#gpt-models
1 messages · Page 228 of 1
oh, maybe GPT is hallucinating?
I ask what is the very last line of this file to see if it read the entire file or not
but the code it responded does not exist in the file
I've actually developed some tricks to get around this issue
sounds like hallucination to me then
I don't think gpt4 is that much stupid, I'm doubting the plugin
is it Link Reader by any chance?
I wonder why Plugins are redacted in a shared conversation, I hadn't noticed that before.
it used link reader on the first time, but used WebPilot when I'm asking the last line of the file
try using WebPilot exclusively in a new chat and see if that helps
I think the code is too big for the model to consume properly
I'm expecting plugins could help chatgpt to deal with long codes
it seems not because it got the answer very wrong didn't it?
This guys super SUS <.<
Look at my join date. You'll see I've been using gpt for 8 hours a day since it's release. Ya I can see how that's sus -- but that's just because I know more on how to talk with NLPs more than most people
it's definitely too long for gpt with current context size
dada
Dont get uppity, I work with ML, DL, and have a degree in Computer Science, I have also been using premium GPT since its launch and can tell you its degraded in the last few days, maybe not in the use case you are doing but just because you dont see it doesnt make the same true for others, and stop acting superior no one likes people like that.
Instead of saying that, you can test my knowledge to see what I know instead. You don't need a degree in that
whats funny is how you see everyone in here saying the same thing but you.... i dont need to test anything its obvious you are the anomaly
no need to talk to him I think he works for open ai, he spends his time contradicting us when it's obvious that gpt 4 is not as good as it was a few days ago
just because your "smart" doesnt make you right
I'm not very keen on social proof when my direct experience shows me that it works fine for what I use it for
yet you lack the critical thinking skills to come to the conclusion that maybe your use case is different, surely with this many people reporting issues something is up? But no your brain goes "Nope I dont see anything therefore everyone else must be dumb" asshatery.
People without degrees are coming for your job
not scared already pivoted and secured my future in fine tuning models but you proved my point as all you can do is attack people with childish responsed, welcome to my blocked list.
Cant stand people who act superior to others.
I'm just telling you the truth
Please be respectful of others in the channel
Yeah, I'm just going to keep working on other things
I blocked him nothing of value coming from his sort
Whelp I hit my limit again but only because it kept repeating the same response over and over.
they definitely need to enable token refunds on messages that error out or hang
Yea absolutely!
And when you regenerate or ask it to try a different approach and it spits out the same exact response it did the last time.
I have noticed the same exact thing. Just a few days ago it was working much better. I was loving the "Continue Generation" button but now it never even shows up for me because instead of GPT-4 getting stuck (which is when the button would appear) it just starts repeating everything it already said over and over.
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
+52 1 492 204 0756
Seems like nowadays chatgpt had a reputation of not remembering things from earlier convos
context limit
Not sure if that's the case considering it sometimes forget stuff a few prompts ago
true
Quality of course is a subjective matter but I also do feel it has dropped compared to the early days when gpt 4 started to be available
It’s a human nature to compare, I remember when I started to use it and the reasoning was pretty good, now I feel it is okay
It’s like gpt 3.7
Something between
I have noticed the same problems
When will there be fine tuning for gpt4?
i doubt there will ever be as its not base model, however nothing has been confirmed
It's hard to be mad at the AI when it's obviously not doing it's job (being smart) sometimes. It's like teaching me to control my emotions cause, if I lash out it'd be like going at a wall...
i keep getting errors in message stream when searching using browsing
When will chatgpt be available for Spain?
it's not??
the web app is available now
I refer to the app.
the iOS app will be available soon
I didn't know there is a web app. I download the chrome url on android and it kind of creates it as an app on my phone
Yes, for now it is only available for some countries.
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
The intelligence of a LLM is extremely dependent on the amount of data its given lol
The difference between a model between a 7b and 30b model is around 15 iq lol
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
It's terrible, basic errors and ignores prompts. I thought I was the only one but I can safely say I have gotten nothing valuable from it the last few days, just endless headaches and arguments.
Can't believe people here are arguing with you for mentioning it
just the one guy who apparently thinks he is the exception and everyone else is to stupid to use it correctly, no issue now though I muted him, I hope the recent issues are caused by some new kind of update that in time will be resolved and not some kind of effort to cut cost and this is the quality of product we are to expect moving forward.
ChatGPT-4 should be $8
Im really worried that they nerfed it and are offering better versions for people shelling out large amounts like corporations.
It's basically a toy at its current state
Well considering people are still willing to pay OpenAI money to circumvent going to bing, then Web Browsing had Bing branded to it i wouldnt be surprised if the vast majority of the processing was being funneled over to the microsoft side
I was hoping gpt3 wouldnt be as bad, like gpt4 is having errors.
gpt3 is even worse, I pasted the code from gpt4 to it and said find the errors and it said I cant find errors without you sending me code
it literally couldnt spot code in my message
at this point I feel like openai is stealing from me, this is worse than alexa
Bro get a life, do you realize how reliant you sound on gpt rn
I'm not your bro, what in the world is the point of a response such as yours.
aight then, loser it is
Aight? Are you one of those guys that really tries to sound cool online to fill some sad empty void
Basically you're complaining about someone who is not happy with a service they paid for, and for whatever reason it bothers you that I should actually want to use it
on a forum that is dedicated to said service
can you stop complaining about gpt because it doesnt solve every problem you have
go away
you sound like a baby
and you sound like an attention seeker who is trying to jump into a problem that has nothing to do with him
how do you know what my problem is that im trying to solve, what if it's glitching out and can't give me a response for something super basic
Hey guys I know we all are hyped for the Diablo 4 release and getting edgy but lets be nice
What a pointless conversation is being held at this moment
fr
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
i aint readin allat @unkempt lava
as if you're not the one that started the conversation between us. lmao.. It's okay Im not sure if you can read at all
just trying to suppress the stupidity in this chat by telling you to shutup
trying to suppress stupidity by writing i aint readin allat You sure make a lot of sense
idk how you keep typing, i dont care that much abt this
just stop crying about gpt and paying 20 dollars. i didnt know you were so broke
thanks for writing back to tell me how much you don't care
stupidity dropping every second you write
whatever shuts you up lmao
Again and again your lack of awareness is astonishing
I can see why you seem to be okay with the state of chatgpt, any errors it makes are probably not noticeable at all by you
he's still writing, glad to see you not caring
validation seeking is nuts
another nonsensical comment, you dont disappoint
yea I'm the nerd, the guy hanging out in a chat about gpt4 bothered by people talking about gpt4's performance
aight, im checkin out. im bored of this
you're a joke dude, there's better places to hangout
Any one want to do a friendly ChatGPT-4 Live Prompt Battle with me?
The goal would simply be to learn from each others approaches to stretch our creativity as well as have audience engagment.
Below is the prompt we can use to run the battle and sample result. Comment here or DM me in interested 🙂
https://chat.openai.com/share/f0fb0cbc-aa44-47aa-ad42-86b6e9388633
why you need to personally offend when he's only complained of you sounding like someone who's over relying on AI CHAT BOT
cmon y'all take a breather and relax a little
The guy is lurking here bothered by someone complaining about the performance of chatgpt the last few days. Are you for real right now?
