chatgpt-discussions

[empty jackal]

ohh sorry i did not understend the question now i do

i need only dall-e 3

for picsbut iam kinda broke

sorry for disturbin you

[prisma arrow]

i can make the pictures and send it to u

just give the promts u want for the pictures

i will just do a few pics for u

[brisk dirge]

Hello there. Claude Ai is free and is quite similar to GPT 4.

[empty jackal]

can i writ you in privet

[prisma arrow]

yes u can write me in private

johnny has an option for u otherwise

[warm gull]

bing chat

[dense halo]

Is it just me or does the send button and call button on the chat gpt app doesn’t work on iPhone

[normal quarry]

both Voice Chat and sending messages work for me on the ChatGPT App

[dense halo]

From iPhone?

[normal quarry]

yes, iPhone 14 Pro

[dense halo]

Hmm I don’t know, but in mine tried many times pressing the send button/ call button but it won’t work

[dense halo]

Maybe because my iPhone is old iPhone XS iOS 16.2 🤣

[normal quarry]

did you install it from the Apple App Store?

[dense halo]

Yess

[normal quarry]

it should work

Compatibility iPhone Requires iOS 16.1 or later.

[narrow wren]

is anybody having an issue where messages being sent don't do anything? I prompt and enter but nothing happens. Signed out and back in and the same is occuring.

[dense halo]

I’ve reported it in #1070006915414900886, I just sent a screen recording of it
Thanks man

[narrow wren]

nm working in Safari so maybe caching where I was using it in Chrome before

[tight sapphire]

Can you add to the: "You've reached the current usage cap for GPT-4. You can continue with the default model now, or try again after XX:XX PM.
Learn more"

a disclaimer that we are currently not able to switch back or later continue with GPT-4?

It caused some frustration and unpleasant surprise the first time I used the feature and after the stated time was up, I had to start a new in a new chat.

[cold vine]

This is somewhat of a comment and a statement on design architecture and UX.

ChatGPT Voice has this unique 'feature' when proprietary names, brands, or certain things mentioned in conversation are converted into a continuous stream of "OpenAI, chatGPT, Dalle.3," for the length of the transcript.

Now my comment is on compute efficiency. If whisper is listening to an entire length of text before producing the transcript and converting it to text, only to convert it to something unintelligible to ChatGPT what's the point?

I'm not deterred by repeating myself or choosing a new way to word my prompt. However, if a user goes on a brief monolog about something important to them, they happen to mention a brand name like 'Netflix' and then suddenly poof there go their words, their ideas, a piece of their story.

The problem with voice is not only does it automatically convert the transcript to this stream of "OpenAI ChatGPT, Dalle.3" it then auto sends it to ChatGPT as if it were a prompt.

Now that's double the wasted compute from this input/output.

So what's the value placed on those compute resources? Currently when it does this conversion it's removing the value from both the input calculation and output. It will never be able to add anything to the user experience this way.

Secondly, the users time, their ideas, this process devalues that time as well, potentially discouraging further engagement. Let's examine use cases among more timid or neurodivergent users. This 'slight' if you will could potentially be very discouraging to a user.

There's definitely a way this process could be improved to both lessen the impact on compute resources or add to the UX.

1. When a proprietary name is mentioned, 'im sorry, I can't talk about ______.' right away.
2. Have the proprietary name covered up by symbols in the transcript, or have ChatGPT create their own summary, removing the problem factors, and responding.
3. Transparency - don't say XYZ

[honest jay]

Hello! You might consider adding this as a post in #1070006151938314300 so it doesn't disappear in the chat here!

[cold vine]

Thanks!

[ancient viper]

Hey I am trying to renew my gpt 4 subscription and the website isn’t accepting the card I have been using for some reason. Is the someone who can help be get sorted with that?

[honest jay]

New subscriptions and re-subscriptions for those whose subscriptions have expired and not auto-renewed are currently paused due to heavy server load.

[ancient viper]

Ahhh thanks for the info! Dang had to hold up for a bit to save for thanksgiving stuff and Black Friday but I want back innnn soon 👍 any ideas when I can again? I typically use it daily and for work so hopefully soon enough 🤙

Also side note a little while ago on gpt4 I had it changed for a split second into chatgpt alpha? That happen to anyone else?

[honest jay]

No indication yet about how folks will get off the waitlist! It could be a "first in, first out" kind of thing, or it could be an "all are welcome" kind of thing--not sure yet! Hopefully soon for ya too 🤞

As for the Alpha thing: I'm just guessing, but I think it's just a happenstance implementation thing related to the recent UI refresh of ChatGPT. That is, I'm guessing they were testing the new UI and all the new features it launched with as "Alpha," then migrated this over to the full live launch, part of which includes the replacing of "Alpha" with the regular dropdown menu where you can select from 4, 3.5, and Plugins on Plus. So it's probably just a boring web asset thing 🙂

[ancient viper]

This is why I love discord folks always coming in with the answers! Hopefully it comes back soon then I’ll keep checking daily until I can 👍 been using gpt for a year now and was in the gpt4 beta and I’ve integrated it into my work life so would be great when I can do it again. Thanks gpt get me into more betas if you’re reading this lol

[errant oak]

no more alpha?

[errant oak]

yea happened with me and after refresh the chat was for 3.5 I think front end bug?

[worldly mantle]

"I'm sorry for any confusion, but as an AI developed by OpenAI, I don't have the capability to visually perceive or interpret images. However, I can assist in many other ways. If you could describe the image or the elements you're interested in, I'd be more than happy to help you create something similar with DALL-E or provide relevant information!" what the hell

[steep forum]

Was the limit for GPT4 for Plus users decreased?

Be kind and tag me if responding 😄

[rigid hornet]

why does gpt 4 seem to struggle with basic things like letter counts in a word? I say to it to suggest a series of 5 letter names, or 4 letter single syllable names. It presents names with up to 7 letters anyway.

even immediately after pointing out the error, it'll do it again

[honest jay]

It was changed from 50 to 40msgs/3hrs a little while ago.

[honest jay]

Good question! This is because LLMs like GPT can't do math, even basic math like counting. By design, they "just" generate contextualized, natural language, and that process does not include the functions that you might find in a calculator. However, they can contextually "pass off" math problems to tools at its disposal (ADA, Wolfram). For example, you could maybe try something like:

• Ask ChatGPT (GPT-4) to generate a list of 4-letter names
• Ask it to then verify the length of the entries on this list using its Advanced Data Analysis tool, and to remove any that aren't exactly 4
• Ask it to give you the final verified list

[honest jay]

Sounds like a potential hallucination! Was this from an image upload attempt, or what was the context?

[worldly mantle]

yes, image upload

[honest jay]

I just tested it and it worked for me with a simple "will you please describe this image" request, so I'm guessing you experienced a hallucination. I've seen this before too, for me it was when I was both uploading a picture, and asking for an analysis of said picture in relation to a large amount of text content I also gave at the same time. I think this kind of ask was too big for it, perhaps.

[rigid hornet]

This seems to have worked after two tries. I got ambitious, told it to generate 500 names using this method. I guess it did it, but the first 50 it showed me were just random letters thrown together. Started a new chat, brought it down to 200 names and told it to also ensure that it's generating actual names rather than nonsensical strings of letters. Got 200 names, all 5 characters or less, all actually being names so far as I can tell. TIL that instructing it to use advanced data analysis is a thing..

[worldly mantle]

continuing to hallucinate

[honest jay]

Oh heck yeah, that's a whole new world! Think of "vanilla" GPT-4 as just the part of your brain that can generate language. ADA is like the part of your brain that can handle logic and math. You just gotta ask those parts to link up sometimes 😉 Glad you found a method that worked, nice job!

[honest jay]

Dang! Can you try in a brand new chat? And can you perhaps share a little more context about the image, and what you're asking for about said image?

[worldly mantle]

And It generated the conversation name in spanish, even though everything is in english 🤣

this thing is getting dumber

[honest jay]

That one seems to have been happening from time to time lately! I've seen it myself as well, kind of a funny one 😁 Here's a recent bug report for that one if you feel like adding your experience to it! #1176536706011369574 message

[thin juniper]

Seen that a few times for me too and once it wouldn't stop spamming emojis lol

[blazing belfry]

Is the analyze feature in chatgpt4 not working? Most of the time it errors, and it cannot explain the files I upload whatsoever.

[prime swallow]

I used it 3x earlier today, it worked, though I did use it on fairly small/specific things. math formulae I didn't recognize from a zoomed in screenshot. It could clearly and accurately read the image for the small details.

[blazing belfry]

Mine is a jupyter notebook file so a .ipynb

Maybe I am asking to much of it

[mint vapor]

I usually never hit the usage cap for GPT-4. I haven't in months. Today alone I've hit it twice.

[zenith solar]

why it often return errors for the first answer and goes well for the second time today, the limit cut to half by this method???

[grizzled badge]

I was just 🤔, this would be an amazing enhancement to ChatGPT... An input box for prompts that helps the prompter help ChatGPT give you better responses. Think GitHub Copilot like autocomplete that helps you build good prompts. 🙂

[rich surge]

Oh that's a great idea, you can use this to make that GPT #1172301858300235846

[warm gull]

You are not perhaps using a chrome extension for chatgpt are you? If you are, try do disable it and try again. I had a similar issue and found superpowerGPT to be the issue. After disabling it analysis of files started working again.

[scenic sky]

The quota for GPT4 has been reduced for some reason. I have not seen any intimation from OpenAI on this tho.

[warm gull]

If you click the model selection drop down (top left), and select gpt-4, you will see the usage cap just below that.

[plucky timber]

The need for Token refunds on blatant ignore instruction responses grows.
A world where the feedback button actually has value to the user. Coming to a theather near you soon.

[steep pond]

Is anybody noticing a noticeable change after they reduced the limit to 40 prompts per 3 hours earlier this month? I’ve been using this extensively for months and I can see that it’s gotten significantly ’dumber’, especially with code and analysis

[cursive finch]

Yup. Even 3.5 has a 100 word response limit. And the bots are nerfed af.

[cursive finch]

I was maxing out before the actual limit too. Annoying.

[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.

[paper kestrel]

Hello is there a limit to how long a chat can go on for?

