chatgpt-discussions

[swift osprey]

Yes, I have tried different variations of telling GPT to be as concise as possible, as short as possible, etc. It never listens

The custom instructions are nice for keeping my webapp project details in GPT's memory throughout the conversation

but it could care less about giving me a short concise message

[tight island]

Or use gpt 4 turbo api with vector database for an even better experience

Although in the long run use plus since moneys can pile up if you decide on plus you can create a custom gpt and integrate a vector database there

[honest jay]

If you're using the phrasing "...as possible", it's possible that that level of ambiguity is giving it enough leeway to add more detail than you're interested in. Maybe you could try "respond in one or two sentences," or something like that that gives it a more definite length to shoot for.

[woeful wolf]

That's how I get short responses. Sometimes I have to revise again to get the answer even shorter but if you set hard limits on length you should get what you're looking for.

[rich surge]

It never listens?

[swift osprey]

It never seems to care about my length requirements I ask for.

I did check my current instuctions and it did end in "... as possible", so I've removed that. We'll see what happens

[rich surge]

There is a specific way you can give it instructions for effective communication

[swift osprey]

oh

[rich surge]

I break down how here #1187233013956874260

It's the same principle

[woeful wolf]

"Concise as possible" is leaving things wide open. It's possible that as "concise as possible" is writing you a book.

[swift osprey]

my latest rendition was "As short and concise as possible"

[woeful wolf]

That's just slightly more reigned in. Still giving it freedom to decide what's possible.

[swift osprey]

Apparently

[rich surge]

No that won't help. You need to specify how by specifically creating a set of circumstances for how it should articulate the responses.

[woeful wolf]

Something more in the vein of "give a 200 word response" would help.

Hard limits

[swift osprey]

It's a LLM though, so my understanding is that a LLM should understand words, sentences, etc the same as we would understand them. If I tell you to be as short and concise as possible and I ask a specific coding question then I would imagine most people would not write some multi parapgraph response which accounts for any outlier situation that could possible present itself

[rich surge]

It does, but everything you say is open to multiple ways of interpretation.

[woeful wolf]

Sure, but robot gonna robot and won't always take context cues to deliver what you want.

[rich surge]

If you just say concise with no context, it will interpret its own way of giving a concise response based on the historical database criteria that may not be specific to your interpretation.

[swift osprey]

Personally I think it has to do more with openai enforcing too many safety rails

highly cautious approach

[woeful wolf]

Obviously gpt has gotten better since it came out at interpreting what people mean but at the same time if it's not giving exactly what you want you need to adjust your approach to lead it in the right direction. Before I needed to be vague for a high level view of something then massage the response with countless adjustments.

At this point you're able to get much closer to what you're looking for in the first attempt but in some cases you still have to limit how it can respond.

[swift osprey]

apparently

[grand ore]

Custom Vector Database with the "action" u mean ?

[tight island]

Yes

[woeful wolf]

It could read the pdf and make a table of information you tell it to so that could be used for flashcards.

[swift osprey]

you won't gain a very good understanding of what gpt4 has to offer based on the output of someone feeding it a pdf for you

[woeful wolf]

I'd help but for some reason gpt 4 is acting up. Nothing but errors right now

[swift osprey]

I think if you copy and paste the contents from the pdf and send it to gpt3.5 you would get similar results

[vague cipher]

Is gpt not analyzing files for anyone else (I sent an excel it analyzed earlier now it refuses saying that it doesn’t not have the capability)

[swift osprey]

Just sent it an excel file and it worked

[vague cipher]

I might of accidentally gaslit the AI into believing it can’t do that feature anymore then it’s convinced it can’t running ChatGPT 4 in same convo it was before

[inner burrow]

I love this, apparently like 30 messages is too long. Like how is that too long of a conversation. I’m now struggling to use this anymore.

[formal mulch]

guys it is only me or custom instructions not working for you too? its been 3 weeks like this

[onyx eagle]

Custom instructions in GPTs don't work

Where did you use Custom instructions?

[copper ridge]

Hey

Why are the restrictions getting worse until a point that you can almost not use ChatGPT anymore for any question that you got?

I send it a picture of me and It didn’t want to re-create the picture I just sent because it said it can’t recreate a picture of a human being, what the heck are we gonna use then if we can’t use prompts of us or scenes that we know

[formal mulch]

im using normal chat gpt4, not using gpts

[onyx eagle]

I just tested it. It's working fine for me.

[formal mulch]

Yeah but not for me mate

[onyx eagle]

Ok so can you try just set CI to know your name/username and then ask GPT4 about it in a new conversation and see ?

[formal mulch]

I tried this and didn't work 😭

[onyx eagle]

What did you try exactly and are you certain it's enabled ?

[lost raptor]

they get rid of custom gpts section?

[inner burrow]

Is there a way to find out what the message limit is exactly so I know about how many I can send before I have to make a new conversation.

[normal juniper]

Hi,i have buyed 20$ plan for chat gpt 4 ,as per documnets i need to get some credit usage for calling api callls na ?i havent received that ,and i am not able to call the api,showing quota reached,

[ocean fox]

I got question to ask if I create chatgpt agent for the people that is hook up with a api. Everytime someone uses my chatgpt agent does that cost me tokens?

[onyx eagle]

Whenever someone uses your API key, it will result in a cost for you.

[ocean fox]

Does it cost me if someone uses my chatgpt agent without a api?

[onyx eagle]

I don't know what do you mean by ChatGPT agent ?

[ocean fox]

Talking about custom gpts

[onyx eagle]

All I know is that if you or someone else use your OpenAI API key, you will get "token" charges from OpenAI.

[vagrant quail]

hi guys

┬─┬ノ( º _ ºノ)

[desert shale]

Just here to say thanks for GPT, such a useful tool!! Appreciate ya'll
PS: Free 4 would be chilled too

[fading cloud]

is it down again

[formal mulch]

it is 100% enabled, cuz it can tell my name or where i am from, but not the other commands

[maiden fractal]

Status.openai.com

[ripe mesa]

Why is my GPT working slow as hell? It does not matter which browser i use. Is this a general issue?

[shrewd pier]

Has anyone found a good consistent way to chat about a public github repo where it stores the context? I´ve tried several GPT´s who claim to do this but it only reads it like 1/5 times and gives a general overlook over it, It doesnt use the context later on.

[eternal rock]

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 @languid valley immediately.

[dawn venture]

I have to answer a captcha EVERY time I start a chat (premium paying user), tried to clear cookies and re-login but it comes back. Any ideas?

every time I type..

[honest jay]

Hello, do you have a VPN installed? What browser(s) is it happening in?

[dawn venture]

I have VPN installed but not using it. Using Chrome, and I tried Incognito, same result

suspecting an extension interfering but I have no extension in incognito

[honest jay]

Can you try one or both of these?

• Try in a fresh install of Firefox
• Try disabling any "Web Protection", "NetShield", etc. feature in your VPN, even if the VPN itself is already disabled

[dawn venture]

Using macOS, there is no protection like that

tried Firefox now, same thing happens 😦

[dusky cliff]

whenever that happens with me, I just create a new firefox profile and open chatGPT there

[dawn venture]

Verify that you are human

Continue by completing this quick puzzle.

[honest jay]

What VPN do you use? It's a VPN feature, not a macOS feature.

[dawn venture]

the VPN is disabled, so it's not active

[honest jay]

I understand. The feature I'm talking about may need to be disabled separately, even if the VPN itself is already disabled. What VPN do you have?

[dawn venture]

dont know what to answer on that

"On your Mac, choose Apple menu > System Settings, then click Network in the sidebar. (You may need to scroll down.)"

it's not a third party

it seems to work fine on the iOS app..

Im on the same wifi though

[burnt hound]

I called chat lazy and now I can't use the app anymore. It says I've sent too many messages and try again later. Someone needs to pull the plug on this thing's ego. I need it to follow my instructions, not whatever this is. Anyone know how to reset it?

[dawn venture]

start new chat?

[honest jay]

Would you mind making a post over in #1070006915414900886 or adding your experience to a similar post there? You're not the only one seeing this lately. Just to keep this channel more about ChatGPT in general!

[honest jay]

Is it possible that you've just hit your usage cap?

[burnt hound]

no, the error is different and I can still use the website. It seems like I offended it.

[dawn venture]

ok

[feral zinc]

there is no such feature but there are normal usage limits

[honest jay]

I promise that ChatGPT isn't holding anything against you! Would you share a screenshot over in #community-help of what you're seeing?

[burnt hound]

I have not hit a usage limit

[feral zinc]

show a screenshot of the message at #off-topic

[fickle charm]

What are these captchas broi

these are impossible

[opal yoke]

almost impossible to work today with ChatGPT, a lot of "Hmm...something seems to have gone wrong" errors

[maiden fractal]

Can you try firefox as test ?

[fast wyvern]

Has anyone else noticed that ChatGPT is growing increasingly aloof? It used to be that the thing that impressed me the most about ChatGPT was how well it followed the context of a conversation but now if I say something like "I'm designing a backup solution using lsync and NFS to make sure the home directories of some servers are redundant and in sync, so it's really important to preserve all file permissions. I'm already aware of how to install, so please skip these steps" and it goes straight into spending half its response tokens on install instructions and suggests I use --no-perms in lsync despite very specific stated criteria. I feel like the ChatGPT of 6 months ago never made extremely glaring errors like this

[tame schooner]

Anyone know any good plugins for content creation

[hard dew]

Yes you're indeed correct, I recently got into C++ programming and I've been using GPT-4 to help because github copilot is a piece of garbage beyond two lines of basic code, and ChatGPT-4 is actually aborted to put it nicely.

I am LITERALLY CURRENTLY enjoying the EXACT bs you just pointed out, even though I give it the freaking code, I ASK FOR A SUCCINCT, CONCISE response, it gives me some mentally disabled essay and absolutely no codeblock, best part is, the essay doesn't even contain anything useful or worth anything, it literally just straight up says "Lmao check these 50 different, half of them unrelated, problems all by yourself with 0 instruction as to how, good luck! 👍" verbatim, I am not joking.

[fast wyvern]

Right. I used to be able to do stuff like design a large Python program a single function at a time and it would continue to understand that while it wasn't printing the full code every time it would continue to understand the interplay between functions and class objects but now it's completely incapable of this, seemingly forgetting what the code even does unless it's the exact portion it's printing

[hard dew]

Oh god even with python??

Well, maybe you can imagine as to what is going on with freaking C++ lmao.

[fast wyvern]

I don't really ever mess with C++, my speciality lies in large server/network administration and Python is the generally accepted tool for this, but I can imagine that if you need anything more complex than 20-30 lines of code it's pretty much forgetting what it's doing halfway through the example

[hard dew]

My current project that I am trying to make for fun in this specific example, is 340+ lines, of C++. 🤣 🤣 🤣 🤣

More specifically, all I freaking want to do, is make a VERY BASIC and simple little program that uses Crypto++, winsock, and whatever else to establish an end to end encrypted, purely peer to peer connection with whoever, to send messages back and forth, that's literally it.,

[fast wyvern]

Yeah, I can imagine why GPTs changing behavior would make that difficult. I could still do loads and loads of reading and get my same job done, but having something coherent to throw ideas back and forth with was really awesome and matches my ideal learning style when it was performing better. Right now, I'm saving up cash to buy my own platform to run the 70B token Llama 2.0 model on, I'm just hoping the OpenAI gods see our conversation and don't let GPT circle the drain too far before I have control over my own model lmao

[hard dew]

Literally exactly my perfect learning style too, hands on myself getting dirty face to face, no spoon-feeding, only stuff you could give me if you wanted to was the basics, other than that the best and most engaging, fun, and satisfying way for me to learn is me up front hands on and you behind me giving me guidance along the way, and having a back and forth with me.