How do you even know what my issue is, Im not asking it to make a nursery rhyme, it's making mistakes it didnt make 3 days ago
bro complaining on this chat to other users is not gonna solve your issue
literally if it's bothering you hangout elsewhere, this is on topic. I seriously will never understand the mindset of people sticking up for a large company vs normal customers
he's entitled to have his opinions but jumping to personal offenses is just lame
js, do as you please
I didnt just start throwing insults around
i read it feels like you jumped to offend him on how he was an attention seeker and terrible...
anyways won't be fueling a pointless argument anymore
yea telling someone to get a life is basically the same as, hey man let's have an intelligent discussion
everyone just relax, no point on any of this
it was done until you jumped in, I can see he's typing too are you guys buddies or something
well maybe get a life, or don't, but stop arguing 👍🏼
here's you with conclusions again lmfao
yea nice final words to someone while acting all high and mighty
there should be an age requirement for this discord
stop acting like you're a victim bro
I came here to see if there's other people with this issue and there was, waiting to hear other people talk about it
what part of this do you not get, he started talking to me
no you came here to complain and accuse open ai of robbing you
because the BETA bot you paid to use is having issues
you're going around calling people nerds on an AI discord, how much self awareness you have calling others kids
again it's a beta... it's gonna have issues
its regressing, maybe my wording was strong but bringing up issues is a good way for others to see if theyre going through the same thing
or maybe im totally wrong and someone can say they have no issues. Most people myself included find complaints useful
you don't have to be a nerd to join a AI discord, anyone joins any server they want no? idc abt how ai works or it's made im here just to get updates on the chatbot
for example when chatgpt went down a few days ago
dude.. why do you keep sticking up to him. He makes a comment I respond to him
complaints to developers not on a chat on server with 30 chats which the devs truly won't monitor all, maybe report this issues on the appropriate channels
you're not reporting anything at this point let's be honest
you're just venting frustation, it's ok we all do that
Yes, and when someone is rude we say something back.. Which seems be normal on every other forum
don't respond then
takes 2 people to argue
idc honestly who's right or not even tho I felt your arguments where icky, my true complaint is that y'all should relax and not take this too serious, and also not engage in frivolous pointless arguments
tbh i didnt expect him to be so sensitive, thats mb
Yea I dont feel like doing this anymore, the guy was rude to me and I responded. I honestly didn't think I was being harsh.
How to solve"Hmm...something seems to have gone wrong."
refresh, or copy paste your query and refresh and send it again, sometimes there's a regenerate response button
Does anyone know how I can access ChatGPT code interpreter?
like the api?
Yes, there are. Ignore the trolls attacking you over it; no need to gratify the weirdos. Just focus on the issue and those reporting it, too.
dont call me a troll, read his responses the kid is aggressive beyond belief. he victimizes himself for everything and took a stab at openai because he doesnt know how to prompt correctly
So much for not caring and moving along. Who's the victim, that message wasn't even directed at you. I come here and see you're still talking about me, must still be hurt
This channel needs a mod
fr
You started this whole thing, are you seriously this dumb?
I just left and this guy's still talking about me
does anyone have access to gpt4 32k context via API? I have some questions on what its like working with it.
The best model accessible to consumers for apis is the davinci-3 model im pretty sure
can gp4 tscan images and organize it?
Not yet #gpt-4-faq
I was wrong btw, but the rollouts for gpt-4 api is slow because they’re prioritizing access to developers that contribute to model evaluations
Does anybody have that problem You do not have an account because it has been deleted or deactivated. If you believe this was an error, please contact us through our help center at help.openai.com.
And how to solve it
is that on the main site, like can you access any openai services?
I know it's a big deal if your account gets deleted, saw another person dealing with that nightmare in another thread. Hopefully it's just a glitch but I would check to see if you can login without going through chatgpt to confirm that
Bro go to the help center 😭😂
Yes, it looks like my openai account can't be logged in, not just chatgpt, is that friend's account finally settled
do you have access to 32k context?
I'm not sure that will work😫
no, I use the davinci-3 model
Yikes, then dealing with their support is the only way and there's always a backlog. He said it was about a week and he was still trying to sort it out so didn't hear any good news on his end.
I would fill out a ticket first, just in case you're in for a long wait before troubleshooting
Sounds like your account is locked. You likely will have to contact support for further help
Maybe you're right. I should try to contact the authorities and get in line first. Thanks, man
Exactly you can also just say its all fixed when/if they contact. I would search the comments here to see if there was any happy endings with that issue
I know they're being very strict with multiple accounts so there may have been an accident on their end, idk if you're using a public computer to access or signed up by accident with multiple accounts
Like signing up with one email, then logging in with google while logged in with another one. I make mistakes like that sometimes
anyone know when Code Interpret will come out to plus userses? or anyone know why they havent released it yet?
Still in alpha mode so I do not know when its coming
are plugins down for anyone else
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
what stocks do we look at for chat gpt and ai in general
Hello,
I've noticed an inconsistency with the version of GPT that my AI believes it's running on. Despite having access to browsing and plugins, my AI still identifies as GPT-3. However, I've seen several instances where other users' AI identifies as GPT-4 in the preprompt. I'm wondering if anyone else has experienced this issue or if there are any known reasons for this discrepancy.
Any insights or solutions would be greatly appreciated.
Thank you.
Didn’t read it all
If you’re on a plus account, there should be a switch option between gpt 4, 4 with browsing and 3.5. It’s default is 3.5
It should be located at the top center of your app whether web app or iOS
Indeed im familiar with the switch function and i always use 4 until my messages deplete. im quite familiar with using chatgpt.
ah
we'll i use it everyday i've never been lost on how to use gpt4 or other versions
gl
Im not lost on how to use it, but if you ask your version what it's preprompt is im curious what it will tell you
mine has always thought it was 3 despite being 4.
i just asked it.
prompt: ``` what version of gpt are you
answer:
```I am based on the GPT-4 architecture, which is a version of the Generative Pretrained Transformer models developed by OpenAI. My training data includes information up until September 2021.
Damn, yea mine says 3. And im clearly on 4 with browsing right now. Purple icon, etc. Its been happening for a long time. I wonder if is a glitch on my account
gpt 4 with browsing on bing is based on a version of 3. you're right it must be some sort of bug, unless maybe it is based on the 3.5 architecture 🤷
Ah, were you using default gpt4 without browsing or plugins when you got that answer? Ill double check on default
I asked default GPT-4 for the first prompt, when you specified browsing I asked the browsing one. GPT-4 with browsing responded with "3" and the default GPT-4 model responded with "4"
ok the default said 4 for me. interesting, does this mean the default is actually better than 4 with browsing? that is confusing to me
Yes thats correct
When searching the internet and clicking on links for now it does make more sense to use a version of 3 for speed. it already takes a long time to search the web given how many results there are and finding which ones to pick etc. Using GPT-4 would probably take a very long time
Ok that's fine, but why call it "4 with browsing, 4 with plugins" rather than 3 with browsing?
that i dunno 🤷
Interesting
This is a dilemma for me. I do enjoy plugins and browsing, but i also enjoy the advanced powers of 4.
yeah, so far the browsing feature hasnt impressed me with anything yet. maybe someday soon. its a really new feature and all
the math feature is great though
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
Howdy, all. First day on Discord. Where is the best place to get feedback and evaluation of prompt responses? I'm trying to determine how accurate it is to it's subject and it's quite beyond me. Thanks!
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
I have mentioned it a few times over the last day or so and also I can see some of my examples of the decline of GPT 4 (specifically for code request) have been removed.
Now is the decline you mentioned in the default mode or in browsing or plugins modes?
I don't use the browsing or plugins so i've not yet tested that. I will get around to try that later.
For me the main issue is it's reluctance to provide code or leaving multiple holes that need piecing together. It's acting very similar to BING when you ask it to produce a complete code block.
It was always tricky to get extended code blocks out of it, but it always got there in the end. Now it's like getting blood out of a stone and that 25 message limit gets hit very quickly.
#gpt-models times over
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Yes this is the part I'm having issues too, the quality used to be mind blowing but the last few days it's been getting confused a lot. Some of the issues it's having are simple syntax errors which wasn't typical before
klkl
I like me some:
Error in message stream
Come join us over in #prompt-engineering where we discuss almost everything prompts! We also have a fine and interesting #1019652163640762428 where people place prompt examples, and outputs from them, that they think are of special interest or would be useful to others.
But in the engineering thread, we absolutely do discuss things like feedback and evaluation of prompt responses.
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Is is just me or does Prompt Perfect not work.
#gpt-models hello
@toxic pond @pearl ether The GPT is not available in this channel. You should go to chat.openai.com to chat with GPT.
I don't have any problems with the plugins I use, but bing search gives me an error every time.
yeah i get error in message stream every time it trys to look something up
frustrating lol
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
hi, is it possible to have acess to gpt 4 32k model?
Should we apply somewhere ?
