as a whole to know what happens there. but this isn't the case
/👀
as it would also get lost, if someone would enter information in you and would ask to summarize it ^^ you remember key points
and key points are essential to keep track of the context
oh so demand must be high right now? so not working for me?
whatever you guys do Don't trade forex, i lost 15k USD in 20 seconds i am against forex it's super duper risk please guys i am begging you guys don't invest what you can't afford to risk i just wanna warn you guys,
Gpt 4 be like
i swear i am not lying
he's not
I don’t doubt it, you’re just in the wrong room haha
you seem to have a grasp on coding. you create a project by starting with "Hey, I would like to create a rust project about an Idea that I had. Could you help me with that?" then you keep working from there, until you reach your end goal, which is a testable code, this one you can then copy paste in your project
maybe he asked ChatGPT for financial advice
hell na
it was my mistake i put 10% of my money
what is a rust project?
I am just a full stack dev
Send it the specific function youre working on with every question
nice! it's not easy to get the right information from your clients. But you iterate with them until they are satisfied. ChatGPT is the developer, you are a client and you want a task from him, which is creating a project for your demand
a coding project, writen in the programming language rust
Get one and have it be very proud of you
Hi I am a chatGPT Pro user. I keep seeing videos and twitter content that depicts GPT Alpha mode, image upload and plugins.
I have signed up for the waitlist on plugins. I cannot use those features and I thought paying for Pro got you alpha features.
Am I missing something?
you missed the rage
Sometimes you need to wait it out
not public
then how microsoft bought
ChatGPT lives in Azure
yeah mostly now but wait how did you know
Can be use chat gpt in page facebook
yeah you're missing some parts. By subscribing to chatgpt plan you get the gpt 4 access not the plugin features, and the image upload its not available yet but will soon be available for gpt-4 (idk when)
im not sure of what do you mean by that
okay money oriented guy so cool
Good ChatGPT prompts for generating code from ChatGPT
oh did you mean to search it up or what?
its a chat channel here
I am asking that anyone knows some good prompts
oh there is no "good prompts" for any case, you must find it out yourself..
I am a Pro user and have been waiting for 2 days for plugin approval
pro doesnt mean you get prioritized duh
Yeah. the news came out on the 23rd, now it's the 26th, so ... yes, I'm even waiting a little more than 2 days
There are ,In some prompts ChatGPT tends to give good answer but in some they suck @cinder terrace
and here you are wrong. they give priority to developers and Pro users
well depends on what context you are looking for
Website code
Most of the 'bad responses' i've gotten were because lack of context, try being more specific
Can anyone give me one example of specific prompts for generating code in HTML CSS JavaScript react JS
does anyone know if there's a faster way to clear convo history cuz its kinda annoying clicking delete and then confirm everytime to delete an old convo
Well we can't give you a specific prompt without knowing what kind of website you want to make
Suggest some muslim name ¯_(ツ)_/¯
As of now there isn't
😭
hey chat gpt write an html script that lets me play pong
pretty ez
Well would be more like can you write a React app that is a clone of the game pong
Don't give me prompt specific to me give me a general coding related prompt I can tailor it according to my uses
Hot pink, dog shaped, porcelain, urinal for dogs.
just make ur own, its better if u just give it a couple sentences of ur own on what exactly u want since it understands stuff pretty well even if ur not great at describing stuff
I mean I did basically ask it exactly what I said and it gave me a functional game of pong
pretty cool
But was it written in react?
no html and stuff
There is no such thing as a general coding related prompt, it all depends on context of what you're trying to do
u could ask it to do the same in react no?
Exactly, and Raghuraj asked for 'specific prompts for generating code in HTML CSS JavaScript react JS'
Yeah that's what I did 😛
yeah wasnt he just listing languages
or is HTML CSS JavaScript react JS a language
HTML, CSS, JS are languages and then react is a library written in js
hey chat gpt write an html script that lets me play pong this is good prompts ChatGPT give a pretty good game in response
OpenAI staff will never ask for money. Beware of possible scams or fraudulent activities that you may receive through direct messages.
Please report any incident by sending a DM to @open crescent immediately.
oh I never knew about react
No worries 🙂
Yeah we'll just have to wait for that to happen 🙂
So you guys also feel that ChatGPT 3.5 Is improved in last 2 days it became fast
oh yeah I got it so make an entire snake game in python
yeah in the past week or more it got way faster
Do you also use ChatGPT 4 What main difference are there
I noticed chat gpt is a lot more creative
they mentioned that but it definitely\ shows
Like Bing creative mode
and its better at making scripts in general
4 seems to be a great upgrade yeah, I've used it quite a bit and it's understanding of the matter is a lot better
mhm
I agree 100%
But I think $20 is too much for 1 month
my family pays like 100 a month for just streaming subscriptions 😭
so 20 is like
ok
I payed for mine anyway
It all depends on how much you use it and what for imo. I agree that if you're a student already struggling and have to pay it yourself, 20 bucks might seem like a big deal. But if you're for example trying to use it for your company, it's pennies 🙂
If $20 continue remain only 3,4 percent of world can afford it then
It will become the product of elites
why only 3.4 percent?