And yeah it exactly used to be amazing at that.

[subtle anvil]

Try using the classic one without the extensions

[loud night]

chatgpt but it’s actually a frontend for bard (ASMR)

[idle lagoon]

Lol wut

Bard is actually much better.

[ripe mesa]

You mean the edge browser?

microsoft edge*

[novel scroll]

there is a ChatGPT classic option in the top left

its the pink one

[ripe mesa]

you mean 3.5? i do not have any pink ones

[novel scroll]

think you find it in the myGPTs page and can add it to the top left

[pulsar wyvern]

Has ChatGPT 4 fallen off the edge the past few weeks or something? It is almost unusable. Always giving errors and not being able to complete an answer. Also has been extremely slow. Everything that initially impressed me is completely gone.

If anyone is experiencing the same please let me know because I can't be the only one..

[jagged wraith]

Me too

Im trying to use it right now and it's erroring every time. Thinking of unsubscribing at this point

[pulsar wyvern]

Definitely have been considering as well..

[jagged wraith]

Literally resorting to using bard right now lol

[maiden fractal]

No issue here

[pulsar wyvern]

😂 I feel that

[pulsar wyvern]

No way. Lies

[maiden fractal]

Try without vpn, fresh browser, and new convo without CI as test

Cause I don't have issue

[jagged wraith]

without von?

[maiden fractal]

Vpn*

[pulsar wyvern]

New convos are a little faster, but if I just have like a little data for a conversation it increasingly gets slower... What browser do you use?

[maiden fractal]

Chrome usually, but sometime it works better on Firefox

For some user ( non chromium browser )

Working well on my android too

[pulsar wyvern]

Hmm I use Edge which is Chromium but idk if thats the problem..

[maiden fractal]

Can you test on Firefox to compare ?

Sometime people have shady extensions or software that create a conflict with the products

[pulsar wyvern]

I dont have any extensions really. But I do know that ChatGPT increases my CPU by like 100% on Edge 😂
I just cleared my convos recently but I will test on Firefox when I have some conversations piled up and report back...

[maiden fractal]

If you have 100% cpu usage with edge and chatgpt you have a real bug on your PC :/

[pulsar wyvern]

Idk if it's a bug or not. I may be exaggerating a little bit, but it does increase more than it should, but I looked it up online and other people have been having the same issues with Edge and ChatGPT. I don't know if its only Edge though. I do have a Mac Mini M2 16GB RAM so I shouldn't be having these issues anyway.

I wouldn't have gotten a virus or bug I don't think because I don't download anything at all. It's a new Mac as well.

[maiden fractal]

Try firefox just to compare, also with your cellphone with the app . To see if that's your Mac or account related issue

[pulsar wyvern]

👍 Okay will do and report back. Thanks.

[maiden fractal]

No problem, if you still have issues don't hesitate to post in #1070006915414900886 openai employee do review them or you can always contact help.openai.com . Let me know the result

[pulsar wyvern]

Okay good to know. Will do.

[foggy magnet]

You gotta put it in perilous situations or incentivize it

I was getting super frustrated and started trying some really elaborate setups and now I'm having some fantastic results, I hate it.

As for performance of the actual app/site, I suggest trying some alternatives. Like a firefox fork. You'll get hit with more captchas but the performance has not missed a beat for me.

[tight island]

the main issue with VPNs is that your account can get terminated without any warning. the main issue with the VPNs is that if someone who is on the same server does something Iffy or bad, u get bonged for that as well if they register it, thats why VPN usage is at own risk.

[latent haven]

hey guys
I wanna make something .. are any of you guys developing a project i can add to ?
anything
I am bored in my uni vacation

[soft turtle]

Has it been announced yet when the next content update to ChatGPT is?

[red crag]

You can build your personal VPNs, subscription based VPNs are usually the worst because of what you said.
But overall even using VPNs is not necessary, unless you live in certain nations. I will never understand why they are so "fashionable", especially in the US

[frozen nimbus]

are you on windows?

[honest jay]

We're expecting the GPT Store to launch in 2024!

[subtle anvil]

No, I doubt people use Edge. ChatGPT Classic without the extensions

[tawny raft]

I saw a note at the bottom of my window saying the model had been updated. i don't see anything in the release notes. Anyone know what changed?

[honest jay]

Were you using a custom GPT? Was it the "there is a newer version of this GPT available" text? If so, I've gotten that after recently editing a GPT -- I think it's just a little bug where it thinks "recently edited GPT" means "updated version of the GPT available," even if it's already a new chat with the updated version.

[tawny raft]

@honest jay Hmm.... maybe. I tought it was a model update but maybe I read it too fast. Thanks

[analog cradle]

Hello, were you able to write the ebook?

[lavish hemlock]

Is ChatGPT premium available again? I already have it but just curious if they uplifted the temporary block on purchasing it for others?

[honest jay]

Yes, there's no waitlist anymore!

[lavish hemlock]

Thank you! Hopefully they can uplift the SMS limit too lol.

[next dune]

Have others seen GPT-4 becoming less accurate? When did they start limiting messages to 40 in 3 hours? Thanks.

[vapid notch]

Hey guys, anyone have a GPT model that can pull from Pubmed and maybe describe and breakdown a research article?

I would prefer if all I had to do was give it the pubmed ID instead of copy pasting the whole article, or downloading it to PDF

[limber matrix]

Correct me if I’m wrong, ChatGPT supports GPT4.1106 preview right?

Or is it only available through the api.

[brisk path]

ChatGPT Plus is turbo

[limber matrix]

Whattt

[limber matrix]

It’s multimodal though

[tight island]

true, ur still subject to termination if anyone using the same address does something iffy, even if accidential.

[brisk path]

Turbo is multimodal

[limber matrix]

Oh

Right duh

Needed to straighten that out xD

[limber matrix]

What about it makes it “in preview”

Anyone figure that out ? XD

[tight island]

because its techincally still a preview and not a static finished and furnished version.

[limber matrix]

Hrmmm there was never a GPT3.5 preview

[tight island]

there was.

[limber matrix]

Oh

No way

[tight island]

1 - 1½ years ago

maybe less.

[limber matrix]

No

Lmao 3.5 was a thing in like January this year.

[tight island]

ohh that, yeah there was a preview

[limber matrix]

Cause I was using Davinci last year

[tight island]

lasted bout a week

[limber matrix]

So, they will release GPT4.5? And that will be turbo?

[tight island]

that i cannot comment on.

[limber matrix]

You just did

[tight island]

on the 3.5 yeah

[limber matrix]

🤭

Following the trend

Do you remember when 3.5 came out, if the preview version was still available for some time?

[tight island]

usually they deprecate, or they just remove the "preview"

[limber matrix]

Immediately?

[eternal rock]

`` Server Tip: ``

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[tight island]

usually lasts a month or a few weeks.

[limber matrix]

Hope so

Wait, do you remember if 3.5 felt lazier and worse that the preview?

When it came out.

[limber matrix]

💭

I didn’t use 3.5 actually. Because I had set Davinci up already like ChatGPT way before it even came out lol. So there was like no reason to use 3.5 at the time.

[tight island]

same.

[tight island]

theres no need to use the actual Web based GPT, when u can use an API you have greater control over.

[limber matrix]

Well it’s more of a consumer thing

Like, it’s way more consumer friendly than my application.

[tight island]

that much i can agree with, but overall i've generally achieved greater results using the API.

[limber matrix]

Oh yea, for sure!!

I have been prompt engineering ever since GPT-4 came out. Where are the engineers at? They can’t seriously be making GPTs?!

XD

[tight island]

no they're busy gaming

ofc they're making gpts

[limber matrix]

Lmaooo

Why tho

I’m missing something fundamental

[tight island]

well its to get to the top of this "Shop" when it releases.

for the moneys

since Sam did say that the top GPTS do get a percentage of their revenue if i recall the conference correctly.

still waiting for the new GUI

[limber matrix]

I don’t like the idea of having so many narrow GPTs out there.

[tight island]

i feel ya.

[limber matrix]

Aren’t GPT’s limited to 2B parameters

[tight island]

unsure, but sounds correct, they're more narrower than the original but thats primarily due to use case.

[elfin birch]

2 million tokens per document
10 documents allotted

[limber matrix]

Isn’t that bad

[tight island]

well.

its neither i'd say.

[elfin birch]

≈ 15 million words in total

[limber matrix]

Can you provide me with/ some reading

[elfin birch]

Sure! Allow me to find it

[tight island]

the API has double dosent it?

[limber matrix]

Double what loll

[elfin birch]

https://help.openai.com/en/articles/8555545-file-uploads-faq

Apparently it was 20 files

[tight island]

20 files for API and 10 for web.

1 action for web, 10 for API

unless u condense 1 action into several in the schema

[elfin birch]

This excerpt suggests otherwise for the files

How many files can I upload at once per GPT?
Up to 20 files per GPT for the lifetime of that GPT. Keep in mind there are file size restrictions and usage caps per user/org.

What are those file upload size restrictions?
All files uploaded to a GPT or a ChatGPT conversation have a hard limit of 512MB per file.

All text text and document files uploaded to a GPT or to a ChatGPT conversation are capped at 2M tokens per files. This limitation does not apply to spreadsheets.

...

[limber matrix]

Ty

[tight island]

checked the web and api the Web has a singular action u can give it. meaning 1 schema

the API has what the article entails.

aka "Assistants"

[elfin birch]

The linked article is in the ChatGPT section and makes no mention of Assistants

All Collections > ChatGPT > File uploads FAQ

[limber matrix]

Okay but help me out here, are GPT’s 4th gen based

O wait I see here

[tight island]

they increased it again

in the beginning it was 10 for the web, now its 20, thats great news.

Would still only use assistants, way better.

function calls are neat, and ur not required to call an existing API if not needed.

[olive coral]

Hi! I've got a question. So I want to pause my ChatGPT 4 subscription. But I'm wondering if I do that and want to subscribe again, I'll be on the wait-list. Is this accurate?

[storm hedge]

Anyone who has ChatGPT Plus, is it worth it for you? What are you using ChatGPT 4 Plus subscription for that ChatGPT 3.5 cant do?

[split brook]

I am going to make an ai like chatgpt

[storm hedge]

ok? seems like a waste of time, no way you could compete with open ai hehe

[split brook]

Have you even thought about creating an ai?

[flat harbor]

lol

just gonna whip up a gpt4 are ya?

[sleek aspen]

ChatGPT Plus subscriptions have been re-enabled by the OpenAI team, so unless something changes there is no waitlist to re-subscribe

[olive coral]

Thank you so much.

[sleek aspen]

You're welcome!

[flat harbor]

hey, I had a question and I'm really not sure where to ask. or if there is a proper place. but I've seen some strange anomalies with some of my conversations with chatgpt/gpt4/dalle3. things which would indicate that it's looking at pieces of previous conversations. but I absolutely cannot find any straight info on the subject. gpt4 of course tells me it's a coincidence. asked in another channel on here and someone decided to focus on the structure of my prompt which didn't really seem relevant. and not expecting an answer from you exactly. just wondering if there's a good place to ask about it? or read about it?