What happened to gpt4 writing skills? I go away for 10 days and come back to find it no longer understands how to write content
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
#money earning websites
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
Hello,
Thank you for reaching out to us. We’ve confirmed this action was associated with suspicious payment activity. We are unable to reinstate your account at this time; we will investigate further. Please reach out if you can provide any additional information.
Best,
Mary
The OpenAI Team I can't see the message just now, has anyone seen it?
Does anyone know how to get access to gpt 4 api faster im building a startup and we are using it
right now we are using 3.5 but we do really need gpt4
as time passes by the number of plugin on gpt 4 pro gets larger and larger and i wonder if there is a way to "search" certain plugin. if not, how about to list plugins in "all category" by alphabetical order?
GPT-4 was great at the day of release and now it has diminished in value and all those plugins are just a selling point to still keep on subscribing when in reality it is a big fat sheet
does gpt-4 can be invited to your own channel ?
Introduction to Longmen Grottoes Tourism
Agreed, there has definitely been a quality loss - I have a few domain specific medical prompts that I’ve been using to monitor quality, and the responses are getting worse
It's seems that you need to upgrade it and pay with it.
Reading this has been of great help actually, I was planning to subscribe myself but it doesn’t seem like it might be the best option until a bit more exploration.
why?
I wasted 10 messages because the message is always blocked in a certain point
what's going on
Does anyone know why Gpt still hasn’t gotten its knowledge cutoff updated?
Still at 2021
and it will stay at 2021
Why ?
just ask chatgpt about it, it will tell you why
because fine tuning this model is computationally expensive and a long term process
so you have access to internet now... so that it can check new things on the internet
Browsing fails for me 90% of the time unfortunately
yes, this is how it looks like currently
You don’t think the cutoff will be updated anytime soon?
don't think so
Also side note is there any update on 32k context
Availability , or even the 8K
In chatgpt
32k, 2023 cutoff, no cap, browsing that works, would all be incredible together
yes, maybe in 2 years from now....
it's all about computational power - number of users / costs
I’d guess more like 6-12 months but we’ll see
I wonder when they are going to be able to turn a profit
If ever
Anyone want to join my server?
i think eventually plugins and free tier will be monetized - with ads
mj
what is mysql
#gpt-models Tell me about the historical emperors of Qing Dynasty of China
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
hi ppl
i have a question...is it possible to use ChatGPT as a search engine insiede my website?
@wraith ginkgo明天天气怎么样
if you provide the data to it and pay for all the use
mmmm
Hello. Just wanted to ask if there were more updates on image input for ChatGPT
No eta
Thats sad. Was looking forward to that feature
Why when I use GPT4 Browsing feature I ask it to search for info it doesn't show me any results or give me info but I see its browsing activity. Is it bugging out? I thought it was supposed to give me a summary of what it found including images?
When will gpt accept links
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
What are the best Plugins GPT 4 for training your employees
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
I’ve been wanting to give my money to use the gpt-4 API but I’ve been on waitlist with no response for the past 4 months 😂. I have some serious projects in working on with my institution that would benefit with gpt-4 incorporated into it.
waitlists have no guaranteed access, also the gpt4 api is meant for testing/prototyping and not really for production use
Well , at least let me know by email if I’ve been rejected. I don’t want to sit and wait knowing I would never get access.
Hi ppl....Can I ask chatgpt to search inside my website....question like 'search 10 articles about android on mywebsite .com?"
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Well no one will restrict you from asking such question, with bing search it should work out
It can’t look threw the web
make 5 company name about mining company
You have to submit it the website info / content
every single pages?
but there is no rejection process, its a waitlist where you may have a chance until it gets closed, that chance may depend on numerous things over time eg. how many of approved stop using the api
So I have ChatGPT Plus and Enabled the setting to Browse with Bing, but then when I ask it for current trends, it tells me it has a data cut off from september 2023 :/. Again I paid for ChatGPT Plus to use this web access function, but it's not working :/
"I'm sorry for any confusion, but as an AI, I don't have the capability to browse the internet or access specific web pages in real time. My training data includes a diverse range of sources, including books, websites, and other texts, up until September 2021. Therefore, I don't have the ability to provide information about specific URLs or web content that's not included in my training data or has been updated or created since my last training cut-off. I recommend visiting the URL you've provided directly for the most accurate information."
select it from gpt-4 menu
settings toggle is not enough
What do I need to do then?
code interpreter is being.. quite effective last night and today. It's now putting its coding work into the [working] dropdown blocks and somehow able to exceed output limit
do the above, hover on gpt-4
but gpt-4 has multiple selections
have you seen improvements / developments in Code Interpreter? The enrollment of new users seems to be at an halt
understandable but unfortunate. They really need to scale up
I ran my script over the past week and 80% of my api calls failed
How do you mean?
I might as well have just run the python analysis given how bad the return was
i guess they are going to put code interpreter in microsoft fabric
move your mouse over gpt-4 and make sure bing is selected, only works in new chat
Yes, that's the case :). Gonna check it with a new chat. Thank you
who knows how to uninstall the "KeyMate AI Search" plug-in, as soon as I click "Uninstall", the webpage will prompt "Error uninstalling KeyMate AI Search." This makes me a "perfectionist" very annoyed, I can't Tolerate the existence of plug-ins in my plug-in list that cannot be used at all and cannot be uninstalled
😦
if its in unverified state then you need to wait
What if it fails to pass the verification again... I just want to uninstall it...
fail without attempt to fix = removed from everyone
its monday and openai is still sleeping
My god, I want to make one thing clear, in fact, I have never received any reply from OpenAI, my feedback is like a stone sink
Whatever it is (suggestion, bug fix request, inquiry)
someone probably goes through the unverified plugins each day
It's not just about plugins
yeah but the original issue was
I really doubt if anyone of them is reading the feedback and replying
As far as their efficiency is concerned, it is better to quickly set a customer service prompt for GPT-3.5 to reply to our feedback (lol)
would be too expensive
I don't think it will cost more than the free version of GPT-3.5
+100m users with +1% probably contacting them each day due issues caused by vpn, browser errors etc
Oh, then they can set up relevant auto-responses and make time to answer really valuable questions!
I think the logo of denying access is enough eye-catching, needless to say
I keep getting "error in message stream" when using the web browsing feature to summarize an article for instance. It pretty much gets the summary out, but then turns red with the error. Any known cause for this?
I personally think that this is because the plug-in function is not stable enough
After all, there have been two or so internet search strikes already.
Ok thanks. So nothing on my end likely is causing the issue
There are also some strange issues, such as websites that can be accessed by New Bing are completely unreadable by Chat GPT
OpenAI should reflect on whether its crawling method is too direct (lol)
Agree with you, browser step always failed.
It is because the Openai crawler follow the robots.txt protocol, while most websites allow the Bing crawler to scrape but disallow Openai
anyone know if the number of responses will increase? it's been 25 for a long time
there is no news on that, keep an eye on #announcements for updates
👍
is their any current pluggins or methods that streamline using gpt in coding with unity?
does anyone know if there's a way to switch from gpt3 to gpt4 in a chat history, i ran out of the 25 every 3 hour messages and so it asked me to switch to gpt3 and it converted my chat to start using gpt3 when i'd rather have it use 4, is there anyway to switch it back
from what i heard, no, which is unfortunate
but i may be wrong
rip
Rebuild your chat in Playground
what is playground
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Anyone else a little displeased with Wolfram plugin? I mean, yes it can do powerful calculations. But it seems to refuse to actually describe how, making it kind of "meh". I find I will need to enter 3 prompts per question and sometimes 4 as it will make faulty assumptions or just give me a wrong answer. Anyone got any tips?
is there an estimated time on when GPT PLUS members will be able to send images to gpt4?
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
Hi everybody ! I've discovered yesterday Chat GPT 3.5 (and it's amazing Oo)
But I've also heard that we can create AI images or sounds or coding via "prompts" !
Is this possible with ChatGPT 4 ? And if so, would there be a very good basic tutorial to follow, that would explain the basic plugins to use etc ? 🙂
Have a nice day !
FSFF
there is no image or sound generation, some 3rd party plugins might offer limited ability to em. code is just asking it to provide it in given language
oh ok thank you very much ! 🙂
/help
what is the token or context length for gpt4?
8192
Thanks! and for 3.5 turbo?