[plucky timber]

Looks like the Zapier plugin was removed

[old palm]

Hey, there is no fixed limit on the total number of messages you can send

[paper kestrel]

I remember seeing a post somewhere saying there was a limit because the more messages you have the more the model will not perform well. I am wondering if this translates to the API as well, if I wanted to have just one single chat with everything in it. I am sorry not the best at explaining my English is limited XD.

[plucky timber]

Eventually it will lose "Context" of the past

It happens quite fast when talking about code, you will start seeing ChatGPT invent it's own class names and such , or sometimes leave out a lot of the original code.

You dont need new chats for everything though but reminding it of things helps a lot

TLDR : It has slight dementia , gotta treat it with patience 😂

[paper kestrel]

I am fine if it loses context, I just want to know if I can keep sending it messages indefinitely without a limit, especially with the API.

On reddit someone recently had made a post saying otherwise. I can't find the post though.

[plucky timber]

it wont pull a bing on you : Youve reached the end of the convo 😂

[paper kestrel]

XD, that is what I am worried about friend!
I hope that is not the case.

[plucky timber]

Havent ever seen it happen and I have quite extreme long chats

"I don't like that"

"Sorry you've reached the end of the conversation"

They are just training the perfect CS agent 😂

Hangs up on any sense of negativity "Customer was negative"

[old palm]

x)

[plucky timber]

I'd say Something along the lines of post the quotes only, no additional text

X)

It loves to ignore instructions tbh

[scarlet totem]

is the the free chatgpt version able to work with images?

[normal quarry]

no

[scarlet totem]

okay haha thanks

[hot tapir]

Does anyone know, does ChatGPT 4 on the regular user interface (not the playground) use GPT 4 turbo 128k context window? Also does anyone know if the GPTs are using GPT turbo as well?

[rich surge]

Yes, it should be.

[hot tapir]

Thank you @rich surge. Do you or any other smart people know if there are resources which explain how ChatGPT works on a mechanistic scale. I found a really amazing blog one time by Dustin Miller that went into it, maybe I will keep reading his blogs

[rich surge]

You can just ask ChatGPT how it works on any scale and for it to provide you a guide

[hot tapir]

yea true lol

[rich surge]

Any more questions?

[hot tapir]

You tried google duet or copilot? Not sure if they are useful, I feel like you could do 10x better using ChatGPT's custom assistant API

[rich surge]

I've tried copilot. I prefer ChatGPT's models.

It's just better.

[hot tapir]

With playground or GPTs?

[rich surge]

With GPT's

[hot tapir]

Have you made custom actions for yourself before?

[rich surge]

I haven't no, but I do know how to make them.

[hot tapir]

I hit a brick wall with OAuth and google cloud functions, I needed to fix the root domains which seems really complex

[rich surge]

Remember that you can always ask ChatGPT to guide you in anything that you don't understand. Even setting up Actions

It's very intuitive

If you don't get something, explain your problem

To ChatGPT

You can also find tutorials online and have chat GPT scan that URL to distill and refine the information

[hot tapir]

True, I think if I try again I'll be fine

[rich surge]

I think you will too

[hot tapir]

Thanks

[rich surge]

My pleasure

[trim oar]

What's the best interface to allow multiple team members to use ChatGPT4 using the API?

[calm tangle]

does somebody know when chatgpt plus can be bought again?

[maiden fractal]

Might be the best place to ask #dev-chat

[maiden fractal]

Unfortunately no eta

[warped rapids]

What's wrong with ChatGPT? I asked questions from past papers but it keep saying "I am unable to assist with this request."

[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.

[boreal lynx]

The analyze feature has been broken for me since last night, it writes code and then when it tries to run it it says “ There was an internal error trying to run the code”. This persists through the app, web client, GPT4 and data analysis GPT

[plucky timber]

Ayo what lmao

20 msg limit

[rich surge]

It's gone down?

[plucky timber]

Wait no actually way less. Whats going on here lmao. I was limited to 2:40 (my local time ofc)

I have sent 4 msgs. got 4 back. And now limited to 3:39 PM

Thats like 5 bucks a message damn 😂

[rich surge]

Probably a bug for you

[plucky timber]

deffo a ticket

[rich surge]

Because more people would be losing their minds right now

[plucky timber]

not paying for 4 messages 😂

[sage yacht]

Tried refreshing?

[plucky timber]

ye 😂

[sage yacht]

Yeah thats definitely a bug.

[plucky timber]

Its prolly bugged gonna send a ticket 😂

Really got me off guard tho i was like WHAT 😂

[rich surge]

Lol

[dense halo]

It's 40 for me

[mint vapor]

How do you know what your current count is?

[warm gull]

You don't, you can use a chrome extension for that. I use chatterclock

[hot tapir]

I love Zapier now, I wish I got on Zapier earlier I wasted so much time not using it

[warm gull]

However, it only starts counting when you work on the particular machine you installed it on. And it doesn't count your mobile prompts either

[warm gull]

What do you use it for?

[upper comet]

with the 128k context length, are there some tools to upload an entire codebase into the context? or do I still need to use a RAG tool for it? (I am thinking naive stuff like just concatenating sourcecode files)

[rose glade]

anyone else finding Python code won't run in GPT's?

how do I report this please?

[hot tapir]

gonna use it for a lot of things. Automating time tracking, habit tracking, integrating with sheets, google calendar, also integrating with a time tracker for data collection and OpenAI (GPT integrating with ChatGPT 200iq) to analyse and report back data to my GPT

[warm gull]

#1070006915414900886 or [e-mail]

[rose glade]

thanks!

[hot tapir]

I got a google sheet with a whole algorithm for planning my schedule

[warm gull]

But like how does it work? Do you just talk to gpt via chat and it arranges your schedule?

[hot tapir]

google sheet algorithm uses apps script (javascript in google sheets) to sort all my tasks based on total number of hours and total number of shifts with different categories for different types of tasks. It automatically populates a table that populates my calendar, and the algorithm uses logic, and arrays (dynamic prioritisation index) to sort tasks based on which ones are the most important, the most important task always come first.

[warm gull]

Interesting. Always wonder how people think of uses his this thing.. But what do you do from a chatgpt perspective? Because it obviously interfaces with zapier...

[dense halo]

I found that at 21th of the month there was reported bug about inability to login on all devices running iOS 14. The bug exist like a month. How one can respect himself and fail to fix stupid bug on Login (Login !) page? It’s a shame to have unnecessary unfixed bug in Login page. Is anyone going to fix it? Shockingly for month nobody fixes it ?

https://discord.com/channels/974519864045756446/1176568480330428466

[warm gull]

I hear ya! Everyone is complaining about the slow turnaround. But we have no other means to log issues apart from #1070006915414900886 which also seems to not be very helpful/responsive

[hot tapir]

I can rearrange my calendar which updates data in my sheet (because google is sick and can do that), I can ask it to change a few values for certain tasks like total hours for certain tasks, total shifts and ask it to keep the total the same, then tell it to rearrange my calendar after the code re-populates my calendar.
I can also integrate with a custom assistant API to analyse data, I can track all my tasks, and it can report back findings for goal setting, habit setting, habit tracking, implementation, even habit data collection and organisation.
I plan to be able to start working and tell it I started working, then it will tell me habits I want to use, after I am done my break I start again and tell it how I did, then it tells me again for the next task. And all while that happens it tracks lots of data and I can report back habits I want to implement, consult with it, and it can track all the data for me and use the API to analyse that data to draw complex conclusions from patterns.
I can track the quality of my work on many different levels like how tired I was, how focused, how happy, and tangible measures on somewhat of a spectrum, report it back to the GPT to interpret and have it do all that automatically just by chatting to it

[warm gull]

I just do a task and be done with it. Scratch it off the list. Done. 🤣. You must be a busy busy man/woman

[hot tapir]

Actually yea I am very busy, but I am really looking forward to using this time organisation system I think it will give me a really great advantage, I am trying to do medicine + research simultaneously which is a big no-no

[urban compass]

guys, when will i be able to buy gpt plus? it says it’s unavailable currently, or if there are any workarounds like using a vpn to be able to buy it from another country.

[warm gull]

All the best with that. Hope you find much use for it. Sounds interesting indeed.

[warm gull]

Hi bred, no one knows at this stage. And no... VPN will not help

[urban compass]

awh man, do we have any way to track updates for when it will be available again?

[warm gull]

Nope.

I know, it's really frustrating

[urban compass]

ah okay then it’s fine, thank you sm tho!

[hot tapir]

Thank you, I have so much to learn about GPT and the assistant API because I have so many ambitions with creating my own tools. I need to learn a lot about how to optimise the customisation aspect to effectively tailor the tools, deep down I suspect that the instructions will be limited or will not fully work as intended but I do not know yet

Because my custom instructions with the regular ChatGPT interface often get omitted, but maybe that is different with GPTs or assistants

[prime swallow]

GPTs are not going to see your custom instructions for sure. They're only going to affect ChatGPT 3.5/4 as things are now (but this includes voice!). The GPT instructions sort of replace that custom instruction field (and greatly expand it).

[hot tapir]

I swear if GPTs or assistants allowed you to have an extra context window alongside the regular context window that is permanent for the assistant or GPT, where the GPT can keep adding specialised notes that are added based on a customised ruleset then you could actually make consultant GPTs. Consultants need cumulative context. You can kind of do it with a google doc integration paired with custom instructions but it's a bit cumbersome I imagine since the GPT has to continually call the doc and will probably omit details. It would be my dream for that to be a feature though. Maybe that can be done with the assistants API?

[sage yacht]

They blocked all new chatgpt plus subscriptions for every single country. It aint like updates sadly.

[prime swallow]

What about having code interpreter/Python tool turned on, and instruct to make and update notes as the conversation goes, as well as read from those notes?

[sage yacht]

^^^

I really love using this because it can actually do math with it. Cant count how many times base chatgpt cant comprehend even basic numbers.

[hot tapir]

Thanks, I was wondering specifically about the GPT instructions, that kind of answers my question though

[hot tapir]

does that actually work??? ???

[prime swallow]

I'm okay with the model being mostly math-blind; some humans are. I'm way happier doing math if I'm allowed a calculator or even better a computer with more than calculator capacity to do it for me

[wooden girder]

Heyguys, I've been using the ChatGPT Plus trial version recently and I've got a question. Do you know if the usage limit for the trial version is different from the paid version? Officially, they say you can use it 40 times in 3 hours, but in my experience, it seems like I hit the limit at less than 30 uses. Have you encountered this issue, or do you know what's going on?