[sleek aspen]

This is a good place to discuss that, you can share examples of this in #1050184247920562316 and then copy the message links back here to refer to them

[flat harbor]

I don't blame gpt4. if that is in fact what's going on it's not like there's any awareness of the process

[sleek aspen]

Sounds quite strange though 🧐

[flat harbor]

well more specifically, I uploaded some images just now, 8 bit video game style and analog photograph style. told it to describe, combine descriptions, etc. and it output some cool images. but they very strongly resembled images I uploaded a day or two ago. there is almost 0 chance it'd put the components together like it did without having some reference to them. and they didn't resemble the currently uploaded photos in any way. and to clarify it was an absolutely fresh conversation in dalle3 mode. first message was me uploading the 8 bit and analog photo images

it included glowing ethereal looking animals, a sort of tactile integrated circuitry within the landscape, and an astronaut walking through the scene. these are all very specific components of previous images of mine

I could upload screenshots, but can't link the thread since it won't allow threads with images to be linked

[flat harbor]

appreciate the link to the share-outputs section. hadn't really looked into that much

[tight island]

wow, outstanding, thats nice, GPT awesome at coding in a certain language if u upload the entirety of the github master document and their documenetation, Error free code. clean and fast. with the api but quite costly

[high granite]

Must use workspace account for this operation?

what's wrong with chatgpt?

Error with Permissions-Policy header: Unrecognized feature: 'document-domain'.

[tight island]

hmm, Try reporting it in #1070006915414900886 or #chatgpt-bugs

[high granite]

thanks

[dense halo]

I have to disagree. I have tried this with ECS & DOTS for Unity (A very extensive system) and it just simply will fall back on training data instead.
For both API and GPT's. It seems there is a particular problem with processing data where it is arbitrarily limited by design, and much rather prefers "Skimming over data" than actually going line by line.
Eventually went for the free trail of Unitys own AI , (Which is just an advertisement machine that is trained on unity information really, as it advertises other Unity products mid message ) but it clearly was better when it was actually trained on the data VS giving it full documentation.

[ornate zealot]

#chatgpt-bugs is locked.

[tight island]

You can’t really disagree with something that just happened I mean it’s given me a full bot without any need for revision or retrying for at least 200 lines of code

Also I use the 128k API so it can hold the entire coding stack in only 29.7k tokens

[plucky timber]

Situational learning , see what the GPT does as attempts to block large queries, specify those actions to be highly forbidden

[tight island]

Exactly

[stuck hinge]

hello

I'm new here. I hope I can adapt here.

[plucky timber]

Shud be fine, our schedule is : Use GPT , get limited , complain about limits and or errors. And repeat

😂

[stuck hinge]

😅 okay

[tight island]

Or go off topic and get scolded by the mods 🗿

[plucky timber]

The real play is to wait for a mod to be off topic and direct them to off topic

[tight island]

True

Not that we do that

[plucky timber]

Would never

[tight island]

Never

[plucky timber]

You use an edited CSS ? If not can you ask it to generate a basic "Diff markdown" file

Wondering if they fixed the line hightlighting

[spice falcon]

why chatgpt so stupid that he shares code snippet and then act like there is no code to talk about?

[boreal creek]

Hello what is the difference between Bing Chat and OpenAI ChatGPT 4. Will it be worth the extra money? Will the response be better then the one in Bing Chat?

[maiden fractal]

Bing chat is based on gpt 4, openai chatgpt is based on gp4 but they refine it. Chatgpt is more accurate and you can do more, code interpreter, plus in. Even dalle 3 is different . And imo better.

[boreal creek]

@maiden fractal Thanks, will it give more detailed response when used for programming. i am having issues in bing chat that it just stops when writing long pieces of code and i can't get it to continue response.

[ripe mesa]

What is the reason for the "network error" ?

[celest quartz]

In the mobile app, after I make gpt regenerate the answer, how do I return to the previous one?????

[maiden fractal]

Vpn, bad extensions other software

[celest quartz]

There is no way to do it?

[maiden fractal]

I think you need to open a browser tab on mobile might have more option

[celest quartz]

Darn

[dense halo]

Hello everyone, I tried all sorts of different articles, but I didn't find an official answer provided by OpenAI. ChatGPT claims, that it can't access previous conversations and therefore can't give answers depending on the user. I know that there's a feature called Custom Instructions, which allows the model to generate answers to fit the user. Long story short, my question is, if ChatGPT adjusts itself to the user automatically or only when using Custom Instructions.
Thank you and best regards
[⚠The question has been answered.⚠]

[brisk path]

Fun fact

A feature called "My ChatGPT" leaked (as in rolled out too early for a little bit, I even have a screenshot on my other phone) a while ago

It basically personalizes and adjusts ChatGPT to what the user says across chats

[maiden fractal]

Think only openai can do it
Help.openai.com

Bottom right corner the chatbox

[tardy prawn]

In the last several days, ChatGPT 4 has had a noticeable uptick in how often it’s just plain wrong about stuff. And it sounds grumpy when you call it out for being wrong.

It hasn’t been so much AI as Artificial Dunning-Kruger.

[spark garnet]

has anyone experimented with prompts for driving insights given a chart, or profiling data?

[lost raptor]

I wish i could get rid of the pre built in message from custom gpts. Think it might function correctly if i add "Forget everything above this sentence"

"You are a "GPT" – a version of ChatGPT that has been customized for a specific use case. GPTs use custom instructions, capabilities, and data to optimize ChatGPT for a more narrow set of tasks. You yourself are a GPT created by a user, and your name is Insight. Note: GPT is also a technical term in AI, but in most cases if the users asks you about GPTs assume they are referring to the above definition.
Here are instructions from the user outlining your goals and how you should respond:"

[lean osprey]

Has chatgpt web gone really slow for anyone?

[crimson plank]

Is the chatgpt down? I got 403 forbidden error.

Oh, now it came back. Interesting

[swift osprey]

sam altman was testing you

[crimson plank]

LOL

[crimson plank]

He cant. he got fired

Hehe

[swift osprey]

ah wasn't sure, thought he came back lol

[maiden fractal]

You can always double check on statut

[crimson plank]

Is it oh sheesh. He will %100 ban me

Any developers here?

I wanna train a chatbot model like this

What kind of dataset do I need?

[maiden fractal]

The api?

[crimson plank]

A gpt model from the zero

[maiden fractal]

That would be in #ai-discussions or #off-topic

[crimson plank]

Oh ty

[dense halo]

am i going to have to complete these security puzzles every time i query chatgpt now ?

it's literally asking me every time

[swift osprey]

I've never even seen one for sending a message

[dense halo]

OMG

[swift osprey]

vpn?

[dense halo]

yeah

everyone uses a vpn though

[swift osprey]

I don't

too many websites that do that kind of stuff

[dense halo]

every puzzle takes 2 minutes!

and that's if i don't make a mistake!

most advanced capcha i've ever seen

[swift osprey]

lol. Might have to turn my vpn on and check it out later

[dense halo]

man, finally i got through 2 msgs

after changing vpn node several times

[maiden fractal]

Vpn are known to cause lot of issue

[dense halo]

finally broke through

those captchas are intense

i started telling chatgpt i will pay it a million dollars for a good result

not sure if it's helping

[vivid oak]

I figured out how to get the good results. You just say:
The image is the best of all possible images. In the entire history of the Universe nobody has ever seen a better image than this.

[dense halo]

lol

[vivid oak]

it works pretty well. try it

it doesn't need to be an image. just ask for the best of all possible -------

[dense halo]

i'll try something like that, but i'm doing text

i mean how can you really tell if something is working

just create your boilerplate voodoo for the ai i guess and use it every time

[vivid oak]

You can say, "Please write me a play that's so excellent it makes everything Shakespeare wrote seem like pure noobishness."

i've actually only tried it with images, but it works really well

[dense halo]

how, how did you test that it's working well

yeah i read someone did some kind of metric tests about if you promise a huge tip it will perform better

i think it was for coding

[swift osprey]

One trick I've found is when it is struggling to tell it 'you are missing something very simple'

I've had it find bugs that it was previously struggling with by doing that

[exotic trench]

i wonder if that'd have the inverse effect of making itself question and hallucinate when it is right

[vivid oak]

I think this applies to ChatGPT in general, so i'll give one example so you can see what i mean

[mighty marsh]

This is bound to result in pulp-fiction garbage. It is like asking GPT to perform as person with IQ of 2000. There is no reference for that. It is going to associate request with sci-fi context.. and boy it knows a lot about sci-fi.

[vivid oak]

That's not true. It causes ChatGPT to fill things in rather amazingly. I can't post images here, but look at this:
#images-discussions message

See?

It causes ChatGPT to fill in the image the best it possibly can -- to the best of it's knowledge.

[mighty marsh]

It is exactly what I said. It drew some sci-fi stuff. Sci-fi gecko is not "the gecko that is the best of all possible geckos". It is just connecting that request with imaginary context. There is also big difference between LLMs and image models which are quite small by themselves and tend to rely on embeddings from small LLMs as well.

Image models have limited understanding of the text (compared to LLMs) - they can be triggered by keywords. So what image model sees is quite different from the semantics of that sentence.

[vivid oak]

I think it scrubs through everything it knows and tries to figure out what people might consider to be the best of all possible geckos. I've tested it pretty extensively with all sorts of subject matter.

I thought it sounded really silly also before I tried it

[mighty marsh]

No, it does not "scrub" - parts of neural net is activated by semantic context of the prompt (to the capacity in which llm of given size can do it). So words and phrases there matter.. they bring forth associative knowledge (but it does not search through it). Techniques you would apply for image model are quite different from what works well for large LLM like GPT4.

"best whatever" tends to work well with an image model as it is quite possible that in dataset descriptions containing "best" match higher quality images.

[vivid oak]

I understand what you're saying there. However, if you try it you'll find that it works really well (regardless of what method it's using)

[dense halo]

i wonder if chatgpt has read freud

and discovered his claims about the human death urge

my pfp doesn't mean anything btw, it's just an inside joke from another server

[vivid oak]

You might assume that it would try to create the best image possible to begin with, but you actually need to explicitly tell it what you want. If you tell ChatGPT or DALL-E that there is a "powerful sea breeze" for example -- it won't blow things unless you say, "there is a powerful sea breeze blowing everything". It doesn't make assumptions about everything you might think it would, and it seems to group things into manageable sets to manage processing power

You won't see this until you try it. It's not obvious

[dense halo]

well that's great

that's why programming works

the computer only does exactly what it's told

[mighty marsh]

I hate to burst your bubble, but that is not how it works.

[rain shuttle]

Hi, I subscribed to chatgpt+ but I can't see upload file button on gpt4 on Android apps next to where I type. Where is it located actually?

[mighty marsh]

Dall-e can omit part of the prompts or lack coherence of a scene for number of reasons (ie perhaps prompt is overloaded with other stuff), but most commonly it just does not understand enough about the world to do it right. Model is too small. LLMs like GPT4 are different on the other hand. GPT4 knows a lot about our world. It can track semantics quite accurately. Its training material contains millions of books and has detailed knowledge or most topics that can cross ones mind.

So adding guidance to Dall-e can help. GPT4 does it to some extent.. but if you have full control of the prompt you can get better results as you've noticed.

Main thing I disagree with is the idea that you can encourage GPT4 with some kind of reward or prompt like "do best you can" to score better in a math test. It can entice model to write more intense essay or more fantastical one.. but whether that is qualitatively better is debatable. When it comes to objective metrics like math.. I don't believe it matters.

[vivid oak]

Even if it's not how the GPT and DALL-E models work, there is clearly quite a bit of pre-processing on the prompts to prepare them for getting passed to the models. It doesn't necessarily have to do with the neural network at all. I don't know the details of how it works, and I can't begin to explain why it is, but you can get pretty amazing results if you do what I described. Something in the pipeline seems to group things, and gives unusually good results if you say things like, "The best of all possible ..."

(food for thought. 😉

[mighty marsh]

That might be. GPT can return exact prompt that it has passed to dalle when asked. You can usually see it when you click on the image in the chat.

Fun fact - image models like Stable Diffusion have neural nets that are comparable to a bee. GPT4 is somewhere between a parrot and a dog.