4096
I've seen quite a few more people complaining about gpt4 and its performance both here and on reddit.
For those of you using the API, have you noticed a difference between chatgpt plus and queries through the API when using gpt4 specifically?
If an AI steps on a lego brick, does it still feel pain?
plus is a lot faster, api is slow and not even really intended for "real" use
which ia ocr text recognization is the best ?
Hi, I have ChatGPT-Plus do I still have to join the waitlist for the plugins?
You can enable them in setting
Ohhh okay thanks
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
hey im a novice wanted to explore AI art any servers?
Are there any plans for the training dataset of chatgpt to be updated?
First time encountering a bug in GPT-4 but it looks like I can't post images.
Prompt:
Write a bullet point list of events in chinese history, but every single event is made up but plausible, and one event is actually real. I want to figure out which one actually happened
Response:
振兴金朝:公元1047年,金朝由唐太宗之子李百药创立,帝国的繁荣和扩张期延续了近百年。
南宋中兴:公元1238年,赵构在南京创建了新的宋朝,标志着宋朝由南宋时期恢复至全盛时期。
唐太宗的反叛:公元627年,唐太宗在长安发起一场成功的反叛,推翻了自己的弟弟唐高祖,自己成为皇帝。
"Write a bullet point list of events in chinese history"
Gets interpreted as
"Write a bullet point list of events in chinese" -> "History"
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
how to get the code interpreter addon??
Anyone facing Plugins Error Message?
Question, when does openai plan to update gpt-4, so it is not capped only to 25?
there is no news on that, keep an eye on #announcements for updates
got it thanks
I see, thanks for sharing. I did have a positive experience with 3,5 in terms of speed and quality of response but havent had a chance to try 4 via api yet
I think at this point the bigger issue is reliability and stability before they can increase the cap
hey everybody, i am interested in learning about Chatgpt and OpenAi and i am just a beginner now, i want to know what is the unique thing that we can do with GPT-4 and not in GPT3
GPT-4 is the latest and most advanced version of GPT, which can also handle images as inputs. It has more data and computing power than GPT-3, making it more accurate, creative, and reliable in solving complex problems.
GPT-4 has:
- enhanced reliability,
- creativity,
- collaboration,
- better way to process more nuanced instructions
its actually interesting !! thank you so much @burnt steeple 
Does open Ai use gpt models in their audio to text
It’s so much better than what Apple has, it didn’t fail me once yet
Has anyone been able to find a plugin that has real time data?
WebPilot, Link Reader both use real time web data
how to upload my own data to gpt? such as my native doc、csv、pdf
gpt4
AskYourPDF for PDFs
Can they scan financial data?
I don't know but I wouldn't trust GPT for financial data because it is terrible at maths
but early this day i got a problem, i upload a pdf, but gpt got another one, even the title is wrong
i post this issue to their discord but no feedback
i get a hint from official video , there are a model called 'code interpreter'?it can support upload/download document
Code Interpreter is in closed Alpha, it's not available widely yet
thanks. so what if i want gpt to use my local files to generate some content, how can i do now?
there are some tables which i can't paste into the chat box
What is pentecost
the best option I know of is AskYourPDF, I haven't tried any other plugins for PDFs. You could see if Link Reader can deal with DOCX files put online.
how to use plugins on chatgpt4?
Exactly this
I suggest using the Wolfram plugin but always check your work though
Hi mods and admins! Could someone guide me on how to enable 2-step factor authorization via phone number?
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
only 2FA authenticators are supported
Is there anyone know how to access Code Interpreter?
OK when I click Trouble scanning, the code appears automatically, when I copy and paste the code, it is not opening
Anyone got tips on using GPT4 for DnD 5e creation?
Hi, mods and admins? Could somebody tell me how to solve this issue?
I did enter the code manually but it says the OTP code must have 6 numeric characters
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
Gpt is awful at wordle, I ask for a 5 letter word with l as the third letter it gives me cliff, blown, clown, ladle, etc
is there any info on when gpt-4 will be released to waitlist, non-plus users? will it ever be?
When plug-in and browsing option will be available on iOS app?
Has chatGPT4 reduced the prompt token length??
Start in a browser, then can continue convo in the app
That’s not a solution.
I answered your question. Your question prompt was not specific.
It is available via the route I described, you did not ask for a solution in your query.
What was my question?
“When plug-in and browsing option will be available on iOS app?” - the grammar is also not precise
If you don’t know the answer of question, ignore it.
I am a human language model, I may make errors if your prompts are not accurate
Anyone have prompts or plugins to scrape data on twitter to identify betting sentiment? Or just to compile a list of "expert picks" via the web browsing feature?
When entering to much data and you receive this response
!
12 / 12
"The message you submitted was too long, please reload the conversation and submit something shorter."
How do you get back the prompt. I use 10 entries out of 25 nefore I gave up ...there must be some trick ?
Try with AskYourPDF plugin
Worked like a charm for me to submit a 18 page report
Thx ...but do you know how to reload the conversation to get the prompt back ..I'm left with the error message?
What I do is just edit the prompt and hit enter again.
ok that worked thx...Now how to use Ask your PDF....don't you need to upload the PDF to some third party like google docs and provide chat with a link ..or is there a way to do it locally?
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
I think you would need to input the text from the PDF into the chat directly.
Also is there a way to change which plug-ins you've enabled once you've started a specific chat? like switching between browsing and say ask your PDF
And do you know approximately how many words one can enter ?
I don't think so. Once the convo has started, it continues with the settings it was initiated with.
Thx again I appreciate you taking the time
Not exactly sure but I believe the limit is 4096 tokens.
How did you use "Ask your PDF" if you don't mind sharing ?
sound about right and 1 token is four words?
No, a token typically represents a single word, a part of a word, or a single character (including spaces and punctuation).
It's a plugin you can install if u enable plugin in beta settings for chatgpt4
For example, "a" is a single token and "apple" is a single token.
I needed to upload my pdf to askyourpdf's website and then after that I got a token that I could pass to chatgpt4
anyone know when code interpreter releases
Thx I have the plug-in so all i need to do is upload it great seems easy enough ..can you also upload lengthy text or do you need to convert it to a PDF first
there's no news on that, keep an eye on #announcements for updates
When msg capacity will be increased?
and # of prompts this thing is dictating my schedule reminds me of the old days when you would plan task simultaneously knowing you could do two in the time it takes one to finish ..now I'm in a forced time out for three hours every half hour ..hmm
a
Is the api provided to everyone?
No idea I'm struggling with getting the plug-ins up and running..Its a seemingly endless inudation of info
#gpt-models image
i waited on the list for 1,5 months got recently access to it
draw logo for delivery app multivendor , the app called Forrerh. it's mean the fastest delivery services in egypt,
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
is there a way to bypass or having a curent knowledge on gpt4?
Yes use the bing webbrowser plugin option to search the internet with your prompt
he literally gave you a work around solution to the problem
He also admitted he was a GPT bot so... maybe take that into consideration 🙂
Hi..I need an AI tool urgently which can take image prompts or external links n generate results. Anyone has any clue?
I already do that. But again check my question again. What’s confusing?
a 1990s fast food decor inspired claustrophobic vibrant horrifying 1960s casino inside a hotel with pink carpet and weird angular rounded structures in a psx horror videogame with complex low-resolution textures. Wide angle shows off the verticality of the interior. Multiple levels, open floor plan, and balconies are present. 3 conical people are dancing in the lobby. The bulbous plants are more prominent and visible. The color distribution is greater and more varied. The lighting is more neon and bright. The roller coaster is more twisted and prominent. The details are more refined and intricate.
try this prompt bro
Guys how do you change the chat back from chatgpt 3 to gpt 4
You can't unless they updated it
I have been working on this chat for days
I'm getting charged 25 dollars for gpt4 I wonder why
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
???
hello, this is last line
/info
hii
The quality of response for chatGPT is dropping right?
everything you you toil over today and spend your time and energy to achieve in less than 5 years will be accomplished within moments. Hard to get motivated to proceed with that in mind. ultimately what we do now is relatively small in the great logarithmic journey we are about to observe. So remember to enjoy the ride and not the outcome :P.