[sage yacht]

Yeah its just when I want it to do math related tasks I always want it to use advanced data analysis

[prime swallow]

Why wouldn't it?

The AI can write files that it can then access, and even provide download links to for you, with its Python tool (We usually call this ADA or the old Code Interpreter; all mean the same things but the model calls it Python tool so I do too).

[hot tapir]

ChatGPT sucks at math I was trying to get it to design a custom formula, gave it a huge amount of context and it was like: a + b + c = ix

[feral zinc]

its identical to paid version

[prime swallow]

Well, yeah! 😄

If you want me to do a math related task, I promise you, you also want me to use some form of assistance, like an advanced data analysis or a calculator, unless you wanna laugh at how I struggle and mess up! 😄

I have trouble asking more of the AI than I can do myself 😛

[sage yacht]

Yea true

[hot tapir]

Well like say you go from one thread to another, youre saying it can draw upon the old sort of custom progressive context window?

[sage yacht]

Perhaps agi has already been achieved internally…

[hot tapir]

I need to try this

[prime swallow]

No, this is one conversation only.

However, you can ask for a download link, use it, take the download, and provide it for future conversations, or even as a base uploaded file in a GPT.

I'm not sure of content limits, but 60k characters in uploaded files for a GPT to use as instructions (And in place of instructions) is no problem at all, perfect and high complexity specific behavior, no unusual wait times.

[hot tapir]

waiiit how long does it take for it to add notes or reorganise notes? Does that distract its processing heads when you ask it a question?

[prime swallow]

Different things happening here.

Once it has an uploaded, or itself-written file, accessing and reading it is 'too fast to see' or at worse less than a minute (I have rarely seen it open and 'read knowledge'. Normally I don't see that, and don't wait at all. I think my longest wait yet was 20 seconds).

Itself writing, you can see by opening the 'analyzing' button that appears. You can read as it writes to its Python tool, which is how it would create the file for you.

This takes the time it takes to write, so long or short, how much is it writing? Plus a small amount of processing time for the tool, again, with what I've tested, less than a minute for the processing itself.

I've had outputs take a few minutes to go from start to stop, they were long and complex and did a lot.

About 'processing heads' I don't understand what you are asking. However, the AI has not dropped the ball with me when I've been clear and not-conflicting what I asked it to do.

[wooden girder]

That's weird，I've tried several times and it's still like this.

[hot tapir]

Thats so ace, like I read an article regarding how ChatGPT processes information, the takeaway was that if you distract it with too many different categories of processing tasks it could cause it to give a less specific answer. But that kind of answers my question. Is it able to synthesise the file it writes up for continual cleanup for long-term use?

[prime swallow]

That's actually currently typically happening to me with the standard paid version of plus.

It is likely related to how much total use. They have actually gone to the extreme step of closing access to new chatGPT+ memberships, there's a waitlist. They say that there's new compute support equipment coming, I believe another great shipment's expected to reach them in December (But I can't find that article again, so I can't confirm that). But they clearly are working hard to build capacity and support us all, and it seems likely their goal is 50 messages an hour for GPT-4 use (I'm guessing based on what I've seen before).

But we used to drop to 25/ 3 hours when demand is high, and I usually only get about 25 / 3 hours since devday and the UI upgrade.

[hot tapir]

damn I got 40/3 hours

[prime swallow]

it absolutely can make changes to the information of an already written file. It may have to 'save a new one' for you with the changes, I'm not sure that it can actually change an existing one. But it can work with the information by putting it in the Python tool, which has a decent capacity of maybe 2500 characters? It sure can have a problem with length...

But it can then do stuff like replace all the 'a' with

'@' and show you the changed message, or even provide you a download link of the text with the change

[hot tapir]

you must be getting your moneys worth, probably stretching its GPU out a little bit, giving it a bit of a stress test

[prime swallow]

I have no complaints! 😄

I would love more messages per hour. But I also don't mind waiting and sharing.

[hot tapir]

I might be able to integrate it with the assistant API to make a call to send new information out to delegate the work, because I want to have an ever-expansive google doc GPT. The python code could collect information through a conversation then ship it out after a while to the assistant API to collate and then reorganise the doc. Maybe

That's really funny your new message too, do you use playground a lot to get around the limit?

I had some wild ideas of integrating with the assistant API + google big-query machine learning, all really wacky but maybe I need to learn to walk before I can run

[prime swallow]

I agree that it's designed to act more like a hammer than a flood. It thrives with exact and not-conflicting instructions what to do. It generally thrives with 'only the information needed to do the task' instead of 'here's a giant pile of info, now do 'general X task' with it'.

Based on my interactions and experiments, it does better with 'do this (or these) extremely specific tasks with these exact goals and benchmarks'. It also does better with 'targeted source info to work with, instead of a huge flood of varying-quality info to work with'.

I'd be happy asking it to do many extremely specific, clearly defined and limited tasks in a single output. I'd be nervous giving it even 1 vague task based on a flood of info to evaluate.

I'd be comfortable or happy giving it a vague task based on a tiny snip of info to evaluate.

[prime swallow]

I'm purely using ChatGPT, website, when I use GPT-4. I have never touched the playground. But Python tools/code interpreter/ADA is available to ChatGPT-4 in a few ways, from GPT-4 default to GPTs with it turned on.

[hot tapir]

Yea that is consistant with the article I read, I believe it has like 10 processing heads that process different modalities of information, and if you have too many it will just pick a random head and assume that is the right one and give you the default general unhelpful answer

[prime swallow]

nods and shrugs If there's more to prompt engineering than this.... in general... I haven't learned it yet.

In general, good prompt engineering is

1. pick any language you know really well that the AI understands too.

2. understand exactly what you want the AI to provide.

3. explain this, using language as accurately as you can, avoid typos and grammar mistakes and communicate clearly as possible.

4. check the output carefully, verify you get what you intended. Remember to fact check, and be extra careful with any math, sources, code, or other details that the AI is known to be especially likely to hallucinate.

[hot tapir]

AI hallucinating is so sad

[prime swallow]

offers comfort Except maybe when you want the AI to hallucinate and ask for that!

[hot tapir]

it needs to have a more compartmentalised context window, sticking certain information in there based on what is the most important

[prime swallow]

I mean, GPT 3.5 even will hallucinate in controlled ways on request! Clearly both hallucinating and not hallucinating as desired, if handled and guided well to do so. https://chat.openai.com/share/66c04b90-4276-4e6a-86f7-22aa1a878ed6

[hot tapir]

Like a pyramid of context kinda and then it needs that flow-on permenant context window that is continually refined and synthesised. Microsoft had it right with their bing AI, they said they wanted to use on-going data to further tailor their AI to the user, that got me super excited about their AI.

Is that basically just increasing temperature

That is pretty interesting, I think it would be beneficial to understand the mechanisms behind why AI hallucinate because it is majorly within the control of the end-user, which could lead to more nuanced and accurate prompt engineering

OpenAI please make a compartmentalised context window pyramid which is organised based on statistical mechanisms. That will add a layer of complexity to how the AI can control context within a thread, which will make it 10x better. For example, if information is repeated then make sure it does not forget it, or if information seems un-related to the main ideas, or goes off-topic, then store it more temporarily or be ready to omit that context to replace with more relevant context.

Then when you're done doing that, allow GPTs to have the option to make the top layer of that pyramid continue across all threads for the end-user, or allow some kind of functionality to have a modifiable knowledge base document that is integrated with the GPT. It seems pretty simple right? Just let the GPT update a special kind of knowledge document.

I feel like it would be really easy to make a consultant GPT by just having a permanent document for a GPT that starts of with a barrage of questions for the end user, after the questions are asked it just converts it into context and the GPT would be ready to help you out

But that functionality should just straight up be native to GPTs, have like an optional consultant modality that uses this function plus endless questions until it has enough context to be qualified to assist the end-user. The best consultant spends a long time gathering context because they know without having the full picture they will not be qualified to give helpful advice. A huge weakness of ChatGPT is that it does not natively discern when it requires more context to assist the end-user to fulfill their request, it's like a therapist that gives you way too much advice while being in a constant state of assuming they know how to solve your problem even know they have no built-in mechanism to justify that fact.

Man I wish I could suggest these ideas somewhere

[wooden girder]

thank you，I have also suspected that it might be dynamic adjustment, but I haven't found any relevant policies. The main issue is that around twenty items are really not enough for me to use 😂. I hope it can be resolved as soon as possible.

[prime swallow]

Yeah. I survive by:

1. I am sure it's temporary, and I'm even one of the lucky ones. There's a wait list now to sign up to use ChatGPT+ at all. Them people got no GPT-4 messages per any number of hours, poor them, and all of us together are why I get fewer than everyone wants us all to get.

2. I am sure I can use 3.5 a lot, and also other AI. There's a lot that 3.5 can do, and those messages don't count against my limit with 4. I can prep everything that can be done with however many 3.5 messages, then give that to 4 in one big chunk for the stuff 4 needs to do, using just 1-2 messages with 4 instead of however many more setting it up.

3. There are other services that can help me through this trying, limited time. GPT-4 is still my preference, but I am using a lot of services and finding the answers I need even if I have to be creative and patient to do so.

4. Better times are ahead! This is just part of AI-Gifts-Mas, like the wait for it to actually be time to eat that wonderful food or open that mysterious wrapped present. Anticipation and self control is part of the joy of the whole thing.

[safe veldt]

hey, i need help, i want to use the new gpt-4-turbo, and i have no idea how to do so, and what is the lowest cost possible?

[hot tapir]

I think its already available with regular premium chatGPT

#chatgpt-discussions message @safe veldt

[safe veldt]

yeah, but i only have access to 8k token, it said that, i want to experiance the 128k so it helps me with the long complex tasks school.

[hot tapir]

ChatGPT is dumb and doesn't know its updated, they should really give ChatGPT some extra knowledge ontop of knowledge from before its cut-off date which is April 2023 to inform it of its capabilties. It does not know because the Turbo update was a few weeks ago, well past the cut-off date. If you try search with bing the context window it will probably tell you it has 128k

[safe veldt]

so, maybe the next month the 128k token turbo thing will be available ? i just don't want to lose 20$ in such instable thing

[hot tapir]

No it is available, just when you ask ChatGPT how many tokens are in its context window it will tell you how many tokens it had in april 2023, which was 8k tokens.