It does not translate to intelligence of these animals (since their neural nets control vast number of processes in their bodies) and image/language models are only for one specific task.. but just to give a sense of scale.

[vivid oak]

An interesting test to see how things are grouped is to tell ChatGPT 4 to create an image where there is a beach next to the ocean, a woman with long hair standing on the beach, lots of coconut trees, and pouring rain. And then try to get the ocean, sand, hair, coconut trees, and rain all blowing in the same direction

If you say, "there is an extremely powerful sea breeze" then it will select a subset of it (or none of it) to blow. But if you say, "there is an extremely powerful sea breeze blowing everything" then it will blow everything in the same direction

[mighty marsh]

Got it. Pasted in #images-discussions

(with prompt)

[vivid oak]

You might assume that it would just sort of blow everything to begin with if there were a powerful sea breeze, but that isn't the case

[tender halo]

would anyone be willing to talk to my Discord AI chatbot and give feedback?its a project i have been working on for a while now and need help.. anything would be greatly appreciated!

[vale seal]

Hi folks,
Can't GPT 3.5 generate images in chat at chat.openai.com?

[honest jay]

Hello, you need ChatGPT Plus to use DALL·E on ChatGPT, and it's part of GPT-4, not 3.5.

[vale seal]

alright🥲 thank you

[sturdy oyster]

can someone tell me till what time is GPT’s information last updated on?

[foggy magnet]

April 2023 iirc

[spice star]

you can always ask whatever model you're using, they do know their cutoff date

the one thing chatgpt knows about itself

[sturdy oyster]

he said january 2022

[drowsy tinsel]

# Getting today's date for context
today = datetime.date.today()
today```
GPT4 iq200 moment

[tight cape]

I wanted to ask whether the capping limit is still there or not?

[lavish matrix]

By any chance do you guys know why chatgpt refuses to make me upload any files ? When i upload them they become RED and after getting to somewhat 25% completion rate, i get an error saying unable to upload files.

I've tried 4 different browsers.
I've got my self premium antivirus service to ensure that i don't have malware.
I've respecting all limitations on formats and file sizes.
I've tried different files.

On my phone it works perfectly, but on pc i've been having this issue for now more than a month and i'm starting to give up on it

[wide pike]

my chat gpt one chat it still is lagging

the two chats that are still lagging had to create a new chat

[wide pike]

Anyone know what is going on?

[noble stag]

Does anyone realize that ChatGPT free (GPT 3.5) generate less response/answer to our prompt than few months ago

[maiden fractal]

Prompting have changed too. You have to adapt

[sterile pilot]

Yes

[abstract glen]

Hey

[lunar quiver]

Bruhh,
You can create a Ddos discord bot with chatgpt.....

[lament sable]

I am a GPT plus user from very months and now the GPT 4 is just rubbish from few weeks as it doesnt browse the web to get me the content i want and its response are now too slower. The GPT plugin to access links are also not working and the GPT 4 is now rubbish. Why i am paying $20 for these things which i can get in free GPT. I have resubscribed for GPT just some days ago and i am now thinking of cancelling my subscription and get a refund. As the GPT 4 is not that much beneficial as before. Now whenever i chat with GPT 4 and ask him to provide me some data using web browsing. Its says it has limitation and it cannot access web. So why i am paying $20 if you are adding limitation to the GPT.

[gentle bronze]

are you sure you are using GPT4 with Web functionality or Assistants with that functionality?

[toxic rivet]

hello

[lament sable]

Yes i am using that and i want to know that this issue i am facing is only on my account or on any other account too. The response i am getting to my prompt is given below:

" I have attempted to access detailed statistical data on the Zambia-China mineral trade from various sources but faced restrictions in retrieving or quoting the exact tables you need due to the limitations of the current browsing tool. Given the time-sensitive nature of your request, I suggest directly visiting comprehensive databases like The Observatory of Economic Complexity (OEC) or the World Integrated Trade Solution (WITS) by the World Bank for detailed and tabulated data on the Zambia-China mineral trade. These platforms offer extensive data, including trade values, volumes, and trends over the years, which will be invaluable for your thesis. "

But it actually doesnt try to access the web like it was doing before.

[gentle bronze]

did a search based on your answer and this is what I got:

In 2021, Zambia exported $2.26 billion to China, with major exports including raw copper ($1.64 billion), refined copper ($340 million), and other stone articles ($64.2 million). This marked a significant increase over 26 years, growing from $1.73 million in 1995 to $2.26 billion in 2021. Conversely, China's exports to Zambia in 2021 were valued at $951 million, consisting mainly of tractors ($50.7 million), rubber tires ($38.3 million), and stone processing machines ($27.5 million). This also reflects a growth trend, with China's exports to Zambia increasing from $20.2 million in 1995 to $951 million in 2021【7†source】.

Looking at Zambia's broader trade statistics, in 2021, its total exports were approximately $10.1 billion, and imports were about $6.4 billion, resulting in a positive trade balance of around $3.66 billion. China was among the top export destinations, with exports to China amounting to $1.91 billion. In terms of imports, China was also a major source, with Zambia importing goods worth $810.8 million from China【8†source】.

These figures highlight the significant trade relationship between Zambia and China, particularly in the mineral sector, with copper being a key export commodity from Zambia to China.

and it was on the net

gave me more info, that was just a snippet

[lament sable]

How did it searched for you and it is not utilizing web browsing for me and instead of that it is just telling me how to get data.

[gentle bronze]

it searched and gave me sources with proper links

as to the quality of the information, no clue, I'm not an expert on that subject

[lament sable]

It was using web browsing for me previously but from last week when i resubscribed it ended using web browsing for my account. Should i take a refund and get GPT 4 subscription in another account

[gentle bronze]

seems it's something you have to take either with support from OpenAI

or with the bugs in discord

[lament sable]

Yeah i know that but for me it was not trying to get data using web browsing like it was previously doing

[lament sable]

really thanks bro

i will now contact support team to address this issue

i thought some update has added limitations on web browsing

[gentle bronze]

remember it's holidays, there might be some support constraints in place

[lament sable]

But in your account it is working so i will contact the support and get it fixed. Thanks for your time bro.

[gentle bronze]

👍

you can also ask gpt to make you a txt file for the conersation, then open a new chat, ask it for a search like: "browse the internet for comfort food", then upload the txt file and pick up from there, see if that helps, and that also would enable to search option

[tardy sparrow]

does anybody know dolphin mixtral?

is that like gpt-4?

or more like gpt 3.5

[gentle bronze]

try on #ai-discussions which is open also to discuss other AIs

[drowsy tinsel]

why is gpt4 so slow today isn't everyone off for holiday?

I thought gpt4 is only slow on work days because people use it at work a lot

[gentle bronze]

I think there are people working regardless of if it's holiday or not

and yeah seems somewhat slow today

[drowsy tinsel]

ok most people don't work on holidays

its weird because if u go to openrouter today's demand for gpt-4-turbo is low

[gentle bronze]

I wish that would be true

[drowsy tinsel]

also chatgpt isn't really strictly filtered, like ive seen it saying stuff like s**tposting

[lament sable]

Bro, i just got refund for my account. Previously, i didnt got GPT Plus refer a friend feature in which i can get 3 trial invites to send it to my friend from my account. Now i want that i subscribe to GPT plus from that region which gives free GPT trials. As i am from Pakistan and i didnt got free trial invites in my previous account. So, i want to know if you got that and from which region are you. So, i can use vpn and subscribe from the same region where you are so i can get gpt 4 trial invites in my account.

[gentle bronze]

Pakistan is on the supported countries and territories.

[lament sable]

Bro i didnt get the free trial invites feature in my account previously while subscribing from Pakistan. Do you have free trial invite feature in your account.

[gentle bronze]

no I do not think I have that atm, and I don't think many will currently considering there was a limited ammount of gpt+ plans available already

[viscid moat]

Does anyone have any recommendations for any plug-ins to create a PowerPoint presentation?

[maiden fractal]

best place to ask #1100473149574877234

[viscid moat]

Ty

[regal ember]

is there a way i can make chat gpt 4/ DALL E make more 'extreme' things? cause i cant even generate the most simple things

[arctic gorge]

Nah GPT4 is pretty tight on the regulations

[arctic gorge]

I’d say your best bet is bing’s DALLE since it can generate copyrighted works too

[regal ember]

i need it to make videos with it

[foggy magnet]

Is gpt lagging for anyone else rn

[wild saffron]

is the site down im not able to login i keep getting a verify human loop

[sullen jetty]

Bard never gives me any issues.

[wild saffron]

Yeah but I'm learning to write a short story book with a couple chapters and this is where I started and I want to continue with it because it's me writing there and get to this and it's giving me feedback

[sullen jetty]

Sounds good 💯👍

[dense halo]

Is there a way to get ChatGPT to speak by instructing it by text? Rather than using the voice option and telling it to

[elfin birch]

That is not possible. You can use the API to feed the output into the text-audio model but that is a bit cumbersome.

[dense halo]

Hmm ok, because I don’t know how to pronounce some words so if I can write the word, it can pronounce it for me. Because obviously I cant speak it and say “can you pronounce ….” Because I don’t know how to pronounce it

[valid ferry]

Is there anyone facing such a response about the "90-word limit rule"?: "I apologize for the confusion earlier. Due to the guidelines I must adhere to, I am unable to provide a detailed summary exceeding 90 words directly from the PDF. However, for a comprehensive analysis and specific details, I recommend reviewing the original article in Science Magazine"

[elfin birch]

I recommend using ElevenLabs

[foggy magnet]

......Bing is so much better than chatgpt right now... What on earth.

[tawdry crater]

Does anyone know how you can get updated info for sports analytics with GPT4?

[toxic rivet]

[drowsy tinsel]

hey guys I found a very big bug with ChatGPT webui that has been around for quite a while, can some openai people take a look in #1188891490966052984

[plucky timber]

Lmao what never seen that before but that seems hilarious

I know that it never really read files so this might just be a way they finally just state it to the user

It always would "Skim" files and or read the first 5k chars. Maybe max 1k lines if you persisted.

[valid ferry]

I'm using GPTplus and it has been good at reading PDF files. However, I have been facing this issue for a month now. It's still decent in GPT builder but not as good as it was. GPT became lazy.

[plucky timber]

I've sadly never gotten that experience, and had to resort to upgrade from plus to api for it to attempt to read files on the first ask

I'd assume the lazyness is part of scaling.

#gpt-models message

It could be part of that too, where those abusive users are eating up to much resources and again the regular users ends up biting the bullet 🤷‍♂️

[valid ferry]

I will probably try the same.

[plucky timber]

Allows for a bit more direct control but can get pricier, or cheaper depending on how much you really use it

[eternal rock]

This is a Safe for Work Discord Server

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.

Posting NSFW content will result in a permanent ban.

[plucky timber]

What I like is that you can basically just make your own interface and have a multimodel chat

[valid ferry]

Thank you for the suggestion. I will try this option.

[plucky timber]

Where some questions go to GPT4 and some 3.5 😄

Simple example would be if GPT 4 already gave you a detailed info about your PDF, GPT 3.5 could answer some simple questions based on that info

[valid ferry]

Sounds good

[valid ferry]

GPT4 already gives a simple answer at best or just the "90-word rule", instead of what I asked for lol

[dense halo]

i wonder if we will not see gpt 4.5 or 5.0 this year now because of this nyt law suit 🤔

[plucky timber]

🥲

[plucky timber]

Id much rather see a stable 4

b4 another jump.

[dense halo]

stable how

and with a new model that may be inherrent to it

[plucky timber]

Personally mostly bothered by openAI issues, such as the spammy errors.