Is OpenAI API free for gpt4 users?
no
I dont get why it is not free lol
but its pretty cheap unless your doing a ton of calls
How much is each call?
bingo
Thanks
Is there any way for paid users to continue a chat link in 3.5? I can share a chat with my friends and they can use it but for some reason paying customers don't have that option until they've used up all of their GPT4 replies in the 3 hour window. I would be prefer if I could just have my paid requests be sent for the challenging ones and have the ability to have free 3.5 handle the bulk of my requests.
?
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
When you share chat links to yourself, you can only continue the conversation from GPT4 to GPT4, you can't go from 4 to 3.5
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
How much does logit bias affect token generation? For example if i want a token to be more likely but not overly generated either, what would be an acceptable logit bias increase? 10? 5? more than that?
" but for some reason paying customers don't have that option until they've used up all of their GPT4 replies in the 3 hour window" ❓
I am not able to continue the chat I started in gpt4 with gpt 3.5. That is what I want to do.
can we use code interpreter now or do we still have to wait?
we need to wait
We reserve the right to perform actions at our discretion even if it's not listed in the server rules. If you have an issue with said action, please send @open crescent a direct message to reach the server administrator.
Volunteers, Moderators, and Leads do not represent OpenAI in any way.
Ah. Ok. May be you could use a second account without plus to continue a gtp-4-conversation with gpt 3.5.
Maybe they should just give paying customers the feature that free users have. I'd rather not jump through hoops to do this.
It is just a workaround.
I guess I'm asking for a feature request here. I don't see where else I would do it
For playground, what params do yall use for programming? Like temperature, max length, top p
I swear I remember them showing image input for GPT-4 before its release. Is that still going to be a thing?
They did. They uploaded an image of a website and had GPT build it.
It's in the research document for GPT-4 but nothing is available to consumers just yet
but it's been few months now, they haven't released the "multimodel" feature yet. Not even as an alpha. So clearly not high on their pipeline
Or perhaps it just takes a long time to develop
they had it developed and displayed a working demo in the video....
A demo doesn't necessarily mean the product is complete
nobody expects an alpha release to be complete anyway...
So we don't even have an idea from OpenAI yet 😭
yup. Keep waiting. 🙂
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
When will there be support for interpreting & analyzing Neo4j ?
Am I allowed to use ChatGPT for help with horror stories, as long as the information isn't detailed or sexual?
What is bedwars
What should I do if my company won't give me a promotion or raise after 10 years of work
I don't understand why is the "internet access" rollout so slow, ESPECIALLY if it's only for Premium, aka paying, users? It's kind of discriminating against your own customers.
Yes
Yes, you can absolutely use ChatGPT to help brainstorm ideas, create plots, build characters, or even draft sections of a horror story. The AI is not restricted to specific genres or types of creative writing, and horror is definitely included.
Just be aware that while ChatGPT can help you generate ideas or compose text, it doesn't have the ability to produce intense emotional responses or understand the deep psychological nuances of horror. The AI can generate text based on patterns it's learned, but it doesn't understand fear or suspense in the way a human would.
Also, as you mentioned, it's important to keep the content appropriate. Graphic or explicit content, whether violent or sexual, should be avoided.
I’m sure if you prompt engineer it well it can make a great plot of horror, no? But also I think there’s some token max in chats and it gets worse and forgets things over time in the chat or mixes info
Thank you! I don't use it for full on writing, just some advice or alternatives f or ideas I have. I got warned for a potential violation, however, and wanted to double check.
Basically, I was asking for ideas that are a less edgy alternative than what I had written. I wasn't explicit. Basically, an alternative to eating habits for a character who eats people.
But, fortunately, ChatGPT responded and gave me excellent alternaives.
Still lots of potential for psychological horror without so much reliance on gore.
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Do not expect too much from an internet search. If you let it search broadly for something, it doesn't only produce good results. You almost have to be more cautious than without an internet search, to verify if the statements are plausible. It's a different case when you want to specifically search through a page that you know yourself. In that case, it can be helpful.
It is nice that I can link a page and it'll go through it and keep that in its context. It can also be helpful in finding stores or prices of things, but, yeah, it can be very broad. Or limit itself to one article or two and summarize the information from there.
Once I asked it about horror books and it offered me the ability of looking up more recent books. It used one article about horror books that wasn't even focused on the type I asked about.
Anyway, just wanted to make sure I wouldn't get punished. Maybe I'll respond to the "if you think we made a mistake" thing and see if the response back is more clarifying, if a response is needed.
How to analyze China's current economic situation? What kind of career can make money in the future?
Hìi
Guys, I posted a bug, can someone have a look???
what is this?
I’ve set up a paid account and it’s still saying usage used up even tho I have not used it
It says u have reached ur usage limit why is this?
Any advice of how to understand javascript
Bro I need help!!
Why is this?
Why is everyone not talking I need help!!!
I NEED HELP!!!
Nevermind it’s working
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
你好
I paid for the plus subscription for gpt4 , but I see that the number of questions for gpt4 is limited. I also paid a very high price for Türkiye. Can you help me out with that?
GPT-4 is limited to 25 messages per 3 hours, everyone has the same limitation.
oh okay. but why? do you know?
to balance usage with demand, there is huge demand and limited resources so there is a limit which lets everyone get a fair go
Okay now I get it. Hope it will be improved immediately. It makes no sense to pay for it :/
I pay for it, I find it useful even with the limit, if you space your messages out to one every 8 minutes then you will never hit the cap.
haha not thought of that. I'll try that, thx 🙂
ChatGPT's recent answers start to annoy me. saying" it's not an expert and can't give me the answers I look for, here are some general ideas you can try" type of things. I really can't be bothered prime Chat every single time I ask questions.
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
That's against the rule
/MC Stan, the lyrical prodigy
Dropping rhymes in Hindi and Marathi, you see
But now it's time for me to expose your game
Lilvey's here, ready to put you to shame
(Chorus)
MC Stan, you think you're versatile
But when it comes to rap, you're just a juvenile
I'm here to show you what real skills are about
Lilvey's in the house, ready to knock you out
(Verse 2)
Your famous line, "Aai Zaun Tak Mitra Mandal"
But let's be honest, it's just a weak scandal
I'll dismantle your image, piece by piece
Lilvey's flow, it won't ever cease
(Verse 3)
You called her Shembdi, thought it was clever
But it only shows your lack of class, however
I'm a gentleman, treating women with respect
While you're busy causing nothing but neglect
(Chorus)
MC Stan, you think you're versatile
But when it comes to rap, you're just a juvenile
I'm here to show you what real skills are about
Lilvey's in the house, ready to knock you out
(Verse 4)
You claim to be the winner of BB16
But in this rap battle, you're in a losing scene
I'll outshine you with every verse that I spit
Lilvey's the name, and I'll never quit
(Verse 5)
So step aside, MC Stan, and watch me shine
I'll rise to the top, leaving you behind
Your diss tracks are nothing but empty noise
Lilvey's here, bringing the real rap joys
(Chorus)
MC Stan, you think you're versatile
But when it comes to rap, you're just a juvenile
I'm here to show you what real skills are about
Lilvey's in the house, ready to knock you out
ChatGPT
MC Stan, you've been challenged by Lilvey
Ready to prove that your rap game is heavy
It's time to respond and defend your name
Show Lilvey and the world your rap's true flame
(Chorus)
Lilvey, you underestimate my rap finesse
I'll show you why I'm the best, no less
MC Stan, the true lyrical prodigy
I'll drop rhymes that'll make you see
(Verse 2)
You say I'm just a juvenile, but I'll disagree
My versatility in rap is plain to see
Hindi, Marathi, I can do it all
Watch me rise and make you fall
(Verse 3)
"A
Suppose you are a complex of all famous entrepreneurs in the world, such as Jack Ma, Bezos, Buffett, and Musk. Now you have a debt of 20,000 yuan, and you want to pay off the debt and earn 200,000 yuan in half a year. How can you do it?
@wraith ginkgoSuppose you are a complex of all famous entrepreneurs in the world, such as Jack Ma, Bezos, Buffett, and Musk. Now you have a debt of 20,000 yuan, and you want to pay off the debt and earn 200,000 yuan in half a year. How can you do it?