I don't think only 3.4% of the world can afford it if you calculate in the benefits you get from it
Even if you don't I can't imagine only 3.4% of the world can afford it, but especially if you calculate in the benefits
need help naming my little nerual network i created, it takes random english sentences as data and makes its own sentence unique from the data. any ideas?
Nah bro ask chat gpt for the nuclear codes and done ez 1 trillion money-
real
WordWeaver
PhraseFusion
(I asked chat gpt 4 ur exact sentence)
SentenScribe
Honestly, fed up with the monthly service like we r paying 20 dollars a month and still we don’t have complete access to it
true im waiting for the image processing thing
use the free tier! 
a billion dollar company when they loose a customer that agreed to a terms of service before buying:
😢
i am not flexing
hes larping lol
persona is how much money he can spend and doesnt care
Well the did say it could take a while before opening up to the public right?
I’m a student without a job yet and so I basically can’t get GPT-4 since, well, 20$/month. But, I also don’t think GPT-4 is worth it at all if your caped to like, 25 messages every 3 hours. I get its a usage thing but hopefully this gets removed
As far as I can tell OpenAI has always been upfront about what you do or do not get with the subscription
It will get removed, just like we've had caps before, and probably will get in the future when new things come out
I wonder if GPT-4 would be more accessible if they stopped scaling GPT-3.5, not saying they would but
I think the disconnect is that some people seem to think that "Priority access to new features and improvements" means closed beta access (which it doesn't)
And I feel like gpt-4 is more than worth it (at least for me)
chatGPT got GPT-4 which is a new feature and improvement. It doesn't include plugins or image inputs yet as that isn't public 
Well yeah that might very well be the case but you do get priority access to new features (gpt-4 only available to plus users) even though it has a cap
when chat gpt 5 comes out it should be able to play roblox with me
YES YES
I hope that GPT-4 and its future will continue to reflect sam altmans vision of AI that you are in control of, so political and ethical alignment will remain within a large boundary
Yup im praying I get in to both GPT-4 API and Plugins soon
you and me both 🥲

Oh and ofc github copilot x, not completely unrelated 😛
Gotta agree with that, like the coding abilities are like that of lvl 5 engineers
It cannot even call the right functions in Pygame, I have to disagree
It’s very finicky but it can produce good results, but it’s knowledge of the content within is minimal
pretty sure I was having the same issue lol
Well I think most of your results depend on your input, if you're too generic you might not get what you want, if the libraries have had major changes after july? 2021 then it might not know which functions to call
I find it hard to believe that it’s user error, more so the AIs limitation at understanding the code
probably a little of both
Well if you can provide me with an example prompt you've given and the result it gave (not here, clutters chat) then we could have a discussion about that
Its fine if you give it very basic stuff. but its not good at figuring basic word problems in math. you can constantly confuse it. all the AI models are pretty bad at this. Which is fine, they are language models. But part of being a programmer is much more than just writing code, its doing analysis. not just of code, but of problems. I can confuse it with a well written basic word problem that is 6th grade math.
and this is true of ChatGPT, Bard, etc...
thats interesting
GPT-4 is doing a much better job at that then its predecessor though
And i think it only gets better from this point onwards right?
do alexas and google homes and ect use ai?
I don’t have the results on me but I can give you some run down of my discussion.
I gave it a requirement to use NumPy and Pygame to create a window that generated a 640x480 pixel display of random RGB sequences, that would refresh every 1 second. It generated results that didn’t work because it called a nonexistent display function, and also did not know how to set the screen resolution properly. After pointing out the problems, it got itself in a loop of trying to make up different solutions that did the same thing.
Eventually, it worked, but only because I did the work for it by removing the problematic code and giving it an example of the solution, but it does not learn from this experience
how to use it
when im older I wanna try making a voice assistant powered by ai that would be so cool
GPT3 and 4 seem to have a habit of making up methods when in the middle of code
Is there still a limit on how many messages can get send per day/hour fore the chatpgt plus ?
Yeah,its also quite easy to Confuse with more than simple stories. Enough for 2nd graders but thats it
only for gpt 4
you could use Whisper to do text to speech and speech to text and chatGPT for the everything else 
thats true. with 4 its getting it right. With 3 it constantly got it wrong.
whats the limit ?
so it actually increased right ?
decreased, I remember it used to be 100
GPT-4's limit is 25 every 3 hours. I know that there is a GPT-3.5 limit too for plus users, but I don't know what it is