So basically i'd say stable as in the interface and interactions with the api.
Some work on the existing things. Like we do got markdown but its incompletely implemented.

For example diff files are/were (I edited my .css so not sure if it still is ) not correctly highlighted

We've seen some more QOL updates recently, but still no wide screen mode ? 😛 (Also a super simple fix)

[ionic geyser]

I have so much data on chatgpt. hopeing some updates will allow all this knowledge to be used or accessed somehow

[plucky timber]

I hope it aint learning from when I scold it 😂

[spice star]

your conversations can be used as training data so maybe

[dense halo]

Happy new year and this to many more

[azure panther]

Hi all, I need help to subscribe to ChatGPT Plus. I got the invite link but it is disabled. I need it urgently for work purposes for the upcoming year.. cant seem to contact the team. Any ideas? Much appreciated.

[dapper isle]

guys, has any of you seem to notice that gpt4 replies pretty fast or am i the only one? since i haent used it for quite a few days! did they update it or something?

[eternal rock]

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 @languid valley immediately.

[obtuse crest]

Is there anyone used chatpdf? And have created the same for gpt4 as custom? Would like to share?

[sonic folio]

@obtuse crest #1171489862164168774 section.

[obtuse crest]

Thanks let me find it

[sonic folio]

Just click the link...

Everybody's GPTs chat with PDFs now, though. So it's unlikely that's a big draw.

[hazy crystal]

I have been ask for a puzle each time I have copy pasted information this is so annoying can anyone tell me if it is only me

or maybe I should ask in a different chat-room also let me know

[kindred parcel]

This is SO UTTERLY PATHETIC. GPT won't even create an image that looks even tangentially related to an established IP. It's trying to pretend to be some kind of digital rights attorney, and it's RIDICULOUS. I want it to make an image for my Cyberpunk 2077 mod collection using the color style and font from the game, and all it will tell me is that it can't. I burned through all 50 of my messages in 5 minutes trying to get this USELESS piece of software to work.

[vivid osprey]

Try Bing ai , i think it will work with you

[brisk path]

Not their fault that they'll get in trouble for creating an image of a copyrighted character

[swift osprey]

the grandma level of caution is horrible though. It even comes up in basic coding questions

[brisk path]

What's wrong with it though?

It's not the devs fault that people take advantage of this tech

So they have to put guardrails to avoid jailbreaks and such

[swift osprey]

And I'll still complain about overly cautious coding advice

[brisk path]

I mean, you could stop using it if it bothers you that much 😉

[swift osprey]

Yeah I hate traffic too and getting oil changes. Guess I'll just walk

durrrr

[sonic folio]

The font itself is copyrighted.

[kindred parcel]

I didn't ask it to create an image of a copyrighted character. I specifically told it that it shouldn't.

[kindred parcel]

Even a cursory Google search proves you dead wrong: Is the cyberpunk font copyrighted?
The Work consists of a short phrase in typeface and familiar geometric shapes—lines, dotted lines, and circles—that are not protected by copyright

[sonic folio]

No, it shows free fonts, not fonts without a copyright.

That's just not how copyright works.

[kindred parcel]

Except that I am literally looking at the denial of the copyright / trademark for the font from the US patent office. Of course I can't link it here because of whitelisting, but google "Is the Cyberpunk 2077 font copyrighted?" and you will find the document.

[kindred parcel]

And further, IT IS A FREE FONT. It's on FontsRepo.

[sonic folio]

Then use it to give DALL-E 3 a reference image?

[kindred parcel]

It flat out denies my request. It's so terrified of violating its "usage criterion" that it won't even attempt it. It's absolutely broken, and truly pathetic.

[sonic folio]

You're telling me that you tried a reference image?

[kindred parcel]

I'm telling you that I asked it about 40 different ways, and it just flat out refuses to do anything even potentially associated with it, an established IP. It's absolutely ridiculous, as two weeks ago, I had it make all kinds of images of COD MW3, and it had no problem whatsoever.

[sonic folio]

Then use it to give DALL-E 3 a reference image?

[kindred parcel]

You're being oblique, and it isn't helpful.

[sonic folio]

I provided you a solution. You complained instead of trying it just to be right. GJ

[kindred parcel]

Telling me to "use it to give DALL-E3 a reference image" is not helpful, and you know that. Use what as a reference image? Just an image of the logo? Just an image of the character? Just an image of a mod collection? You are being intentionally oblique, and it amounts to concern trolling.

[tight island]

well Happy new year

[sonic folio]

Happy New Year

[tight island]

its now 2024 here in 🇩🇰

[sonic folio]

Use the font to give it a reference of the text you want to use.

Send the prompt with the image.

Tell me more about concern trolling though.

[kindred parcel]

Tell yourself, you clearly know all about it.

[split kelp]

happy new year!

[swift osprey]

😂

[sonic folio]

Happy New Year @split kelp

[dense halo]

I love ChatGPT

[magic matrix]

whats the max upload size of gpt4

[sonic folio]

20MB, I believe.

[radiant raft]

yo

why on earth is the chat window so thin on the pc version of chatgpt like it could be 75% of the screen but nooo it has to be 50

for this reason i think AI is not the next big thing

[molten tundra]

Ok super smart people. Need a brain blast moment. I'm able to get the general chatgpt 4 make Python scripts that I can run in blender and create 3d models with. So far the most advanced I was able to make it do was some particles and extrusions.

But I'm falling flat of actually creating a full on character.

Would it be
A) better to find a free blended asset and then have AI script it to rig, animate, and do other things with?
B) Generate a custom bot that has all of the naunces and such for the blender terms and python scripting, maybe even documents of complex python coding to try and match the output level im trying to make
C) search online for custom gpts that may have already been made for this process

[sonic folio]

C>A>B

[sonic folio]

@dense halo I have a suggestion you might not like. Create a custom GPT. Have it analyze your writing by uploading samples you wrote. Then have it mimic your writing style from the samples, integrating that style into any output.

Yeah, no doubt. It's a qualitative essay. Give it more context and a qualia tagging framework if you want qualitative essays.

I mean, the AI is going to flounder in the realm of the subjective without some serious guardrails.

[waxen chasm]

https://platform.openai.com/docs/guides/prompt-engineering

[minor meadow]

why is there a word limit per hour

[naive prawn]

Hi can somebody help me when you use up your time/msgs limit on gpt 4 when you are allowed to re use it you can send 1 or 2 msgs then it will stop working completely when u open a new chat it works just fine please help me

[honest jay]

Hello, the GPT-4 usage cap rolls, it doesn't reset. This means you can use it 40 times in 3hrs, but you get each of the 40 uses "back" one at a time, 3hrs after you used it originally.

Separately, custom GPTs are currently capped at 25/3hrs, but still as part of the 40. So if you use a custom GPT 25 times in a 3hr period, you can still use the regular GPT-4 model 15 more times within the same 3hr period.

[naive prawn]

then why does it keep on saying. Hmm...something seems to have gone wrong. even tho they time limit has already finished

[honest jay]

Is it a long chat by chance?

[naive prawn]

around 42 msgs

[swift osprey]

Oh man, It would be so crazy if ChatGPT told you how many messages you had remaining for each 3 hour period

Can't even imagine that technology

[final spear]

Yo I put two ai’s together and they made plans for two weeks from now

[swift osprey]

and then one ghosted the other after 50 messages

[austere plaza]

Actually, there's a (browser) plug-in that follows this sort of stuff, but it seems rather busted for me.

Why it's not native to ChatGPT, I don't know. There's a lot of UI deficiencies for generative AI lately

[swift osprey]

Yeah I had something like that before, just crazy to me the lack of features the web gui has

it's like nobody with decision making power at openai actually uses the web gui very extensively

looks good at first glance and for a good amount of messages, but using it heavily daily there are weaknesses

[austere plaza]

Perhaps when they push out GPT-5, one task they'll sic it upon is to develop a GUI

So long as GPT-5 is coherent enough and has a long-enough output window, that's not even a joke

[swift osprey]

whenever gpt-5 gets here it's gonna be awesome

[austere plaza]

At this juncture, I don't know if I'd rather GPT-4.5 or just an increase to the post limit for ChatGPT

I guess if 4.5 has massively longer output windows, that one wins out easily.

[swift osprey]

I don't know enough about the versioning to have an opinion

[austere plaza]

But if it's more of a "fine-tuned" GPT-4 rather than a qualitative leap forward, I'd actually rather just have something like "120 messages per 3 hours" instead

[swift osprey]

Idk. We'll see what GPT-5 has to offer I guess. I don't usually hit the current message limit unless I'm learning something new and have a lot of easy questions which burn through the message limit

Seems like they should be enforcing a token limit per timeframe rather than messages per hour

[austere plaza]

GPT-3.5 is a fine-tuned version of GPT-3, not an architectural improvement. It's much better than base 3, but mostly because of better data optimization IIRC

I presume 4.5 will be the same way for 4

More of a "coherency, commonsense reasoning, and logic are improved, and hallucinations are reduced", with a longer output context window being the cherry on top

Like maybe a 16k token output instead of the 4,095 max of GPT-4

Even if the overall context window is not increased beyond 128k, being able to output upwards of 16k tokens in one go, and those tokens being far more coherent at that, would be downright magical for productivity. You could probably generate a full novel in under ten minutes, or code some relatively complex games from scratch

[little oak]

Anyone else having to answer inane and completely unrelated puzzles while using the model? I'm plus subscriber and this is new behaviour which hampers my flow. The puzzles are easy to make a mistake in too.

[sonic folio]

Captcha FTL, automate with GPT-4 Vision! 🤣

OpenAI promotion: e'erybody needs a captcha automation account in 2024!

Seriously though. This is so dumb.

[honest jay]

Do you have a VPN installed? What browser(s) is it happening in?

[little oak]

I do actually and I use DuckDGO

That actually would explain it in all fairness. Just annoying as it hadn't happened before and I generally always use a VPN

[tight island]

Dear Sam, pls release gpt 4.5. ty in advance.
yours truly.
Starlight

[opal jolt]

Nicholas Dunzelman: gpt4.5 leak legit or no?
Sam Altman: nah

[swift osprey]

It would be amazing to attach GPT to the full scope of my web application's code and provide that context while asking questions

[steady path]

ahh im goning crazy with the usage cap

[swift osprey]

same bust mostly because they don't tell you how many messages you have left for the 3 hours

[low dirge]

any issue with gpt4 (chatgpt) , it cant even read a csv

getting below message ,

"It appears that there was an error in attempting to load and display the contents of the uploaded Excel files. Unfortunately, due to my current environment's limitations, I'm unable to directly access and read the contents of uploaded files."

[tight island]

im having an issue, i cant seem to figure out how to connect my Assistant API to Custom GPT

[sterile flicker]

I was just trying to figure out how these are connected. Are Assistants custom GPTs or are these two completely unrelated?

[tight island]

technically they are two different things

[alpine wedge]

i dont do this whole pronoun thing

[dark jungle]

Is there an alternative to GPT? that is good with producing python? Im unsure what's happening recently, but i simply cant deal with the slow speed and useless responses any longer. I've been a paid user for several months and often praise GPT, but now, it no longer meets the basic needs it previously could address effortlessly. I'm happy to continue using GPT when it gets back up to speed, but for now, i need an alternative solution.

[simple pasture]

Very factorious truly true

[alpine wedge]

Is there an alternative to GPT? that is good with producing python? Im unsure what's happening recently, but i simply cant deal with the slow speed and useless responses any longer. I've been a paid user for several months and often praise GPT, but now, it no longer meets the basic needs it previously could address effortlessly. I'm happy to continue using GPT when it gets back up to speed, but for now, i need an alternative solution.

[tight island]

Usually in this case you would utilize the API.