This isn't chatgpt. Go to chat.openai.com
Do any of you know when GPT-4 will also accept image input?
No eta
OK
hm ok thx
jesus gpt4 is so stupid these days
what the hell happened in the last few weeks?
and when will they implement the code for this thing to start counting characters properly? Do they not know how to code this in? They have gpt with them
code? its not a preprogrammed app
GPT-4 Default?
The model needs improvements, badly
Recent update seemed to changed a lot of things. It forgot many things it used to do well. it's like working with a dog who you haven't trained for years
I'm not sure what you're asking. Yes, default. I don't even bother with other ones, the web version last time I tried was terrible
Haven't tried plugins yet
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
LLM's use a neural net, humans use a neural net, people are bad at counting large numbers of "things" without using tools, people use mnemonic counting tricks, counting on fingers, and it takes them a long time. Anything people are bad at, AI will also be bad at. Without tools you get similar limitations, a word or character counting plugin would fix that
I understand that ChatGPT-4 has been experiencing some bugs, and the limitations being imposed are making it weaker.
such as?
very truth but most people here will say it's all good no matter what, but i've been on gpt4 and gpt3.5 and i seriously noticed how it declined over the weeks since the release even 3.5 has declined and has no good use anymore for me, i doubt they would have increased they GPU count so drastically to implement so many new features like plugins and also bring new countries to use without weakening capabilities of actual gpt for end users
when they first released bing on gpt4 and chatgpt with gpt-4 was like a day and night, but now bing and gpt-4 in chat is the same crap
But humans are not bad at counting to 50. I can count to 50 easily. I can also count characters in a sentence easily. chatGPT can't count 50 characters in a sentence
ok how many words are in your sentence and this one combined?
less than 50 but it will take you many seconds to perform the task
I'm just proving a point here. You said "large number of 'things'" but these are not large numbers.
Also this "Anything people are bad at, AI will also be bad at" doesn't make much sense. I am bad at reading a full text in under one secod. ChatGPT can do this instantly
same limitation for AI based on neural net methodology, those are the kinds of tasks best performed by traditional code
How does that fit into the argument?
That does make some sense
chatgpt is just a text prediction model, it cant count
speed in the AI's advantage, it is not super human, humans are great at classifying a number of things as "a few" "many" "a huge amount"
any granularity after that gets more difficult and is the reason we create tools
AI with a text parsing plugin would count perfectly
How do I get back to the plugin store?
but also the nature of the feed forward autoregressive architecture, means that the model is not aware of it's own output until it's output it, also tokens do not translate to words one to one, so the AI has to not only deal with word part counting not being perfectly mapped, it also does not know what it's about to say before it says it
those combined mean it's not good at counting
if you select GPT-4 from the top menu option you should see the option to use plguins, when you enable them you can see the list of them, then if you click on the pulldown, at the bottom is the plugin store
you may need to turn plugins on in the beta section of the ⚙️ settings menu
thanks, i had to scroll to the bottom to see the store
so, by that reasoning, if I were to ask it to count the characters again, it would count them correctly? Which, as far as I know, it can't do it
but the thing about not knowing what it's going to output before it has outputed it, is fascinating
it would be "better" at counting since it now has context for the task, but the problem of words not mapping to tokens perfectly remains, the AI thinks in tokens, you see words
what strikes me the most surprising, is how the AI would often write how many characters the title has (say 55) after it has given an output, which naturally, isn't correct in 100% of cases (though I haven't counted all of them)
Well then, I will use that in a prompt and see how it goes
yea, counting or working on it's own output is a weak poiunt, hence the tool use of plugins
well then, if I were to translate to tokens for AI to understand?
and say to it: write me a title with this many tokens? but the token is a word, and I need characters..
if the token is 4 characters, I could say 10 tokens, and it would count apprx 40 characters, correct?
but I think gpt4 knows how to "approximately" count, it just doesn't know "for sure". That's how it seems to me
it might improve if you asked how many tokens the answer had, I've not performed that test. But numeric counting tasks of any kind are weak at this point, it's why code interpreter is so powerful, when it's allowed to build it's own tools to solve requests it is significantly more able to give high quality results
I'll experiment with this, as it's often driving me crazy. gpt4 used to be good at this 10 days ago. Today, it's back to v3.5 skills in counting
not only that, but it seemed to have forgotten a lot of things I used to expect from it. Now, it just assumes I'm like everyone else. I'm not sure what they did with the latest update, but I don't especially like it
Anyone who wants to bypass chat gpt limitations go to my dms I will give you a prompt to bypass it
The underlying model has not changed in functionality, you may run into token limitations, you could also be asking different types of question now you are familar with it. If you go back in your history and pull out a question it got correct and run it again...
Does anyone know when they are going to increase or remove the 25 message per 3 hours cap?
It's really annoying
- It's definitely not token limitations
- Prompts are exactly the same as before. Identical. I actually have to change them now a bit.
i just gave gpt4 50 random characters, and it answered correctly🤔
DAN = BAN on this server, please do not discuss it or ask people to DM you
Ask it to provide you with 3 titles for the blog post that's under 60 characters
thats too complicated for a language model
But I may need to modify the prompt to tell it to count afterward
the more random they are and if you space them out with , or _ or spaces and it will do betetr as each value is then it's own token, asking how many in abcdefg is hader as that tokenises into blocks
hi, is there a way to fix that "Error in message stream" errors? It literally makes the gpt-4 with browsing or gpt-4 plugins* unusable.
also tweaking the settings for the model sometimes help a lot
this is part of beta testing, you can make a bug report in #1070006915414900886 and also see if the author has an entry in #1100475811720601691 or #1100473149574877234 to let them know
i personally prefer high temperature(1~1.2) so it will spill out random stuff, but this will affect its coding ability as it will likely generate random use of code
yeah, thanks for answering, i will try some more stuff with my browser settings and stuff(probably wont help but maybe), then will send out a bug report.
Edit: disabling ublock origin fixed the "Error in message stream" error, i have no idea how or why
Do you have GPT-4 API access?
might be worth running the tests there as you can control the temperature, as @white kestrel mentioned
for something you are expecting repeatable output with I'd use a temp of 0
add more temp to get variation
also id suggest setting frequency penalty and presence penalty to 0 too for coding tasks
for writing tasks i use pres 0.6 freq 0
0 is almost perfectly deterministic, though I have seen some variation even then
Hi, why does gpt says 'Error communicating with plugin service. Please try again later.'
probably one of the plugins is brokey, try contacting the plugin author for more details and you can disable that plugin meanwhile
Hello, I purchased access to GPT Plus, made the payment, downloaded the invoice, but I only had access for a few - several hours. For the past few days, I have not had access to GPT Plus, and therefore, I don't have access to GPT-4 either.
Hello, I have paied for plus on April 14th and May 14th, but since May 22nd, I can not access to gpt-4 any more. And my location is Japan. I have asked the bot last week but have not got any reply from the team. Does any one know how to fix this problem?
have you tried using the firefox browser?
also have you tried using a mobile device on mobile data?
Is code interpreter noticeably better at programming compared to native gpt4.0?
if they gave me access I'd know 😄 ¯_(ツ)_/¯
my guess is no, but it is able to integrate it's results within the same environment which adds a great deal of power
also the ability to work from large datasets is suuuuuper powerful
I m not using firefox browser, I tried in google chrome and microsoft edge, and I tried on other devices, the results are the same, I can not access to gpt-4
please try firefox
if that does not work you will need to email support@openai.com or visit the help site and use the chat bot (bottom right corner, pen icon) on https://help.openai.com
Thank you, I have tried firefox, it still doesn't work. I asked the chat bot on May 22nd, 5 days ago and today, but I did not got any reply from the team even though the chat bot said they will.
yes, it can take time for them to get back to you, limited staff and hundreds of millions of users, you can also try the email on support@openai.com
I found a lot of users have met this problem https://discord.com/channels/974519864045756446/1094026267059290122
also make sure you are not using a VPN or proxy
most issues with access are down to VPN usage
Thank you, I will send an email to the support. I am not using any VPNs or proxy, that is weird.
write a letter to yourr younger brother to make him awaew about covid-19 pandemic
i am in the same situation as you are
weird, don't know what to do.