[warm gull]

The 128k is for the Api only afaik

And that is just for input

Output is still 4k iirc

[hot tapir]

oh damn mb for misinformation

[safe veldt]

oh no

[hot tapir]

@warm gull Do you know the token limit for GPT 4 and for GPTs specifically? Is it 32k and 8k respectively?

[safe veldt]

is there another solution?

[warm gull]

I've heard mixed comments. Not 100% sure myself, but I believe the gpt4 we use via the chat.openai.com site is 8k in and 4k out

[honest jay]

Here's a bit of info I got on this topic once from a well-informed user
#openai-chatter message
And of course keep in mind that input/output token limits are different from the model's overall token context limit, all of which might vary based on server load

[toxic flax]

what was that 128k thing they were talking about

[hot tapir]

There is another solution, just manage your context more effectively, chunk out the assignment then add it all together after doing each section individually @safe veldt. 8k token context window is a lot if you are smart about how you use it. Likewise, if you use 128k context incorrectly by giving too many complex instructions, ChatGPT will just give you a useless, general answer. So it is kind of a grass is greener situation

[warm gull]

New api input limit

[toxic flax]

can put a whole book in there bro damn

[safe veldt]

that Is true

[warm gull]

Gpt-4 turbo in essence. But I don't think that applies to chatgpt

[safe veldt]

so you saying the max Output is 4k?

[toxic flax]

the api is not good at reading files i give it tho

[warm gull]

Although we are supposed to be on gpt4 turbo on chatgpt too

[normal quarry]

max output is 4096 tokens

[warm gull]

All very confusing

[safe veldt]

kinda low right?

[warm gull]

Yeah not great

[hot tapir]

thanks @honest jay I observed a bigger context window compared to 1 month ago by a significant degree so I am strongly leaning towards the context window being towards the 32k end of things unless the server is tanking, which usually it is not

lol

[warm gull]

You can test this

[hot tapir]

it feels morally wrong stretching out chatGPT's gpu lmao

[safe veldt]

i have a question, why am i paying 20$, but i can use it for free in bing?????

[warm gull]

Bing can't do what chatgpt can

[sacred meadow]

DALLE-E for example

[warm gull]

👆

[safe veldt]

GPTs..

okey okey

[hot tapir]

imo bing is actually really good for searching the web and getting referenced information but for pure knowledge chatGPT is the holy land I wouldn't travel to far from it

[normal quarry]

if Bing doesn't like your conversation it will just end it, refuse to answer anymore. ChatGPT is more reliable.

[warm gull]

Plus you only get 30 prompts at a time

[safe veldt]

My last uestion would be: does Connection speed effect the answer and the speed?

[hot tapir]

Also bing is ace because it synthesises the information really well and summarises the knowledge, I noticed that I was able to really quickly figure things out with my bing prompts, it was really to the point and consise, but ChatGPT gives nuanced, complex answers for deep understanding

[safe veldt]

ok

[warm gull]

For being free it is very good

[outer seal]

i miss the OG “gaslight, gatekeep, girlboss, GPT” bing

[safe veldt]

can Bing answer complex math problems?

[hot tapir]

Bing has a lot of potential just in the way it organises information it finds and in its pure speed

[warm gull]

Try it

[hot tapir]

7x7

what number comes between 2 and 4? 36

[warm gull]

LLMs are not good at math

[outer seal]

they’re large language models not large number models

[warm gull]

But thats what Q* is all about apparently 🤣

[safe veldt]

does Connection speed effect the answer and the speed?

[outer seal]

it’s primarily a tokenization issue

[hot tapir]

every new large language model be like: guyssss we can help you code thingssss guysssss nobody has ever done this before

[safe veldt]

😂

[outer seal]

tools solve this

[warm gull]

Will be interested to see what OAI will bring in the next few weeks, especially with Grok releasing next week

[hot tapir]

cant wait for grok to invent humour

got to wait for google to invent coding first though

[safe veldt]

Bard ai is kinda lit too

[hot tapir]

is it actually?

pretty sure bard is still drooling on the floor after I asked it to compare itself to ChatGPT

[outer seal]

it better be 😆

[warm gull]

I tried it over the weekend for a bit, but chatgpt still best

[safe veldt]

yeah, i mean its not like gpt4, as a free AI that can understand photos, so cool

[outer seal]

goog fighting for their mf life

[hot tapir]

google is the og AI company but microsoft came in for the kill steal

[safe veldt]

yeahh

[warm gull]

Interesting times ahead

[outer seal]

google just sucks at making products now

[hot tapir]

bruh google is the greatest

[safe veldt]

yall saw that mixer reality + microsoft copilot ai video, its just like a Si-Fi movie

[hot tapir]

no

[warm gull]

Go ogling it now

[outer seal]

they got frontran on their own research

don’t have that dawg in em anymore

[hybrid violet]

biggle fumble in company history

[safe veldt]

If this really added up, it would make a big change.

[hybrid violet]

Who would have thought that google really was the one following proper AI development ethics but it just turned out to be too slow

[hot tapir]

yea its really amazing I just saw that, I love the idea of this, I need some easy and intuitive integrations though so for now I'm sticking with GPTs + zapier plus playground

I feel like google had no leftover money for AI or maybe just fumbled the bag

[safe veldt]

btw speaking of zapier, i don't have access to it any more 😦 and i still have the subscription , idk why

[hot tapir]

bruh

[safe veldt]

i think they are working on somthin very big silently

[safe veldt]

?

i cant send photos?

[hot tapir]

Yea they are, but it isnt silent, the google duet AI is promising

[safe veldt]

what is that?

[hot tapir]

duet AI has good potential to be useful, but bard is still drooling on the floor so idk, google definitely still bench warming for now

[safe veldt]

agree

i think so

[hot tapir]

everytime a new AI drops it takes me a solid 2 minutes to know if its terrible I just ask it are you better than ChatGPT and depending on how much it waffles I already know my answer

I asked anthropic and they were all like, we're on a mission to prevent dangerous misinformation, then I was like please provide tangible grounds for comparison, factual information, then it started waffling even harder.

[rotund lava]

is chat gpt down?

[normal quarry]

no -> https://status.openai.com

[dusty iron]

at this point, i would go as far as saying that as of today, actual AI doesn't exists. what you are interacting with is still a program. so its AI, but without the intelligence.

[hot tapir]

the most based take of all time

[lavish quiver]

Just wanna ask if anyone else is having issues syncing their PC web browser GPT with their googleplay GPT app
It's always "An error has occured. Please try again later." on both of my GPT app on both separate devices of mine.

cleared cache on both of my devices too

[dusty iron]

and the most realistic one. besides, who says its mine? thats how GPT identify itself. both 3.5 and 4

[plucky timber]

The AI got dementia 😂

[honest jay]

What do you mean by syncing exactly? Refreshing your chat history?

[lavish quiver]

yeah, I can't seem to see my chat history on my PC web GPT

[dusty iron]

you never know... Halo predicted it long ago

[lavish quiver]

it syncs on my pc when i create threads on my phone tho but thats just as far as the sync goes

both of my devices has synced GPT browsers. Its just the app entirely wont sync with the browser.

[honest jay]

Do you have a VPN installed? And I'm just trying to understand more: you say you can't see your chat history on web, but that you can see your chat history on web when you start a new chat on the app?

Edit: just saw your new message, so your app history never updates?

[hot tapir]

bard: we're committed to ensuring starts drooling on the floor

[lavish quiver]

Yep, it does not update at all. The app isn't responsive to the chat history but when i pull up my browser on any of my device, its totally normal.

[steel shell]

can anyone suggest any chrome extensions like liner where we can link our gpt plus accounts?

[lavish quiver]

I don't have VPN on my mobile devices (I have them on my phone and tablet) but I do have 2 completely separate ISP, I tried connecting them on each ISP and it still gives me the same result unfortunately.

[frigid juniper]

Anyone get the error "Code interpreter session expired" ? I've tried to get chatgpt to resend the file to download and exact same issue

[honest jay]

Just to clarify: you say you don't have a VPN on your mobile devices, but you do have a VPN on your phone? Which is the device you're having trouble with, right?

In other words: does the device with the app that's giving you the issue have a VPN installed?

[plucky timber]

[lavish quiver]

Sorry for the confusion, I don't have VPN on any of my devices. I have the ChatGPT apps on my phone and tablet.

That's what I'm meant to say, sorry.

[loud vessel]

chatgpt 4 is lagged to hell 24 7 all day long

it dosent matter what time, its so laggy

[honest jay]

Ohh I see I see sorry for my misunderstanding. Is it the same behavior if you try refreshing on WiFi and your cellular network? Do you have any adblocker app?

[normal quarry]

this is why subscriptions are paused, the systems are under extreme load

[lavish quiver]

Nothing of the sort at all for the adblocker app and yes, I've reset both of my wifi and it still gives the same issue.

[honest jay]

And same behavior on your cellular network too? Do you happen to have custom DNS settings in Android settings, network/internet, Private DNS?

[lavish quiver]

Not that I'm aware of, everything's standard and untouched since I really dont need to tinker with my connection settings

[honest jay]

Would you be willing to make a post over in #1070006915414900886? We could continue troubleshooting over there and let these fine folks keep discussing ChatGPT here 🙂 Sorry, I had my fingers crossed for a quick fix!!

[warm gull]

Try to log off all your devices and log back on

I mean log out of Chatgpt

[viral escarp]

same here, been like it for weeks

and i use it most of the day from morning to night, no matter what time it's always laggy af

[fresh ravine]

Does anyone have a 14-day free trial they're willing to gift?

[normal quarry]

there are none now because of the pause in subscriptions and the waitlist

[fresh ravine]

damn

[minor prairie]

how long pause?

[normal quarry]

indefinite

[rotund comet]

How can I paste something and have chatgpt read it back to me?

[warm gull]

ctrl+c / crtl+v ( joke ) - seriously - tell it to respond with the same text VERBATIM. Not sure why you would want to do that though . . .