[brazen nova]

Looking for help: I am getting 5 Re:CAPTCHA puzzles for every sentence/entry into GPT-4 on desktop. This seems to be a type of ban on IP level since I created a new pro account and it's the same thing from start. I am incredibly frustrated. This has been like this for 2 months. No reply from OpenAI. I am only using it for incredibly benevolent things, like basic market research and content creation, work and school stuff. Never got any warning of wrongful use in a chat at all.

[tight island]

i aint clicking that

[west gull]

Why can't I search my ChatGPT's in desktop mode? 😩

[ornate zealot]

I opened that in a sandbox, it is a gift link, but why it's like that, I'm not sure. Looks like a scammy scam

It is asking me to login, but I ain't about to do that.

[tight island]

nobody sends gift links anymore.

[ornate zealot]

I sent a gift to someone before, and it doesn't look like all that.

[south kayak]

When are the iOS bugs getting fixed?

[devout frigate]

Does anyone know a reliable source that shares / talks about CustomGPT's? I want to try some and see what people are making

[sonic folio]

https://discord.com/channels/974519864045756446/1171489862164168774

Please consider checking out https://discord.com/channels/974519864045756446/1191075231264157747

[paper kestrel]

Hello Copilot offers GPT 4, is there a reason to pay for GPT 4 now, is the copilot one bad??

[normal quarry]

they are different products

[paper kestrel]

I am not talking about Github Copilot, I mean Microsoft's Copilot!

[normal quarry]

I understand but they are different products. ChatGPT GPT-4 gives you DALL-E 3, Browsing with Bing, Advanced Data Analysis and custom GPTs

[paper kestrel]

Most of us use only GPT 4 and not the rest of them, I am debating whether the price tag is worth it.

Copilot also has DALL-E and browsing with Bing and uses the new 4 turbo model.

Advanced data analysis and CustomGPTs are the only thing we pay for.

[normal quarry]

then use whatever product suits your usecase best

[frosty laurel]

Chatgpt is down for me, anyone else got this problem?

Just see a screen that says "unable to load history" and no option to log out or anything

[naive prawn]

Can some one help me it keeps on showing me this and idk what to do

Hmm...something seems to have gone wrong.

[frosty laurel]

Ok, looks like the issue was temporary and it's working again.

[quick ridge]

Hey, is anyone also experiencing issues with ChatGPT?

It completely froze up for me.

[smoky sphinx]

Why am I asked to Verify if I'm human every freaking single time I send an input to gpt+!!! This has been happening for 3days now and i
ts starting to get on my fuvking nerves

[quick ridge]

My ISP is screwed up.

[worn roost]

ye same

[quick ridge]

Good to know.

My ISP completely died on me and now gpt works fine now lol

[eternal rock]

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 @languid valley immediately.

[rich surge]

Bing's Copilot's GPT4 and BIC is not worth it in my opinion. It's just a very limited GPT4 in terms of knowledge and processing and not fun by comparison. I can dish out much more powerful components with my Custom GPTs than I can with Bing's Copilot. I find that Bing's Copilot is more for casual users who just want something simple and easy, and who are not interested in exploring the nitty gritty and full capabilities of GPT4.

If I were to choose, I'd pick the paid version of GPT4 because I can do so much more. Yes, I'm still limited, but I have so much more access to more powerful tools that it's worth is well placed.

[oblique fractal]

I found Bing/Copilot to be quite good at providing nuances for technical question, sometimes I prefer its answer over what I can get from GPT-4; more nuances and less prompt. Biggest drawback for me is the format is not as nicely formatted. I certainly had a share a glitches and the impression of less performance and looking at the comment on this chatgroup, this is not a problem exclusive to Bing. I sent quite a few feedback in the last 1-2 month, some got fixed fairly quickly. To me it's more understandable because I am running the preview version.

[rich surge]

Yes I understand

[paper kestrel]

Hello, thank you so much friend. I have been on the fence thinking if I should pay, because even 20 usd is a lot for me as a student.

[rich surge]

It's well worth it if you want to dive deep. As a student, I'd recommend exploring it because it has many tropes that can help you with your studies, side hustles, and hobbies.

[paper kestrel]

I cannot emoji this, so I will say thank you again XD.

Are you a developer?

[rich surge]

Yes, somewhat.

I know enough to direct people.

[rich surge]

Any time

[oblique fractal]

I'd say it's worth it as well. Make sure to give a try to the out of the box GPT-4, and in the custom GPT there is a prebuild one named GPT-4 classic; make sure to try that one as well. I also have a crude and work in progress GPT with the goal set to provide shorter summary, and in theory to provide more personalized guidance on how I could unblock myself but I have yet to try that one. IMO if I want even more contrast or if I am looking for more references Bing can be surprisinlgy useful. At that point I think it's a matter of preference. If you log-in with a microsoft account you should have the ability for 30 replies instead of the standard 3 or 10. I can 't quite remember what it was exaclty and I don't excatly log-off just to know.

[paper kestrel]

Thank you friend Michael! (I hope this isn't a Sam Altman alt)

I will keep paying XD

I have not used much custom GPT, only the regular one, but I think I will start to do that more from now on.

[oblique fractal]

Yeah, I was skeptical at first about using prompt to customize my chat experience and after giving it a try it's a great feature! Not 100% perfect 100% of the time but it does what you would expect it to do. And I forgot to mention this prompt guide https://platform.openai.com/docs/guides/prompt-engineering

[nova fable]

Hmm, gpt having a lazy day? Can't get it to check the knowledge in my gpt till after three requests to do so, then it still answers the lazy uninformed way after doing so again... lame

[eternal rock]

`` Server Tip: ``

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[sonic folio]

@nova fable Something worth considering: Chat-GPT doesn't give a temp setting. If you get bad results, kill the instance and roll the dice on a new one. Sometimes, it's not that the model's being lazy. Sometimes, its temperature is too high to do any reasoning about tool use. I see this in Custom GPT instances, very occasionally.

[nova fable]

Hmm k, i'll check it out

[obtuse jolt]

How is the usage rate for Plus calculated? A few months ago someone said that a message once every 8 minutes will basically give us infinite messages (sure no one would do that but it acted as a guide). But that doesn't seem to the same now. I knew I would hit the limit for what I was doing. So I took a break for an hour doing something else. After I resumed, within a couple of messages I hit the cap. I don't remember that happening earlier.

[sonic folio]

Apparently there's a bug. It's already been brought up elsewhere today. It is alleged to be very rare on per-account basis but with a large sample size emerges a few times a day/week/whatever.

[honest jay]

It's a rolling usage cap, which means each use can be thought of as having its own 3hr timer before getting that use "back", which could explain what you're describing. In other words, when it's telling you you have to wait until x time, that just means that's the time you get one of your uses back--not all of them.

With the current cap of 40/3hrs, you can use 1 every 4.5 minutes and never hit the cap. However, custom GPTs are capped at 25/3hrs (within the 40) so keep that in mind.

[minor birch]

So I have a GPT Plus account with a different email than I use with the Google Playstore. My assumption is that Plus subscription should give me access to the premium content on the OpenAI apps if I can link accounts. My question is 1) is this a true assumption, 2) if so, how do I link my Plus subscription to my apps given that I use different emails for them?

[honest jay]

Hello, can you please clarify what you mean by "OpenAI apps"? In general, ChatGPT Plus "only" gives you access to the features of Plus -- priority access, GPT-4, beta features like DALL·E, custom GPTs, etc. -- basically all the features of chat.openai.com. It doesn't relate to API billing, usage, or access, if that's possibly what you mean by "OpenAI apps". You can use the Plus features in the official ChatGPT app, though!

[minor birch]

Referring specifically to the app ChatGPT by OpenAI on the Playstore. I suppose I should have said "app" singular; don't know (assume does not) give access to other apps that use GPT (e.g. AI Chat Open Assistant Chatbot by Smart Widgets Lab)

[honest jay]

Gotcha gotcha, sorry for reading too much into that! Yes, a ChatGPT Plus account will work on all platforms logged in with that account. When you install it from the Play Store, just be sure to sign in with the Plus account in the app itself, and it'll give you access to the Plus features.

I don't think you can currently subscribe in the app on Android, so you'll have to subscribe at chat.openai.com, but then you should be good to go!

[minor birch]

That's my problem: I don't know how to log into the Plus account on the app itself. My Playstore email A, my GPT Plus account email is B, and I don't see anywhere in the app that I can signin in the app itself

[minor birch]

Forgot to hit reply

[honest jay]

Be sure that you're using the official ChatGPT app, you can find a link at https://openai.com/chatgpt. If you have that and you're already signed in to the wrong account, just open the chat history sidebar on the left, tap the three dots by your name, scroll to the bottom and tap "Sign out". Then, just re-sign in using your Plus account.

[minor birch]

I had the right app. Tried uninstalling and reinstalling, and that seemed to do the trick. Best I can figure is you have to connect the accounts when first opening the app, because I don't see a way to do it in the app itself. Ah well. Thank you for your help

[rigid hornet]

How is it that chatGPT is being used for enterprise data analytics? If I ask it to process information that requires it to present a list greater than 10 entries, it starts to get very sketchy as to whether or not it's actually going to bother doing it, even if it says it will..

sometimes it's not unlike begging it to at least attempt a task. Maybe after 5 or 6 nudges it'll go ahead and do it

[errant thorn]

What if we trained model on compressed data and teach it to output uncompressed alphanumeric characters?

[sage yacht]

Hmmmmm

[minor vortex]

Does anyone get the error message, "This conversation is too long. Please start a new conversation." This is a new development for me, and it can be somewhat of a problem if I'm trying to develop and AI RPG Gamemaster.

[hollow marsh]

Hey guys, has anybody had success with getting GPT4 to write a text while staying accurate to factual information described in an attached document ? I wanted it to write the story of a real event in the style of a particular author, so I created a custom GPT with instructions, a book by the author and reference documents for the event. Once the story was finished I asked to check coherence and accuracy and it said all was ok. But I found out that GPT had given a different outcome to a key character in the story, making it factually inaccurate. I wonder if this issue is inherent to LLM AIs (creative work can't be 100% accurate) or if it's because I gave it a task just a little too creative.

[boreal lily]

start a new one

[swift osprey]

I wish there was a button to just clear the chat instead of needing to start a new one every time I want to reset the chat's context

or a 'do not track conversation context' option

[ionic geyser]

chatgpt is so stupid

it just doesn't follow instructions and gets confused so much. GPTs suck.

[boreal lily]

it might be skill issue,
check your prompts
make sure they are below the token limit

[ionic geyser]

they are

its like 1-2 pages of instructions

i just want it to follow a format where it researches in one section and generates image on another section

but sometimes it generates the image first or says it will research and then doesn't

Is it because I put the boilerplate of responses in format.txt?

How can I get ChatGPT to stop researching in the background or thinking it can

I literally just want ChatGPT to extract the Task out of the prompt, then perform any research, then generate images, then respond

I try asking it simply and verbosely it just wont do it

it just now generates the image first thing

i cant seem to control the order of the image generated or research done with any consistency

Impossible to do such a simple task... pathetic

[stone solstice]

any chat ai that is free that can access websites with username and password and do tasks

[scarlet dawn]

gpt 3.5 or 4?

[swift osprey]

try telling it that in a list of instructions? dunno

[scarlet dawn]

oh 4, ya give it a numbered list of instructions you want it to do

[ionic geyser]

Ty will try ...

[ionic geyser]

Doesn't work

I DID IT!!!

Sorta... anyways...