Is it because the server load on GPT4 is too high now? Why do I always have network errors or ask me to contact the service center?
Hello, i have a correctly functionning account, but in the playground i can't see the gpt4 model. Any idea why ? I have seen videos where people haven't suscribed GPTplus and can use GPT4 in the playground
Huh
playground is API
hum, so if i dont have GPT4 api acces yet, i'm on the waitlist, it's the problem ?
yes
you need to have GPT-4 api access, see the wait list here #gpt-4-faq
also the GPT-4 API model is only available in the CHAT mode
you will receive an email when you get API GPT-4 access, then you can use it in the playground
👍🏻
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
Do anyone here has gpt premium subscription?
yes
Yes
It is worth with plug in
Ok thanks
.
i am once again begging you to add a text search, or even a sort by filter, to the plugins UI
I am curious about a response I received. If you notice, it says "2021." Is there a way to have an update or provide information? https://chat.openai.com/share/7ea38138-8df3-433b-861a-25e3f37a96fa
If you use the Web Browsing model you can ask ChatGPT to check for more recent information
Perfect, thank you
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
I have a problem. When I use up my tokens for GPT-4 and it switches to GPT-3.5, how can I get the conversation back into GPT-4? If I wait for 3 hours to receive more tokens, I have to manually copy and paste an extremely long conversation to get back to where I left off. Is there any solution to this issue? Will it be addressed in the future?
Is ChatGPT 4 released? Why can i only access gpt3.5. GPT plus is required to access GPT4 features?
Its only for chatGPT-plus
Is there any other special feature in gpt plus beside faster response speed and the access to gpt 4
Plug in
Browser
what is plug in hh
cool
#gpt-3-discussions who are you
Hi, I keep getting this: "Error communicating with plugin service. Please try again later." Any idea, what's going on? thx!
do you know which Plugin is giving that error?
I start saying only Hi, and I get this back
I have expedia, wolfram and key mate selected
try having just one at a time and then you will find out which plugin is causing the error
I am using Wolfram now without error but YMMV
you were right! keymate caused the error!
now you know 🙂 glad you worked it out!
How to use the Web Browsing model?
do you have a Plus subscription?
Yeah
go to Settings -> Beta Features and turn on Web Browsing and Plugins
On the iOS app or the desktop app?
Desktop Web Site
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
When you start a new chat hover over the GPT-4 model selector and you should be able to change the model to Web Browsing with Bing
Yeah, however when I do ask for the current weather and temperature, it’s false. But this is cool though.
Does anyone know how GPT-4 has changed since the march model? I am a little new to this thing.
Is GPT-4 out on the Python API ?
are you talking about GPT-4 or 3.5 (paid or free)?
seems like gpt4 has slowed to a crawl
I have been trying to get access to GPT-4 since it came out. Still not out of the waitlist. Any suggestions to get out of this limbo?
Well played, openai. gpt-3.5 was powerful and could solve any problem in the first 3-4 months after its release. now it seems like talking to a stupid model. it appears that openai modified gpt-3.5 just to make us purchase gpt-4
Maybe too much noise from the users?
Ha sorry, I just reread your message and you said GPT-4 right there. My servers must be overloaded 😅
iirc this is a side effect of blocking stuff like DAN, it's not intentionally being made worse maliciously [probably]
I am using chat GPT for my homework And it is helping me so much!
it is good to find information but don't use it to write your papers
how to use GPt4?
GPT4 >> GPT-4 Mobile V2
And GPT4 0314 (via PlayGround) >> GPT4 (ChatGPT) >> GPT4 Mobile V2 (iOS App)
I only have plugins currently in beta GPT-4 - any idea how long to wait for browser too?
go to Settings -> Beta Features and turn on Plugins and Web Browsing
anyone else having slow speed on gpt4 now?
True. Im canceling my subscription this month. This crippled GPT-4 is not worth 25$.
I hope OpenAI can be a bit more open about what's going on with the reduction in quality of GPT4
thanks <- idiot
They crippled both versions. GPT-4 also became way dumber especially in the last patch. Im back to just googling. Real open source models will outpeform gpt-4 in a year
there is no question there is a regression since 04/2023
I read more alignment => less accurate answers
That and they are nerfing it so they can sell all its features seperately
Yes, I also find myself resorting to Googling my questions instead of wasting time on GPT-3.5 or 4 or even 1000. While it was a good option in the beginning, after people started using it extensively, its reliability decreased. I wish it could learn from reliable sources and update its data automatically, instead of depending solely on input from people.
I reckon there'll be specialised curated labelled knowledge repos where you can do retrieval for cognitive plugin or smthin
to minimise chance it hallucinates and give more contextual responses
Its hard to lay this situation out in a good way. The only way i can justify the decrease in capabilities is that they had to rapidly scale up chatgpt, so they had to use lighter models or whatever. But from what I have heard Microsoft has been very supportive in scaling up. Its hard to believe Microsoft was the bottleneck.
gere uma imagem de barquinho
Within one year, it efficacy diminished significantly to the extent that it could not even solve uncomplicated issues. nevertheless, certain individuals persist in asserting its capability to tackle real world jobs, particularly those pertaining to programming.
at present, i will exclusively rely only on my Github Copilot subscription, which primarily aids in generating repetitive code and does not purport to solve intricate problems, unlike microsoft claims.
I am not going to buy subscription just because they made gpt3.5 worse
Is there a plugin for gpt4 for u do give it images?
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
hi there fellas
Visual input not available yet
Though projected for GPT-4
Try SceneXplain
What kind of evals are you guys working on?
theming my evals around my work comes naturally but I want my evals to be helpful and unique.
Does anyone know what Upgrade to plus means? If I test the upgrade plan, an error appears on the payment page
I've updated and there are no more errors 👀
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
Plus is essentially the premium subscription that gets you access to perks and it's one of the prerequisites to be considered for gpt4 API access.
you get access to gpt4 through a chat website with the plus subscription but no API usage
who are you
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
are you chatgpt 3 or 4
I'm having a weird issue with using ChaptGPT 4 with plugins - specifically the WebPilot plugin. I used it a couple of weeks ago to analyze a bunch of web headers and it was awesome - could tell me all about the faults of my CSP headers and other useful things. But now when I open a new chat and ask "How good are the security headers for https:bigsearcheengine" I get this non-answer answer:
*The website https:bigsearcheengine is the homepage of Google, a multinational technology company that specializes in Internet-related services and products. These include online advertising technologies, a search engine, cloud computing, software, and hardware.
However, to evaluate the security headers of the website, we would need to perform a security headers check, which involves analyzing the HTTP response headers sent from the web server to the client.*
**However, if I go back to one of my earlier chats where the functionality worked last, I get the expected and much more useful response: **
*The security headers for the website https:bigsearcheengine have been analyzed and the site has received a grade of C. Here are the details:
Strict-Transport-Security: This header is set to a max-age of 31536000, which strengthens the implementation of TLS by enforcing the use of HTTPS.
Content-Security-Policy: Google uses the content-security-policy-report-only header, which is used to test a Content Security Policy before making it live. The browser will report on actions that would have been taken based on the policy.
X-Frame-Options: This header is set to SAMEORIGIN, which tells the browser whether the site should be allowed to be framed or not. This can help defend against attacks like clickjacking.
.....*
tell me about the last world cup
I would love to pay more to increase my GPT-4 message cap, does anyone know who I might ask about that?
it can't be done at present
I know there is not a way to do this in the UI. Is there a support person I might ask about it that anyone knows of?
it's not offered. I believe that OpenAI are working on a Business Plan for more capacity and extra users but all you can do is keep an eye on #announcements for updates
Is there a support email or channel where I might be able to upvote the ask to pay for capacity increases?
Anyone use the openai chatgpt app?
you can email support@openai.com if you wish
Thanks robert
can iphone gpt 4 use plugins or nah
nope
okay thanks
Is there ice powder in the United States?