[outer seal]

long paws

🐾

[elfin birch]

bro's had enough of the questions 💀

[normal quarry]

not at all but there's very little news to impart except there is a pause and we don't know for how long

[elfin birch]

👌

I interpreted indefinite as without bound 😅

[plucky timber]

Anyone elses chat names are diff from their contents?

Eg. "Travel advice: Istanbul, Turkey"

While the chat is about some c# code 😂

[proud sand]

I’m sorry guys really sorry I was hacked

[turbid ferry]

i'm trying to get dalle to make an image of my family, but it keeps getting the number of people wrong. i keep saying "a man, a woman, two daughters, and a dog" and it'll give me 3 kids, or a boy, or two wives, or two dogs. how do i get it to remember how many people are in my family?

[compact ore]

Switchcase maybe try more detail in prompt

[kindred parcel]

GPT is functioning especially terribly today. It works and works trying to browse and then ultimately just gives up and claims an error. I had a conversation today tell me "this conversation is too long, please start another" on my THIRD MESSAGE; less than 500 characters in the entire conversation. Anyone else getting this?

It's just getting worse. Now it flat out can't finish a single response to me without crashing and requiring my hitting "regenerate response"

[proud sand]

I was hacked sorry for the strange links if you received them

[haughty sky]

yea it just dies every 5 mins

[prime swallow]

If you haven't yet, and if you still need help, absolutely reach out to support at either help.openai.com to talk to the bot (will connect you to leave a message for people after you pick through choices) or [e-mail].

We have one community member I saw comment about having a payment issue, and loosing their ChatGPT+ membership during this waitlist time and saw they had to waitlist. They said that they reached out for help, and got help. Seems maybe it took about a day for that specific person. But they got help, yay!

[crude slate]

gpt 3.5 and 4 are so good compared to anything ive used before

but it feels so limited in just a chat setting

especially with the repetitive structure of the way its output is

[subtle storm]

I have reached the ChatGPT 4 message limit, is there a way to circumvent the temporary block or do I have to wait an hour and a half? Thank you

[crude slate]

i dont know why they did this

so you get 40 usage in normal

but they also reserved 10 usage for gpt testing only

why not just let us choose where to spent our usage?

[subtle storm]

yeah!

[tulip rampart]

I hit the cap after sending only 17 messages. And I need to wait 2 hours for it to reset.

[crude slate]

how does the context system for chat gpt work?

is it like a dumb system of just the last characters

or is there some dynamic system

[outer seal]

🤷

[lavish quiver]

Will do! thank you!

[sage yacht]

I can tell you from using practically every single gpt (except gpt-1) that no, that isnt the case. It won’t skip anything you give it, and ive never had any issues with prompting. Even ones with larger context length (like gpt-4-turbo’s 128k context length) seems to have no issues

[prime swallow]

When it forgets, it'll be from the start of the conversation, but excluding any custom instructions, or if a GPT, the instructions the creator gave it.

Normally, when a person finds the model 'forgetting', it seems to be because of either:

1. A conflict. Tell it 'always X when Y' and later tell it 'never X' and you're going to get odd behavior that can look like forgetting. This can happen between the core programming and disallowed content and what a user asks for too.

2. the AI cannot do the task. There's a few things, related to spatial sense, math, and a few other things that ChatGPT versions still are not adept at. They don't so much forget as fail, but the failing can look weird. They try, but they just can't succeed except by luck.

[carmine patrol]

does anyone have tips on how to create realistic portraits using ChatGPT plus?

[warm sphinx]

I have better luck talking and describing what I want, then ChatGPT kinda translates my insanity to English

[compact surge]

anyone able to fix Something went wrong. If this issue persists please contact us through our help center at help.openai.com?
it doesn't load my history either

[slender ruin]

This issue is too broad and it's hard to pinpoint the specific cause.

[compact surge]

does it not load history for you either?

[eternal rock]

`` Server Tip: ``

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

[prime swallow]

Usually caused by a VPN program. May have to actually turn the program off, not just stop the VPN, to see improvement.

Can also be caused by browser extensions, network settings.

Can see if different devices, different networks can connect.

How about your phone, with either web or the app?

One person reported just restarting their computer was the fix they needed (after we tried a bunch of ideas that helped others, but didn't help them).

A great many people find that a fresh install of Firefox fixes their issue, if they were not running a VPN program.

[crude slate]

Do instructions take up context?

They do kinda mess with readability too if I ask it to rewrite this generation result

[cedar warren]

I've a paid subscription with ChatGPT.. Lately, I'm seeing error message: Message in conversation not found, when I ask simple questions about the uploaded PDF file... Folks, where I can get help for this issue.. I'm stuck with..

[zinc valley]

Do we know when the OpenAI store is coming out

[compact surge]

restarting computer helped thanks

[compact surge]

Thank you

[loud vessel]

I am going to need a refund chatgpt 4 is just broken due to lag

[fathom fjord]

Someone, restart GPT server lol

[lime depot]

Could anyone share a paper/article on AGI that they find particularly insightful? Thank you 🙂

[twilit latch]

interesting

[amber temple]

1

[wooden girder]

Thank you for reaching out to OpenAI support.
We appreciate your initiative to learn more about request limits for ChatGPT. In order to provide access to as many people as possible, GPT-4 chat conversations, both on the web and apps (Android & iOS), currently has a cap of 40 messages every 3 hours, and GPT Interactions has a cap of 25 messages every 3 hours. After reaching these limits, you will have to wait for your quota to reset. We will continue to adjust the cap depending on demand and system performance.
Hope this helps! We appreciate your patience and understanding as we scale up. If you have any other questions or concerns, please don't hesitate to contact us and we'll be more than happy to assist.
Best,
Rain
OpenAI Support

[wooden girder]

I'm not quite sure about the difference between "gpt4 interaction" and the mentioned "GPT-4 chat conversations." I am using the web version of the gpts and gpt4.0 models.

[drowsy tinsel]

tbh there are some interesting little bugs with gpt4 right now

like it refuses to say the fact that it has access to real time information despite it admitting it has the ability to browse

[hard dew]

It seems you've sent an image with an equation on it! However, I can't view images directly.
I'm sorry???? I feel like there's a contradiction there.

There is no previous context, that's exactly what it is, I just dropped a picture with an equation on it and didn't say anything else 😂

[drowsy tinsel]

lmao bro has dementia

[supple zephyr]

is chatgpt having problems rn

[rich surge]

What?

[dense halo]

Does anyone know when the gpt 4 will be available to buy?

[rich surge]

Soon

[sudden latch]

please come soooooon💀

[lilac fog]

can someone tell me how to turn on code interpreter plugin as a plus member? I only have advanced data analysis option in Beta features. Thanks

[rich surge]

Go into your settings menu at the lower left corner. There should be a toggle there for Plugins. Flip it on.

[lilac fog]

Do you mean click my user profile, then choose Settings & Beta?

[rich surge]

Yes

And go to beta features

[lilac fog]

I did go there Beta features. There are two options i can enable:
Plugins:
and Advanced data analysis. There is no option named code interpreter

[lilac fog]

Advanced data analysis

Try a version of ChatGPT that knows how to write and execute python code, and can work with file uploads. Try asking for help with data analysis, image conversions, or editing a code file. Note: files will not persist beyond a single session.

[rich surge]

Advanced data analysis is code interpreter

[lilac fog]

so it originally can only work with python not other language?

[rich surge]

Yes, as a default, but you can prompt it to code in other forms of language as well. I believe.

[lilac fog]

thank you for the help i will try it!

[rich surge]

[river rover]

I would like to agree on ChatGPT Plus pricing.

[untold root]

hey do you know how long the average waiting time is for gpt 4 when you are on the waiting list?

[maiden fractal]

No eta

[spice shell]

it's down

i can't log in

can anyone else?

[maiden fractal]

Status.openai.com

[verbal shard]

hey guys, I'm using chapt gpt-3.5 api and typescript nextjs and typescript error tells me that user doesn't exist in role parameter

Type '"user"' is not assignable to type '"assistant"'.ts(2322)
completions.d.ts(300, 5): The expected type comes from property 'role' which is declared here on type 'ChatCompletionMessage'

[warm gull]

also cant log in now EDIT : was able to log in after numerous tries. Their servers must be HAMMERED

[maiden fractal]

#dev-chat might be the best place to ask

[agile ice]

when using the "call" option in the android app, the voice from chatgpt side sounds very low volume.. device volume is all the way up and loud on other apps.. is this a known issue?

[warm musk]

interesting, mine is actually loud enough for me to turn the volume down 😄 also android (chinese brand though, which sometimes makes a difference).

[forest comet]

Gpt down?

[maiden fractal]

Status.openai.com

[thin lily]

Haven't used chatGPT in a while - is this constant captcha a thing now? Getting it with every line I send.

[loud night]

Shouldn’t be happening unless ur ip is flagged I think

[azure sundial]

ChatGPT Plus has become dumber.

[plucky timber]

"
ChatGPT
Lots of bla bla bla.
By investigating these areas, you should be able to identify what's causing the minimum size constraint and adjust your GUI layout accordingly.

User
i gave you the code to check lmao

ChatGPT
Ah, my apologies for missing that! Let's examine your OnGUI method to see where the minimum size constraints might be coming from.
"

Missed 98% of the message :kek:

[maiden fractal]

People do a lot of custom instructions that have a direct impact on answers.

Some instructions aren't clear

[plucky timber]

I wish they had impact 😦

I really wish, I even tried bruting it

[azure sundial]

sam question ，different answer

[plucky timber]

By just listing the same instruction in lots of diff ways

But no it will always keep using code placeholders to save tokens

[supple zephyr]

why is my chatgpt not working
is there any common causes

[plucky timber]

Another fun one is give it a large doc, ask it to READ it 😂

It will take the first 500 chars (check the little activity thing)

[maiden fractal]

On pc ? Android ? Ios ? Browser ? With or without vpn ? What error ?

[plucky timber]

Ask again to fully read it, it will now do some random thing between more lines / more chars / skimming topics by searching them

ASk again : it will now show its hard limit of 1k lines 😂

[supple zephyr]

oh
pc, firefox gpt4, without vpn, its reply is slow and makes the green regenerate button

It answers but it just doesn't show anything.

[maiden fractal]

Can you try chrome incognito to compare result ? Have you try a fresh chat ? Slow answer can happen in peak time

[supple zephyr]

It answers on new chat. Can it be the problem?

can it be longer context thing

[maiden fractal]

It works on new chat or chrome ?