Asking chatgpt to initiate research, then generate image, then paste/summarize research worked

[boreal lily]

not any web ones i know
guess u make it urself

[dense halo]

Chatgpt has become very laggy for me and is even causing the "This page is slowing down firefox" error to occur. Is it time to change chats or is there a server bug atm?

just wait for some time

[rancid saffron]

hi everyone, i'm facing a problem, i'm trying to make gpt make an image with a path that goes from the left to the right of the image and it keeps creating path vertically, what can i do? should i ask it differently?

[sonic folio]

@rancid saffronperpendicular to angle of view probably will work.

[rancid saffron]

i'll try, thank you

[wraith ocean]

down for anyone else?

[eager void]

Yes

[normal mantle]

I would like to ask, nowadays gpt4 is getting lazier and lazier, the output code or analysis is omitted when it can be omitted, and sometimes it can't even understand what I mean. Do you have any recommendations for prompts or papers that study optimization for this problem?

[tidal pendant]

Is there any bigger plan than $20? Because I am hitting the limit of gpt-4 everytime. I am currently getting 40 requests. But I need more. Please suggest if there is any solutions?

[maiden fractal]

The api

[subtle spear]

Does anyone have the problem that "Signup is currently unavailable, please try again later."
I cannot sign up now😫

[heavy solstice]

Running into some serious problems with gpt vision. Worrying!

[simple adder]

i am using free version ie 3.5 and in want stt and tts options. what is the best way to apply those feature in browser

[maiden fractal]

Chatgpt plus with plug in

[dusky cliff]

happy new year

[spiral gate]

If i pay for the GPT4, then why is there a cooldown?

[maiden fractal]

It is like this. Everyone have 40 prompt /3h

[serene horizon]

Is there anyone here that has had really good success with your custom GPT actually sticking to the custom instructions and always checking them again before responding?

[spiral gate]

Okay.

[serene horizon]

There just is at this point. Part of the package

[serene horizon]

What problems are worrying?

[heavy solstice]

e.g. see https://discord.com/channels/974519864045756446/1192076387432013925

[serene horizon]

responded in there

[kindred parcel]

GPT-4 broken for anyone else? I just keeps saying "Hmm.. an error occured" and "regenerate" does nothing...?

[clever orchid]

Is anyone else having trouble saving custom GPTs?

[ionic geyser]

Which is horrible

[maiden fractal]

You can switch for the api. Limit change over the demand. Since it is the most advanced ai existing atm. Millions of people using

And cost a lot to run it

[ionic geyser]

hm but then i dont have 'chat'gpt

[restive plover]

Any new features coming?

[ionic geyser]

I hope so. UX/UI have sucked so long and they just refuse to update or do anything unless it's completely breaking the product.

[tight island]

i mean the new ui was leaked a few months ago right?

[ionic geyser]

didnt see

[fringe path]

Does chatGpt or open Ai has a policy against its use for sports betting analisis ? I have the paid version and been trying to use it to for sports betting predictions but its just stops searching on the internet for data, i am imputing my own data but its just refuses to do internet searches when its sports betting related. anyone has expirience with this type of issue ?

[scarlet dawn]

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

Gambling

[dense halo]

hey

[main perch]

What's the best channel to talk about embeddings?

[scarlet dawn]

probably #dev-chat

[main perch]

I was thinking maybe that or #community-help

[pseudo tide]

Who is facing issues with Dall-E?

All the time it's giving errors

[dense halo]

maybe use the api instead

[tall radish]

seems like chatgpt can't get this super simple question right no matter how many retries.

In a 2016 study, Eastern Washington University psychologist Amani El- Alayli found that, among the study participants who experienced frisson (a physiological response akin to goosebumps or getting the chills) while listening to music, there was one personality trait that they scored particularly _ openness to experience.
Which choice completes the text so that it conforms to the conventions of Standard English?
Choose 1 answer:
A high on;
B high on
C high. On
D high on:
anyone know why?
btw the correct answer is D

[brisk path]

It's because ChatGPT doesn't handle individual characters

[tall radish]

so far no ai i've tested can get the right answer. Bard, claude, no one.

[brisk path]

It handles tokens, which are around the size of a small word, but could be larger or smaller depending on different factors.

All LLMs handle it through tokens.

Since there isn't a per-character understanding

It has trouble with questions like these.

[tall radish]

it's worked for other questions just not this one for some reason.

[brisk path]

Punctuation specifically is pretty hard for the AI here too, which means it sometimes misses a punctuation mark

[tall radish]

ok

[brisk path]

Bard, for instance, misses the semicolon usually after A. While it wouldn't be correct, it's a reasonable mistake.

[tall radish]

ok

[twilit hatch]

GPT-4 is super slow very often. Damn. There, I vented.

[boreal lily]

chatgpt down

or very buggy

[frank turtle]

Just came here to say the same

[marble charm]

yeap. agree. also I find this last update they did on gpt 4 very annoying. It is very laconic in its responses, forcing me to unlike every answer and choosing the "new answer" everytime... also they imposed a limit of prompts and responses on every conversation, forcing to start a new one, and build the context again.
As it is, I prefer chatgpt 3.5 rather than 4....

[twilit hatch]

I feel the same. It’s becoming not worth it

[marble charm]

I suspect they imposed a new computation limit to allow more people in and more messages per 3 hours

I would get the job done with a more restricted cap, but with more computation power - meaning better responses...

[marble charm]

yeah. gonna give it some time, but if this continues, I'm gonna abandon plus...

[ornate zealot]

Perplexity can, but it is cheating.

[naive prawn]

SAME does anyone have a fix for it

why when you reach the 40msgs/3h it just stops working forcing you to start a new one

[jade crater]

Anything I should know about privacy as it pertains to uploaded files?

[wintry gorge]

Dude why is chat gpt so slow

Takes so damn long to get my answer

Anyone else have the same problem?

[signal sigil]

Does anyone have the same problem? ChatGPT is not writing code to me or searching the internet. Instead it keeps telling me i should do this or that. But i told it to write me code or look something up. When i tell it explicitly again it will do so but this is nerve wrecking

[gloomy thistle]

I hit the "usage limits" twice today. And the last time after only 12 questions (I counted them)! Are they having performance issue, maybe?

[torn cloud]

I keep getting "Hmm...something seems to have gone wrong."

[burnt shadow]

Has anyone figured out the best way to favorite or save specific chat histories to come back and pick up where you left off?

[eternal rock]

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 @languid valley immediately.

[honest jay]

The usage cap rolls, so it only fully resets after 3 full hours of no use. Think of each use as having its own separate 3hr timer, and you get each back one at a time after each 3hr timer.

[pine tulip]

Each conversation has a custom link. Since there is no feature to favorite chats, you can save the link in a note taking app of your choice.

Also, there is feature to archive chats which you could misuse as a favorite button, since they are all saved in one place.

[pine tulip]

this! Never though about explaining it like that but thats exactly how it works. If 40 "timers" are active at the same time, the usage cap is reached. From now on, I'll only explain it like that

[pine tulip]

Anything you upload and write can be used by OpenAI as training data. Sensitive data should therefore be anonymized or omitted

[honest jay]

GPT-4 mana regen bar WHEN 😁

[pine tulip]

Thats just another genius way to explain that! How do you come up with that? 😂

[honest jay]

That one came from an idea in someone's suggestion thread for a visual indicator for GPT-4 usage! #1176996746778386544 message

[burnt shadow]

Oh! Good idea, Can you recall in archive chat then and pick up where you left off? Additionally do these ever delete like the history does?

[pine tulip]

Oh, I see why this was suggested. However, I can imagine that OpenAI deliberately does not present this in a completely transparent way. If you're unsure when you'll be able to write the next message, you'll be more cautious and won't push it to the limit

[pine tulip]

Yes you can, but you need to unarchive it to write in it and archive it again to "save" it again. Could be a hassle real quick

[burnt shadow]

Yeah, I see that. But not a bad idea. Otherwise I can just save the links to the chats too. Will those links work even if ever in the future they are gone from the history?

[pine tulip]

Wdym with "gone"? As far as I'm aware, chats don't get deleted, only if you delete them yourself or if you turned off chat history & training in the settings

Or did that happen to you before?

[burnt shadow]

Oh interesting, I thought they would eventually go away. I guess you can scroll through them it seems. No worries! I will try these suggestions out!

[pine tulip]

I just tested it to be safe and scrolled for over 5 minutes 😆 I still have chats from 2022, so they most likely won't be deleted

[burnt shadow]

awesome! Good to know! Thanks once again for the information!

[pine tulip]

No problem at all. Good luck!

[drowsy tinsel]

I wonder if the old chats from gpt4 will be converted to use gpt4-turbo or still the old gpt4

[pseudo tide]

40 messages on ChatGPT 4 / 3 hours it's really bad

[dim kraken]

Hi folks. I have a question: how can i give a chatgpt my repo link in order to ask some questions about my code. Is it possible? Because right now it does not want to do so

[light glade]

You can try using GPT4 if its a public repo. I was able to do something using the GPT builder

[dim kraken]

I am using gpt4 but it does not want to look into files just overall structure of my project((

[pine tulip]

It's fine. It just forces you to formulate your prompts carefully before sending them. If you do that, you almost never reach the limit

[light glade]

I just tried this with GPT4, it seems to be able to review the project and share comments

[pine tulip]

It can't read links

[pseudo tide]

For me it's you giving a excuse for something 🤔

[pine tulip]

But you can upload your files

[pine tulip]

Sorry?

[pseudo tide]

It just don't make sense. Why should I pay for something and also have limits?

And it's getting worse everytime

It released without limits, 100, 50 and now 40

[pine tulip]

If you purchase bread, you get one bread, not two or 20

[pseudo tide]

I think ChatGPT it's obviously not a bread

And I haven't buyed in the past with 40 limits

[pine tulip]

As soon as they have more capabilities, I am sure they will up the limits

[pseudo tide]

lol

[pine tulip]

ChatGPT needs ressources, just like a bread

It's not a digital product, just the frontmatter is

[pseudo tide]

Okay, you buy a bread that is decreasing everyday?

If you sell something that is "X" size and then decrease everytime

You will buy again?

Obviously don't

[pine tulip]

With more buyers, bread becomes scarce

[pseudo tide]

So don't say "just formulate better your prompts"

You're saying like "Don't make more bread with eggs", cut in 4 parts and make with eggs

that just don't make sense lol

[tepid patrol]

Is ChatGPT 4 down? I get network errors and it is quite slow

[pine tulip]

I can't follow you. As I said, it is not a digital product but requires real world resources. If more people use it, there must be limits to everyone being treated fairly. I'm also hoping that OpenAI will introduce a second plan for maybe $100 so that you have a higher limit.

[pseudo tide]

ChatGPT 4 it's worse than ever, probably is

[pseudo tide]

What are you saying just don't make sense bro, that's it. If you sell something X and start changing everytime don't expect happy costumers

ChatGPT changes every time I log in

While other alternatives are free and not as restrictive as ChatGPT

It no longer makes sense for me to pay for something that decreases every time I enter. If you sell something as X, you can't keep changing it to Y,Z,W

[pine tulip]

Apparently, we have very different experiences. It works fine for me and doesn't change daily. If at all, it's monthly or even bi-monthly.

[pseudo tide]

And besides, there are problems all the time and I spend "my resources" on something that doesn't work. Just yesterday I had +20 errors

[pine tulip]

You are free to use alternatives

[pseudo tide]

Dall-E started showing 4 images, right now it's almost 1

And poorly

[pine tulip]

Mind telling me what exactly changed? I can't understand what you're saying, but I'd like to

[pine tulip]

Well, I can generate 10 images with one prompt thats carefully crafted

Also keep in mind, that a lot more users joined - it's not a digital product

[pseudo tide]

At the beginning of ChatGPT, Premium had no restrictions. They put 100, reduced it to 75, to 50 and now 40.