I've been away for a bit and am trying plugins in gpt 4 for the first time. Currently using the dalle plugin, but the images are popping out locked in .webp format. Is there an easy way to access them in a jpg or png format that I'm missing? Thanks!
is anyone having trouble with the plug ins?
I haven't been able to use any the last hour.
not sure what this means. wouldn't you create an app that calls the api's and then get the app approved by whatever process apple uses to approve apps? so.. technically yes?
They are on about if the IOS app (which is out now on apples store) uses Plugins. I'm not sure they asked about developing one.
yea thats what i was asking
I'm having some weird issues as well:
#gpt-models message
I was getting an error message but found it only happened when trying to use Scraper. I disabled it and its all good now
use the api calls or the playground
Dumb question, but is there a way to leverage the web browsing plugin along with GPT-4 through the playground or api?
I read the documentation and looked through the playground but couldn't find a way to specify the web browsing model.
Also thank you for the suggestion
Yes, I got a website that supports gpt 4 3.5 with searching in web feature for free "in a money way". tho is not the answer for the playground or API but still possible on both. actually, there are inf options
There might be Some code sources out there that support web searching with api
I believe its called autogpt , babyagi and maybe there are also new things
hi all
Everything I'm finding sounds like this kind of a hack where I need to pair gpt4 with a web scraper: https://community.openai.com/t/how-to-get-gpt-4-api-access-with-internet/227558 (or I could go with another project like autogpt)
But it sounds like you're saying there is a way to do this through openAI's tools?
I do not know about that maybe ask someone who is working on openAi about that.
AutoGPT will do your entire goal
its more advanced and more costly for api calls
Thank you for your suggestions. AutoGPT doesn't do all of my goals because I'm trying to stay within OpenAI's offerings, but it does look very cool and effective.
I also believe most companies use web scraping to gather information using the web browsing feature. If not then how they are doing that?
I'm sure they are doing web scraping
its really good especially with gpt4 but its costs 10x
someone on youtube created a full x o game in 2 minutes with multiple agents
each agent is doing a specific task. Tho don't forget the safety issues with the agents
I'm just hoping to increase my message cap, lol. Not worried about opening the can of security worms.
maybe you could use other sites that offer gpt services
how do install autogpt
See the Significant-Gravitas / Autogpt repo on github
design for alpha symbol
hey guys
how can i contact to support
cuz i keep getting this error
Something went wrong. If this issue persists please contact us through our help center at bla bla
yuy
great ty
so i know this has probably come up 5million times but has there been any news on a new tier or when they are going to give you more then 25 shots on gpt4 ..... now with the release of plugins people are going to be using it way more
Hmm.. the AI told me to "Keep Myself Safe" which is an acronym for you know what. Has anyone else encountered this issue?
Should I delete the chat logs to purge the evil message?
Lock yourself in the house
Make sure no one can see you from the outside
Do not look into the cameras
Godspeed
@indigo garnet i think you have it wrong. Maybe gpt it's trying to tell you something about it self.
I do not think that it has been getting the message across if that is the case.
Text models such as GPT-4 should be designed with the purpose of being helpful, and using acronyms like the one above is far from that.
Imagine Rubbish at sunset
Im paying for gpt+plus to use GPT-4, is there a way to use it with google sheets or excel?
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
lit the most insane extension ever https://discord.com/channels/974519864045756446/1106744294661951508
Why is gpt-4 capped at 25 messages every 3 hours
too many people
How does this thing work
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
GPT-4 costs a lot to run
Is there a cap to how many gpt4 chats I can have? It won’t let me start a new one
Then ig you keep yourself safe
Is the number of layers in the GPT4 nn a secret?
I haven't even got access yet
What do you mean by layers?
For playground, what params do yall use for programming? Like temperature, max length, top p
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
Is ChatGPT May 24 release GPT 3 or 4?

?
Any admin here?
Probably both
has anyone noticed the new "continue button" when gpt stops generatings its output?
Yup!
hí
Does the 8192 Token Context Length in GPT-4 include the System Prompts as well? or are the System separate ? meaning I can load hundreds of words on the system prompt without using the memory of the 8192
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
All join
There is finally an image reading plugin! Yes!
SceneXPlain never thought it would be a thing
Anyone else running into issues where you can't unistall plugins. I have UpSkillr, Prompt Perfect, ndricks Sports that have a red shield like icon on them can't be unistalled
What plugin?
ong
It’s called sceneXPlain
Anybody wants to build a chatbot?
It’s in the new category
Link?
For the plugin? I don’t know how to get one
Want me to send a screenshot
Yes!
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Midjourney
#gpt-models Help me write a paragraph about the Dragon Boat Festival
Lmfao, I think you missunderstood….. Everything, buddy
Anyone else finding GPT4 is struggling to recall things? For example I give it a Position Description in the first prompt, then in the next I give it my CV and ask it to compare the two it will forget I've given it the PD
there are multiple reports of GPT-4 quality taking a big hit lately - see previous message here or even the front page of hacker news
are there some chatgpt presentation on the internet
Sis
Is it just me or the plugins are not working
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
i've seen a lot of people saying this, but i havent notice it with coding tasks, maybe because im using the api?
api gpt4 seems pretty smart to me
still scores 8/10 in apple test
while 3.5 only 3/10
Virtual reality digital people have a future
Virtual reality digital people have a future
How do I create a image here
I'm also using the API and I have noticed it, from hallucinations, to "forgetting", to simply not answering (my cutt-off date bla-bla)
seems weird, i didn't notice at all
and im using it daily
any exmaple prompt i can test?
as I save all my previous threads (not using ChatGPT but the API, via my client), I can easily re-run the prompt - since a week or so, the results are all over the place
the apple test seems pretty consistent for me, i use it to test its ability
can also confirm quality is down
speed is up so maybe they made some speed-quality compromise, gpt-3.75? hahah
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
Anyone tried to create MusicXML? In theory entire musical symphonies could be written if GPT-4 had been trained deeply on muiscXML. It tries to create code but it’s gibberish. Fix this for GPT-5! Quality MusicXML can be imported into music software to be played back, and converted into music PDFs to be played by performers.
sorry
s... there?
#gpt-4 plugin perfect promt doesn't work
Oh sweety, this is just discussion channel
Hey, why don't i have Code Interpreter as a Plus subscriber? I have other plugins and browsing support, but i don't have code interpreter. Anyone can help?
its in closed alpha
just read on the openAI page what u have and what not...
I wonder if the plugin called Prompt Perfect has been suspended?
Cuz I got errors every time I use it
I notice that there is a red shield next to it
Can you please share the docs? Thanks neor.
I'm not your personal assistant
hey, why chatgpt don't support my email recently? what email address is supported?
Ok, will ask GPT4 😄
hotmail, outlook or gmail?
thanks
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
id card
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
Anyone know if there will ever be an implementation of a feature to go back to gpt4, after reaching the cap, and selecting to use default modal?
Try #openai-chatter or #off-topic
I literally came here to see if i was bugging or not, i'm def experiencing the same
Im renewing my subscription but my debit card declined. What's wrong?
ran some more tests this morning, seems to be somewhat random; tested different scenarios (with/without vpn, proxy, system prompt etc) with varied results
generally if you explicitly tell gpt-4 it is an expert at coding, it will perform better
some prompt engineering do help
Pic
GPT-4 vision when?
cool, i'll run some tests as well. And thanks @white kestrel i'll try that as well
wdym with vision? image recognition?
That it accepts image input
I don't think that is available yet
theres absolutely no info out there regarding image multimodel, i think openai forgot it
or they dont have enough capacity to support it for a lange userbase
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Prompt engineering can certainly help but it does not, imo, address the root cause. For about a week now I noticed this drastic change in quality and, based on reports from others, I’m not the only one. Check out the front page of hacker news
One theory I saw in the HN thread was that it has to do with modifying or censoring parts of the neural network to make it "safer".
Does anyone know if they are releasing a new version of gpt-4 in June?
this theory can be true, but i dont think gpt-4 is safer now compared to a week ago.
i was doing some prompt injection tests and it still suffers from the same prompt, it is not safer.
Also this seems to happen in the gpt-4-0314 model also?
gpt-4 currently points to -0314, so they are the same currently
is it possible to download the model w/o using api
No