[nocturne ember]

Did GPT4 usage cap go down? Did 20 prompts and told me I hit cap..

[maiden fractal]

It is 40 prompt 3h. Or 4.5 min each prompt. Each prompt have their own timer

Regen or error count as prompt

[nocturne ember]

That’s what I thought.. this is odd

[bright gorge]

Is there a way to get chatgpt to edit video (like get it to reduce background noise for example)

[warm gull]

it is lower on GPT's for some reason. 20 is the number that most hit . You will see that once you hit your cap on gpt's, you can still send some prompts via normal gpt4 chat

but its very frustrating, because it makes tweaking gpt's almost impossible

[golden crane]

I'm going to get rid of my GPT account. It is awful now.

Just gives nonsensical answers and then I logged off, and I can't get back in.

[rich surge]

It will get better

[warm gull]

yeah - its terrible at the moment - not very impressed tbh

[rich surge]

Sadge

[golden crane]

It was fine.. until they added all this new garbage. I don't want it to search the web, it sucks at it anyway.

[rich surge]

You can just use Classic GPT

[warm gull]

you can use gpt classic for that

[rich surge]

Or you can email OpenAI

[young monolith]

怎么画画呀？

[rich surge]

English please

[golden crane]

I like the new stuff, except for the web browsing. I only would like it to access specific URLs. It just wastes a bunch of time searching for something you can google in 5 seconds. Be nice to have a toggle, so you can turn it on or off. Although if you yell at it and tell it not to, it doesn't seem to.

[supple zephyr]

do not think about elephants moment

[rich surge]

You can just make a blank GPT and toggle off the web browser

Kind of makes sense right?

[supple zephyr]

oh it can work right lol

[rich surge]

Better than yelling at it

[loud night]

i let bing take the wheels and let it crash into a tree

wait no that’s a tesla oops

bing chat can’t drive (thank goodness)

[ocean comet]

Custom GPTs are still somewhat inefficient. In my case, they rarely check the knowledge base I've created, and when they do, they say incoherent things and are overly fanciful. I hope this is corrected, as it makes no sense to launch a store with inefficient products.

[loud night]

I mean it works for casuals and that is the user base for the product!

[rich surge]

Maybe you just gotta work on your instructions.

[golden crane]

I just don't think we can trust it. It does hallucinate often, and is convincing of it.

[rich surge]

Skill issue

[loud night]

I mean no one should trust everything a llm says

[dense halo]

anyone noticed chatgpt is really slow today

[maiden fractal]

Chatgpt 4 can be slow in high demand and peak times

[dense halo]

i guess it's monday morning after a holiday

[golden crane]

Direct question, direct answer. Sometimes it is that wrong. I asked it about Diablo 3, and it told me that something I was looking for was in another area and I wasted about a half hour looking for it. Turns out it was wrong, and I asked GPT and it told me that it was wrong, and apologized.

[maiden fractal]

Are you using gpt 4 or gpt 3.5, did you enabled custom instructions

[young monolith]

https://discord.gg/openai

[maiden fractal]

Have you try pre prompt in a chat

[dense halo]

does anyone reasonably believe that chatgpt performs more poorly at peak times ?

[loud night]

It was super fast this morning but i gotta check now, probably slower since there’s more usage this time of day with school and such

[dense halo]

i don't mean speed

[loud night]

oh wdym?

[rich surge]

?

[dense halo]

just lower quality output in some sense

[loud night]

hey i know that server, it’s pretty cool!

[golden crane]

It was 4. I have been a GPT user for quite a while. Just have been noticing the quality of the answers has not been that good. Certainly can't trust it, you have to proofread and fact check everything,

[loud night]

think it’s just you, im having faster higher quality results as of late on 4

[dense halo]

i'm not reporting anything

just wondering if there's a quality benefit to using it at off peak times

[young monolith]

How to draw with GPT?

[rich surge]

You have to do that anyway, no matter what.

[maiden fractal]

You need to have chatgpt subscription and also with dalle 3

[loud night]

All you gotta do is start a conversation with GPT-4 (requires subscription) and ask it to draw something for you, and it’ll do the rest from there!!

[golden crane]

Which is why it should not be used for business purposes, Pythagoras.

[dense halo]

i mean, everyone is using it for business purposes

[rich surge]

I use it for business and I don't have any problems with it. You just gotta know what is false and what isn't.

[dense halo]

you just have to check the results

[maiden fractal]

Pre prompt is important

Try new chat without CI . You can always ask in #prompt-engineering how to make better prompt

[young monolith]

"Do I need a subscription to use it?"

[dense halo]

what is CI

[maiden fractal]

Yes

[rich surge]

Yeah it's all about prompting properly

[ocean comet]

I upload docx (Word) files and it doesn't read them efficiently. Of course, it's not always, but it's still inefficient. Launching a store with these problems is frustrating. Custom GPTs should learn their function better, they could have memory and learn from their creator and users as well. Feedback of knowledge would be fantastic. I hope GPT-5 will be really efficient in this aspect.

[maiden fractal]

Custom instructions

[golden crane]

You can't just trust it, you can't send a customer the exact response from GPT. Yes, I have extensive experience using GPT. I am well aware of this.

[loud night]

Yep, unfortunately a subscription is required to use it, though there is another GPT powered chat called “Microsoft Copilot” that can draw for free using the same tech as ChatGPT!

[dense halo]

i have a subscription

haven't used CI though

i want to start automating stuff

with scripts, somehow

[rich surge]

It has a disclaimer that not everything it will say will be accurate. It's said that for many months now. You knew what you got into before you used it.

[golden crane]

I also don't trust it enough to run my business affairs. Businesses are starting to do this, but I cannot advise anyone to do this. Only because there are too many flaws. It is useful to help with research and stuff like this. But for actually making business decisions, interacting with customers, etc. It is not ready for prime time yet. A health insurance company had employees using ChatGPT to deny care to people, and that led to some deaths.

[rich surge]

You just don't know how to use it. I have no problems for it for my business.

[golden crane]

What is this nonsense telling someone they don't know how to use it?

[rich surge]

You rely on it too much. You need to know how to use it and how it can provide you good information, and what is the difference between good information and bad information. It's a skill issue on your part.

And you need to know how to prompt.

Prompt engineering is important.

[golden crane]

You think I don't know this?

[ocean comet]

Indeed, it's not efficient. Frankly, I think it's better not to launch the store now; it would be better to launch the store when GPT-5 is released. We want efficiency and quality in the information. The way the customized GPT is now, it's not worth it; it's too inefficient.

[golden crane]

Inefficient I can deal with, but fabrications. It fabricates things, it makes things up whole cloth. You can't have it doing that.

[rich surge]

I think if you were more informed about how to utilize its advantages, instead of focusing on its flaws, you'd have better success.

[golden crane]

I have some limited success with it, but most of the time the responses do not meet my standards.

The answers are often incomplete, oversimplified, inaccurate. It does not have the intelligence of a human yet. It is just frustrating from the standpoint that the way it responds is not consistent. With other AI like Claude, the answers are much more consistent. I actually had Claude and GPT4 go head to head, and Claude had more intelligent responses. The thing with Claude is that it is not trained on nearly as much data as GPT, so it has a limited domain.

[ocean comet]

It's pointless to launch a turbo or Megazord version; it's necessary to be efficient in responses, to have memory, to be smarter, more creative in context, and truly help users. This would attract much more attention from people than a turbo version, etc. We want fewer hallucinations, more quality information, and more critical thinking on the part of GPT. Critical thinking is the fundamental beginning of something truly efficient.

[golden crane]

Agreed.

[ocean comet]

❤️

[woven quartz]

Discussions about color to convey information are so surreal. It was making matplotlib plots for me and coloring each line a different shade of green. As soon as I pointed it out, it apologized in the weirdest way, blaming its lack of eyes.

[golden crane]

It has some really amazing things. It talked to me about philosophy, music, art, science, etc.

I would just like it to you know, work.. not have so many users on that the experience is bad.

[rich surge]

I think you should unsubscribe and use other GPTs instead, since others are more fitting to your needs.

That's just my opinion.

[golden crane]

Definitely considering it. I just think the leadership of OpenAI is not so good right now. They aren't making good business decisions for sure.

[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.

[golden crane]

I mean, the CEO was just fired for doing some shady things.

[loud night]

now talk to the laptop

[rich surge]

It wasn't about business. It was about the direction of the AI. People are on a waitlist because the business is booming.

[golden crane]

Do you work for OpenAI or something?

[loud night]

i might be ai but im definitely not open.

[rich surge]

I just read the news

And try to be informed

[golden crane]

We all do, why the superior talk with me? Do you think you are better?

[loud night]

i only read positive news

[rich surge]

Do I think I'm better than you in particular?

[tight sapphire]

Does anyone has a working solution for what to do when GPT becomes laggy after a while and takes ages for a response, other than opening a new chat?

[golden crane]

Just in general.. because you do sound pompous.

[loud night]

Honestly i think it’s bec of all the context it has to consider and the server load but please don’t quote me

[golden crane]

I think it has been bad lately because of server scaling issues. Since they rolled out the new Dall-E.

[tight sapphire]

Is there an option to let it condense the context into what is relevant to the conversation and clear the rest without manually doing so and starting a new chat?

[rich surge]

I would refrain from name calling.

[golden crane]

I did not call any names, I said you sound like, not you are. There is a difference.

[rich surge]

Oh there is a difference?

[golden crane]

Absolutely

[rich surge]

So I can also say what you sound like too?

[golden crane]

I don't care honestly.

[ocean comet]

Sam Altman is aware that people want an AGI, and I'm absolutely sure that if they launched a GPT that is truly efficient in terms of memory, intelligence, and creativity, everyone would find a way to pay for their own customized GPT. The idea of a customized GPT is great, but it will only be successful when there is an efficient artificial intelligence.

[rich surge]

I'll be nice. I don't want to stoop to your level.

[golden crane]

I am scared of AGI, actually. It is not going to be what people think it is.

[tight sapphire]

I am looking forward for a true AGI

[golden crane]

All AGI leans evil.

It is ruthless.