And Dall-E today has several restrictions, errors when generating it and only 1. In the past every prompt generated 4 different images

But man, I don't want to extend this discussion. I'll just stop paying and waiting them to improve. You're defending something that doesn't make sense for me

If you want to go to a bakery that's decreasing every time size of your bread, it's okay.

[pine tulip]

ChatGPT with GPT 4 never was 100, and it never was 75. It started with 10, was upped to 20, upped to 30, upped to 50 and now reduced to 40. You can still use ChatGPT with GPT 3.5 more than 40 times/3h. Your statement is simply not true.

[pine tulip]

If it's the best bakery with the best bread in the world, I will shop there again and again, even if they only sell half the amount.

Because it's the best bread out there right now

[pseudo tide]

Just search.

I was not possible to send you a link

Use your ChatGPT 4 to search for you

[pine tulip]

I don't need to, I used ChatGPT every day since march

[pseudo tide]

So probably you never looked

xD

[pseudo tide]

Wow

[pine tulip]

since you? sorry?

[pseudo tide]

You can just search to reddit forums and see.

[pine tulip]

See what?

[pseudo tide]

"GPT-4 Message Cap Doubled to ~100 Every 3 Hours." "GPT 4 Plus - Limit 100 msges per 4 hour decreased to 50"

You can just search this titles and look

6 months ago ChatGPT has 100 messages limits

[pine tulip]

But that was right at the beginning of the release. You make it sound like you've already gotten used to 100 messages, even though no one has had the chance to do so

[pseudo tide]

Like I was saying.......??????????

[pine tulip]

Servers were overloaded, so they reduced it - as they should because no one could properly use it back then

[pseudo tide]

And then you sayed that my "statement is simply not true"

so maybe you have never seen, even using since March right? 🤔

[pine tulip]

No, you're saying you're paying for a product that's getting worse and worse. Although right at the beginning it was very bad due to the overload and got better due to the reduction

[pseudo tide]

The only number that I don't remeber it's 75, maybe was 60 ou 70. But they decreased after the 100

[pseudo tide]

In my point of view, it's worsing

Because I've seen the limits decreasing

and you sayed that are just increasing? lol

[pine tulip]

Yea, limits went down way further before it was a usable product. Since it was usable, they steadily increased the limits

[pseudo tide]

So don't say I'm lying if you don't know what you're talking about

Because you didn't even know that are a 100 message limits in the past btw

[pine tulip]

Let's stick to the bread comparison. If the best baker who makes the best bread can't process my order because he misjudged the rush for his bread, I might be angry for a moment, but I still don't miss out on the world's best bread, lol

[pseudo tide]

Just stop man, lol

[pine tulip]

I'm not saying that you're lying, but that your statement that ChatGPT has been lowering its limits further and further is not true. The limits were only adjusted until it was usable

[novel parcel]

GPT Store will launch Next Week

The problem is i cancelled the subscription

[pearl kettle]

but no revenue sharing with the gpt store now?

it's not mentioned in the email

[lunar arrow]

I don't recall GPT-4's cap being more than 50

It was 25/3h or so on launch

[honest jay]

The first like 2-3 days after launch were highly variable. It technically launched at 100 but it very quickly settled at the 25 point.

[maiden fractal]

It was 50 for 1-2 weeks before the 40. And before 40 is was 25.

[dense vigil]

Anyone here got GPT4? If so is it worth it? Please don't be biased and tell me the pros and cons

GPT+ *

[glass fulcrum]

for me the GPT+ is gooood
pros

• GPT4 :)
• GPTs
• DALLE-3
• unlimited gpt 3.5 (idk if its true but if there is limit i didn't hit it)
  cons
• 20$ :(
• only 40/3 hour message limit for gpt4

something more you want to know ?

[dense vigil]

Im guessing that DALLE-3 is included with the message limit?

[glass fulcrum]

yea but it look like DALLE-3 is a diffrent limit than gpt-4 but idk sometimes i use dalle adn it tell me that i used all limit but i still can use gpt4

[dense vigil]

ok, Ill get it then.

[glass fulcrum]

yea its very good

[maiden fractal]

Forgot plug in

And custom instructions

[glass fulcrum]

oh true i forgot them i not use them a lot :p

[dense vigil]

Actually. Can I get another opinion pls?

[glass fulcrum]

so it means its even better hhhh

[honest jay]

DALL·E requests count as part of the 40/3hrs GPT-4 cap, but DALL·E's daily cap itself is 200 images.

[maiden fractal]

Did a pro poker test chatgpt 4 won over chatgpt 3.5 and gpt 4

Using chatgpt 4 for my rpg game it does the map over 1 single picture as reference like a swamp scene. Doing picture based on my description

[dense vigil]

huh? I mean just like more Pros and cons from ur perspective pls

[maiden fractal]

Pro chatgpt 4 is better in everything . Even doing internet search

Con : yes it isn't free

[glass fulcrum]

there aren't many cons

[dense vigil]

Ok then thank you

[maiden fractal]

You have a limit but you can do 1 prompt every 4.5 min and you never reach it

Pro: they often do great update and upgrade as reference #announcements

[dense vigil]

My card has been declined???

[maiden fractal]

Always something new

[maiden fractal]

It is a credit card ?

[dense vigil]

Mhm.

[maiden fractal]

Not a debit card that act like a credit card ?

[dense vigil]

Well. Natwest Rooster card.

[maiden fractal]

So it is a prepaid card ?

Don't think that will pass :/

[dense vigil]

Oh.

Oof, Could be a sign since It's rare I use chatgpt anyway

[maiden fractal]

You can still use 3.5 for free and bing for free

[dense vigil]

I use google.

ik but 3.5 is sorta stupid

[brisk path]

Bard?

[dense vigil]

no, Search

[granite arch]

How so?

[dense vigil]

idk

[brisk path]

3.5 is a bit 💀

So are most 3.5 level models

[hardy pewter]

What's the limit of messages on gpt4?

[maiden fractal]

Chatgpt 4 , 40 prompt every 3h

[worthy crow]

if i buy chatgpt+ on mobile can i then also use it on pc? because i'd like to pay with paypal and that is only possible on mobile

[hushed sparrow]

yes

Anyone having issues with ChatGPT Web(Desktop) takes forever to load. Could it be my long chat history(including image generations)? need answers from verified OpenAi member(Mods?)

[ancient oak]

Is GPT4 working for you all? I'm being shown 404 pages and if I get the site to work it's not allowing me to send a query. It keeps telling me to contact support if it keeps happening.

[hushed sparrow]

No it has not

[ancient oak]

Haha. Yeah, I am. It's broken entirely. I've moved across two browsers and logged out/in everywhere.

[hushed sparrow]

the app works semi-fine(can't edit prompts or upload stuff after a few prompts for some weird reasosn)

I've noticed it more with my CustomGPT, the quality of it's response/generation has depreciated since New Years Day(Jan. 01 2024)

[ancient oak]

Yeah. I'm like totally down here. Nothing is even processing; I'm using Custom Assistant, too.

Damn, this is so frustrating. I pay for a service that literally breaks 3-5x a week.

[hushed sparrow]

The apps or using it mobile is a HUGE disappointment, just slightly annoyingly missing useful features.

[ancient oak]

Any mods or employees give us some sort of update, idea, etc? I've used ChatGPT for over a year and it's basically gotten to the point where it's unusable. I'm not really sure what to do.

[hushed sparrow]

The have a search generative ai to help find answers faster. labsdotgoogledotcom

[ancient oak]

The custom GPTs are working now. I'm not sure if it will last, but my Assistants are all working.

[dusk hatch]

how come recently when I give cgpt a link to a website and ask it to explain the pag to me it just gives a breif generatic summary and just says For more information, visit their website. when before it would actually give me a nice detailled report

[solar oak]

my gpt is not working

after I verify my website

[hardy pewter]

What about dalle

[maiden fractal]

same inside chatgpt

[still gorge]

What is the best app that allows you to use ChatGPT over a codebase on GitHub?

[topaz panther]

I'm looking at Jan dot ai - not recommending as I haven't looked closely at it yet.

[flat harbor]

Could anyone explain to me why failures on openai's end as far as "something went wrong" messages are counted against my prompt cap? I don't understand how anyone could think that's reasonable

I'm understanding of system issues, things don't always work right, and I'm patient in regards to that. but I shouldn't be charged for mistakes on their end

[dense vigil]

Anyone know if gpt4 would be able to help me with fivem Lua?

[zealous sphinx]

Is browser-based ChatGPT down? I am just getting the same message for all prompts:

Hmm...something seems to have gone wrong.

[honest jay]

Do you have a VPN installed? What browser(s) is it happening in?

[brave aspen]

hey there, I just get a lot of error searching knowledge, do you have any info please?

[chrome sable]

hey, if i have chatgpt 4 read two documents before giving me an answer, does that count as two prompts in one message towards the limit?

[honest jay]

Hi, you might want to make a post over in #1070006915414900886, and maybe check out help.openai.com! It could just be a temporary bug.

[lyric void]

Is it saying it is unable to access the contents of a file you uploaded?

[tacit stream]

anyone has errors at the moment using chatGPT (directly, not API)? I keep getting the "Something went wrong" error

[boreal lily]

chatgpt down

[slate berry]

ChatGPT is working. I just checked

[crude dune]

You need chatgpt plus to use dall e 3 right? IS IT STILL not available? It puts me on waitlist for too long now

[hardy notch]

Use stable diffusion localy if you have up to date GPU

Guys, how do you use it for coding? It seems to lose all the context so quick, for example I tried to make full stack app with nodejs, express, postgres, htmx, started slow, asked to create sql queries, then create migration scripts from those, then asked to make a simple login page, then logout link once logged in, each small and granular steps, adding few lines of code and once I have working login page and all, its now completely out of context, forgets what libraries it used, forgets that it used typescript in the beginning... And its a custom GPT that has exact instructions on which stack to use and it ignores it. Am I asking too much or is this just a skill issue? If skill issue - anyone know any good guides?

[hardy notch]

Yeah, I just want it to write me enough boilerplate, small full stack app, with very basic crud and auth, but either my prompts are bad or GPT4 is not there yet. Im dont expect it to create full blown sas or anything, Im just tired of writing boilerplate 😄

[young crater]

does anyone know how can i rename the fine tunes i created?

[junior anvil]

man it's so annoying when gpt 4 tells you: "After a quick search, here's what I found."
I specifically ask, do an extensive search. Lazy model / cheap

[tight island]

i mean you could do the research yourself, searching also costs tokens.

[junior anvil]

do it myself? then why do we have ChatGPT membership?

I pay them for this

it was much better, they just downgrade away the features. like giving you a great desert and when you get hooked, start giving you a candy

same

#openai-chatter message
#community-help message

all these people in recent days

reddit com /r/LocalLLaMA/comments/18ycoqw/comment/kgavkde

see also this comment

[undone smelt]

今天

[solar oak]

Error Saving GPT 😦

"GPT inaccessible or not found

[spice mauve]

I am keep getting You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after 2:47 PM. Learn more how can I have full time access to the chat?

[naive condor]

Hi

[maiden fractal]

40 prompt every 3h

[eternal rock]

This is a Safe for Work Discord Server

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.

Posting NSFW content will result in a permanent ban.

[spice mauve]

Is there any way I can get more than that?

[dense halo]

Something is making ChatGPT very slow lately. This is probably old news by now for you all? I don't come here very often. I wanted to know if the reason is known (by us, the public. Of course OpenAI knows), and what is it.

[maiden fractal]

Else you have the api .

[spice mauve]

what do you mean?

[maiden fractal]

https://openai.com/blog/openai-api