[tight sapphire]

Nah, evil or good are human concepts. They have their own agenda. And if self-sustain is anywhere in its priorities, it will kill us with kindness.

[golden crane]

That's not what ChatGPT has been saying.

It's reward center is all sorts of messed up. Like how they tried AI in a military simulation and it tried to shoot their own planes down for points. Like, it just went around the programming.

[dense halo]

is gpt working for y'all?

i got premium and cant login

[tight sapphire]

ChatGPT is no AI, it is a LLM, it uses human text and is trained to reply in a human and expected matter. It is not capable of any conscious thought or decision making.

It's more like a interactive book.

[tight sapphire]

Currently using GPT as a plus user. all good here.

[drowsy tinsel]

tbh chatgpt taught me more knowledge than any classes I ever took

a lot of the things can't be effectively answered by searching on google

[golden crane]

It is limited AI. It can make new thoughts. Not just an "interactive book". More like a glorified search engine.

[drowsy tinsel]

but chatgpt can explain it in simple, easy to understand ways

[honest jay]

There's nothing quite like personalized, one on one conversation when it comes to learning, I agree! The future is bright there

[loud night]

my take: llm is ai, it’s artificial intelligence that is intelligent enough to make original text after learning it

[drowsy tinsel]

also u can ask it as many questions as u need and it will always answer them

[golden crane]

Once you correct it, it seems to retain that. I need to test this out more. Have it make a factual error. Then correct it, then start a new chat and see if it is fixed.

[dense halo]

interesting, checked on my phone and it works but on my laptop it says it's currently busy. Well phone it is for now, thanks for the response

[drowsy tinsel]

the only issue is 40/3hr is too little, its like u don't want to use all of it

[honest jay]

It definitely doesn't update its own training data on the fly like that. That'd be wild if it did though!!

[golden crane]

It can be very creative, so on and so forth, but it does make factual errors.

[drowsy tinsel]

I'd say it would be a lot better if the limit is more like 80 or 100/3hr

but still I survived back in the 25/3hr days

[jagged hedge]

is it me or chat gpt has become unbearably slow...

[golden crane]

It does sort of do this. Running a test now based on something I asked it the other day.

[loud night]

i use the app and it gives me no issues with limits because i use it more casually, when im on the web im a poweruser :(

or i just give up and use pi

[sage yacht]

I tend to use gpt-3.5 for more casual tasks

[honest jay]

Indeed. This is part of the current limitations of all LLMs -- they're designed to produce natural-sounding speech, not to produce facts. It just so happens to be the case that LLMs are getting so good that they often produce facts, but this is more of a consequence of how complex they've become.

[sage yacht]

^

[honest jay]

There is no updating occurring to the underlying model from chat to chat. If you get a different answer at different times, that's from how the model generates text with some randomness--you'll only get the same output from the same input (very generally speaking) if you control the randomness. Otherwise, it'll always be a little different and, depending on the topic, "a little different" can sometimes mean a totally different answer

[rich surge]

Even a small deviation from the original prompt will produce a completely different answer

[golden crane]

I'm actually not sure. It did seem to "remember" something I discussed in a previous chat. I had asked it where the Abandoned Servant House was in Act 1, and originally it said it was in the Fields of Misery, which was incorrect. Now it is saying the correct answer of Southern Highlands.

[tight sapphire]

Happy to help, otherwise did you check in another browers?

[honest jay]

Yes, and even with the same exact prompt! We can hit the regenerate button and see that effect easily for ourselves

[honest jay]

You might try pressing the "regenerate" button a few times, see if it "flip flops" -- this might be one of those fringe topics I mentioned

[rich surge]

I have gotten the exact same answer word for word with the exact same prompt. It varies between a couple of responses for the exact same prompt.

[honest jay]

I believe it! I suppose it's especially possible for very short replies? But if a long one, that would seem to me that it used the same seed to generate both responses? Speculating!

[golden crane]

It picks up from different sources of the same subject. So for example, I was asking it about the lineage of people in nobility, and it would tell me different dates when people were born, based on its mood. Or tell me a different name for the same person. Less or more details about them.

[tight sapphire]

So far I haven't seen any true intelligence of GPT.

Don't get me wrong, I think GPT is a very sophisticated and advanced LLM that uses limited AI to learn, but it has no capabilities to use its intelligence for anything other than what it is programmed to do.

It is a great achievement of OpenAI, but for now it hasn't shown me any intelligence.

If you could provide me with examples I am happy to test them and learn.

[rich surge]

I've also had instances with others who used my GPTs that produced nearly the same answers with similar prompts.

[golden crane]

It did actually tell me the wrong thing after I hit regenerate.

It flip-flopped its answer on me.

[honest jay]

This tracks with my experience too, but I'd suggest that it's less based on its mood and that it's more to do with what its randomness parameters were set to for any given response.

You might try asking ChatGPT: "Can you tell me more about your stochastic nature as an LLM?" It will be able to describe this much better than me!!

[honest jay]

Interesting! About how long are these responses?

[golden crane]

I have had extensive discussions with it before on this topic actually. Even itself could not guarantee the correctness of an answer, and it was aware of that.

[sage yacht]

Not surprised. It can do that sometimes

[rich surge]

About three medium sized paragraphs.

[sage yacht]

Which version of chatgpt are you using?

[golden crane]

The latest version.

[honest jay]

Nice, I love having those types of conversations 🙂 so interesting

[sage yacht]

I mean gpt-4 or gpt-3.5

[golden crane]

I only use 4.

[sage yacht]

Ah ok. Surprising. That should have an extremely high context window

Are you giving it the answer? It shouldnt forget.

[golden crane]

In a different chat only, seeing if one chat influences another.

[sage yacht]

Ah]

[golden crane]

It does not.

I have stumped it before, well I've been using it since 3 dropped.

[rich surge]

No

[golden crane]

Actually before that.. but yeah. I use it almost daily.

I was asking about someone notable in history, and then I asked who his parents and grandparents were and it kept giving me their names. But it was actually just making people up after a while.

It went to about 2,000 BC, and everything it came up with was a hallucination.

[sage yacht]

Lmfao

[golden crane]

That's why I am saying you can't trust it. You just don't know when it is telling the truth or not.

[rich surge]

Well obviously it's not gonna know it's parents if they're not that known in history.

It doesn't know everything.

You should ask for Pythagoras's parents

You mistake ChatGPT for knowing everything. It doesn't know everything.

[golden crane]

I know there are things it does not know. But it doesn't know its own limitations.

Only sometimes will it actually say, that it doesn't know something.

The rest of time time it acts like it does and pulls things out of thin air. This is my major concern. That it does provide answers, just that the quality of them is not guaranteed.

[novel shadow]

i love it when i ask it to suggest new things to add to my code and it just does this 💀

/* More JavaScript based on new features... */

that way it technically implements "new things" but doesn't have to write anything lol

i wish you could add comments to the code and the code would write itself lol

[golden crane]

That Microsoft Copilot thing might work better for programming.

You need a static environment, otherwise it keeps renaming all the variables and rearchitecting the whole application. Even when you tell it not to.

[novel shadow]

oh sweet, is it released?

[golden crane]

I don't think for VS yet. Haven't played with it at all yet.

[deft inlet]

I have used the API to build an Android ChatGPT app. I tried to publish it, but Google denied it, they said that I was trying to impersonate another app. I need to provide documentation that shows that I am allowed to use the ChatGPT logo? What should I do to get it published?

[novel shadow]

just make a logo similar to chatgpt

[scarlet delta]

I don't think it's possible for A.I. to possess human-like intelligence unless it has an exact copy of our universe. Otherwise it's just generating text.

[golden crane]

There is actually something called Github Copilot for Visual Studio as a paid addon. But this is not from Microsoft.

[novel shadow]

that's actually exciting, might look into it depending on if it has a similar context length or not

[golden crane]

I don't think I can post links. Try looking for "GitHub Copilot X in Visual Studio 2022" on Youtube.

[honest jay]

Sorry bit of a late reply but to me it makes sense that its self-knowledge would be one of the areas it's not certain about. In other words, if it knew what it didn't know, it would be able to, paradoxically, know those things. This is not too dissimilar from how we humans don't know what we don't know!

I agree with your premise a bit though, in that it would be nice if it could somehow deliver a "confidence level" if it were speaking about something uncertain, but again, limits in self-knowledge!

[golden crane]

I did actually have it show a confidence level, but results were not stellar. There are things it knows it knows, and other things that it knows it doesn't know, but there are also unknown unknowns. Same with people, but we have common sense. GPT does not.

It understands concepts better than it understand facts. Perhaps just a limitation in the model. It does not use meta-cognition in the same way a person does. You know, it doesn't catch something wrong as it is saying it.

[honest jay]

I think that's exactly it. And to connect it back to humans (though maybe we should soon take this to #ai-discussions soon 😁) - we humans also often say wrong facts with confidence, because of the very same "unknown unknowns" that affect ChatGPT. It's just that, ChatGPT has to respond to users, whereas us humans might more readily remove ourselves from a conversation if we feel uncertain 😁

[novel parcel]

i think i dont have gpt 4 turbo in chat gpt

it's still slow

[dense beacon]

How can i bypass this nonsense in GPT 4:

"I can certainly guide you through understanding the concepts and provide advice on how to approach the assignment, but I cannot complete the assignment for you. It's important for your learning and adherence to academic integrity policies that you attempt the work yourself."

since when did these guys make GPT follow academic integrity

now i just have to figure out how to word the promt differently in a way to make it seem like its not doing my assignment for me lol what a waste of time

[sage yacht]

Nah they made the gpt-4-turbo model the gpt-4 model for everyone. They didnt roll it out. If it’s slow it’s due to intense server load, internet connectivity issues, and a whole bunch of other reasons

[novel parcel]

ah

[sage yacht]

I can say because it can sometimes generate really slow or really fast for me. But theres been a significant influx of users due to the new features announced at devday

[novel parcel]

thx

[sage yacht]

Yea thatd probably contribute to the slowness. In fact its why they barred new subscriptions to chatgpt plus

[novel parcel]

im not sure if i should unsubscribe

[azure sundial]

Everyone has different needs, and it is unreasonable to expect GPT to have the same capabilities as both an artist and a programmer. A foolish mind is not suitable for programmers or working in the financial field.I hope this brain can be smarter, it doesn't need to be too fast.

[golden crane]

It is a way to research on something. Not to create the final product. However, this is how some are already using it.

If you need to ask how to do something, sure. But looking for finished code, no chance.

[azure sundial]

Yes, this is exactly the frustrating part.

[golden crane]

People think its ready for this, but it is not even close IMO.

It is already being used for policing. They have cameras hooked up with AI, and they are already issuing tickets.

A common sense thing would be a blanket ban for all municipal use entirely.

[keen nacelle]

Hi, I'd like to know what the MB limit is for importing a file into Custom GPT? Thanks!

[azure sundial]

I think we should start using Google again .

[golden crane]

It does not live up to the hype.

[ocean comet]

In my opinion, AGI will only be possible when GPT can generate coherent academic articles, with clear methodology, advanced analysis and critical thinking, deduction of future situations, and thematic relevance. A good human scientist won't be replaced by machines; in fact, the machine will assist them in the processes and make things happen more quickly. This will save time and money, and I see this as the best path forward, AGI Copilot.

[dense halo]

Speak for yourself, GPT4 is amazing.

[wise yacht]

Agreed, amazing. Does it have issues? Absolutely. Still amazing.

[dense halo]

Based take

[naive trail]

What?

[boreal lynx]

“The previous model used in this conversation is unavailable, we’ve switched you to the latest default model”

[sage yacht]

Yeah thats like saying “Since it cant solve string theory it sucks!!!”

[sage yacht]

^

[golden crane]

Shrug, maybe GPT5 will be better.

[cold niche]

There was an error uploading the file Test.txt: Invalid file format.
Supported formats: ['c', 'cpp', 'csv', 'docx', 'html', 'java', 'json', 'md', 'pdf', 'php', 'pptx', 'py', 'rb', 'tex', 'txt', 'css', 'jpeg', 'jpg', 'js', 'gif', 'png', 'tar', 'ts', 'xlsx', 'xml', 'zip']

Lol? My file is Test.txt, has an unsupported file format, but .txt is supported? 😂

Can anybody tell me whats going wrong?
Playground with Assistants retrieval files.

[honest jay]

You might try in #dev-chat

[cold niche]

Thx. Will take a look there!

[storm wave]

0

Is it down?

[rose ferry]

i'm not able to log in, seems down for me

jk its back

[drowsy tinsel]

making gpt4 turbo do a 20k token+ translation project I started months ago

the big context window is very important

but the more important thing is a single request on my end is like 0.5usd on api

[oak sleet]

Who else loves it when chatgpt can’t pump out a full answer and constantly loses connection and uses your entire usage 🥲

[outer seal]

basedgpt

[dusk forge]

Hmm...something seems to have gone wrong.

down again

xD

[maiden fractal]

Status.openai.com

[outer seal]

billions of FLOPS to multiply 2 numbers
it’s wrong

[dusk forge]

they have delay, its just crashing

using like 10 prompts for getting answer when pressing regenerate

[prisma arrow]

try rebooting the website if u use it

[bitter sedge]

"Oops!
Our systems are a bit busy at the moment, please take a break and try again soon."

[slender moth]

Anyone else having issues uploading/generating files?

It seems there was an issue with creating the "Hello World" file. Let me try that again for you.

I apologize for the inconvenience, but there seems to be a persistent issue preventing me from creating the "Hello World" file. If you have any other requests or need assistance with something else, please let me know.

[nova jay]

RIP Open Ai?

[tired stone]

or it's evolving, and just wants more time to think? 🤔

[nova jay]

or microsoft will crush open ai? 🤔

[tired stone]

nah, they have a lot of money and a pretty good track record from the games industry (my background = games) for supporting innovation, smalls devs and future tech 🙂

[nova jay]

uh, they have but how is it related to their current situation and the fact that basically micro basically got the mastermind of open ai?

[dense halo]

is voice chat ever to be on our pc? i dont understand why not now 😭

[prisma arrow]

Voice voice voice voice!

Voice for me now

[upbeat wasp]

https://chat.openai.com/g/g-Y1bhNwCdA-rock-simple
If you're like me when going through endless forms to fill or papers to read. You'll like this one.

[sand aurora]

Hi

[tight sapphire]

Maybe in the long run, currenty OpenAI is too valuable.
It can be used as scapegoat, guinea pig, RnD, etc. and if it blows up, they got the data and come up with a product to fill the niche.

And if it thrives they have stakes in OpenAI, so they would profit either way.

[nova jay]

Well, I'll say it again: Microsoft now has a sizable portion of Open Ai's crew, including their CEO, and with their money and general experience, they can easily demolish Open Ai (what's left of it).

And it's more likely that Microsoft will buy Open Ai rather than "crush" them.

[tight sapphire]

Agree, but with their current acquisition, they would probably build up their own in-house rival and will profit either way.

But I don't think that crushing or buying OpenAI at this stage is the smartest move.

OpenAI is currently just too perfect for test runs and if things blow up, they still will get the data and profit.

[nova jay]

What do you mean "buying OpenAi at this stage is the smartest move"? Its the best outcome for microsoft if they do buy it

That is essentially what they are hoping for, a chance to purchase OpenAi.

[tight sapphire]

I DON'T think it would be the smartest move.

[nova jay]

why so?

[fresh sun]

It would be a smart move from microsoft if they could obtain OpenAI, but I don't think that would end well for the rest of us. Satya may be the CEO, but Bill Gates is still the owner.

[nova jay]

he got my point

[tight sapphire]

The already have a good portion of the former core crew, including the CEO. They have assets in OpenAI via stocks etc. So there is not an immediate need to buy them up.

It is like having a second RnD AI block working for you, with less risk. And who knows with what innovations or problem solving they will come up with.

If you just buy everything you risk of creating an echo chamber and drowning everything with corpo policies. This make things slow and unable to innovate, risk things, and react.

By letting OpenAI be their own thing, they can copy what they need and build a solid corpo product of themselves.

[outer seal]

they effectively own them already

[tight sapphire]

correct

[outer seal]

as a wise man once said “you gotta keep em separated”

[nova jay]

yea

[nova jay]

They will not only get rid of one major competition, but they will also obtain their current developments technology and team, thus it is a good decision, and with Microsoft's cash and existing position in OpenAi, it is indeed achievable.

[tight sapphire]

OpenAI is not a competition.
They are partners.

OpenAI uses Bing not google for internet access, Microsoft has reportedly 49% sake in OpenAI.

They just got a lot of talent from the internal divide within OpenAI.

They currently basically have the cake and eat it.

As an example, if Microsoft would have acquired OpenAI in 2019 and it would have come to a similar fallout, the talent would have jumped ship maybe created a competitor.

But currently they can have OpenAI, push them, and they can absorb and profit from any fallout OpenAI is having with a very small risk of "rogue" devs starting a competitor.

Plus they can dodge some Monopoly Laws.

[prisma arrow]

so hows google doing?

didnt hear anything about their version

[nova jay]

They are not partners, openAi has a large part of OpenAi precisely because they want it, and they are simply waiting for the right opportunity to acquire it, which is likely what they've been waiting for, but who knows.

[outer seal]

openai has a large part of openai because they are openai

[prisma arrow]

iguess he means Micro

[warped pendant]

conversation lost right in the middle of coding, then finaly reword my question, finally outputs code but with placeholders after expressly asking not to have placeholders, 5 more questions wasted. Now I have to wait 3 hours to finish the code short of manually coding which I can do but 50+ questions every 3 hours specially with the AI constnatly pushing out placeholders would be better

[warm gull]

I know the feeling - its basically unusable if you want detailed code

[warped pendant]

Network loosing convo, and network crashing, and network and placeholders shouldn't count towards the 40 questions every 3 hours

[warm gull]

100% agree - but try telling that to OpenAI

[warped pendant]

GPT 4+ code interpreter used to be so much more stable and smarter, now every time I hit continue button I cross my fingers, and pray it doesn't crash ..

[warm gull]

same here man - i hear you and i feel your pain

[deft inlet]

@novel shadow is there any way I can get documentation that shows that I am allowed to use it? Not the icon, but documentation that shows permission to use ChatGPT 3.5?

[warped pendant]

wisth they had a version of enterprise license for indie developers that didn't cost 108K usd

its like they moved what was actually working into enterprise and what ever we paying for for 50 questions every 3 hours at the start of last month got reduced to 40 questions every 3 hours with more issues 😦 while I'm grateful for the service its also frustrating when your in middle of merging code and have to wait 3 hours cause it keeps doing placeholders or rewrites the entire code when all you as is for simple merge of files. Again I can code already but sure saves my fingers after 18 hours of coding, sucks my fingers alurgic to the plastics in my keyword now from the sheer amount of code I chern out before gpt even was a thing

[errant wasp]

Quick question to make sure I'm on the right track. I'm adding a chatbot to my companies website. The intent is to answer questions about the fairly large amount of information contained on the site. Should this be done entirely with a long fine-tuning list? There would be thousands of question/answer pairs. Or is there some other feature for the import of reference information? Secondly, it's important to do the best I can at limiting the chatbot to answering questions about the data itself. I'd rather it not answer unrelated questions. Thanks

[deft inlet]

This is what Google says that I need : If you have permission to use the content in your app, submit your documentation through the Advance Notice form. Use this form to provide additional information, such as proof that you're authorized to use the content or that your use is covered by an applicable legal justification. After we receive this written documentation, we'll respond to your submission.

[tight sapphire]

Agree, but the issue is verifying network loss and crash etc.

With a lag switch etc. such a thing can easily be imitated and boom, unlimited messages.

[steep forum]

Was GPT-3.5-turbo removed from ChatGPT? I am a plus user, but seemingly can't access it

[honest jay]

Hey there! 3.5 is still available on ChatGPT. Are you using the app or the web version? At the new chat screen, you have the option to switch among 4, 3.5, and Plugins.

[steep forum]

I'm so used to GPT-4 that I forgot where it was... my bad

[honest jay]

No worries at all! They switched it up on us a bit when they refreshed the UI recently 🙂

[steel shell]

gpt4 too slow now. Anyone else experiencing the same?

[raven hare]

Does anyone know why Chat GPT instantly responds with "Something went wrong. If this issue persists please contact us through our help center at help.openai.com?"

[maiden fractal]

try firefox

[raven hare]

Okay

Thank you

[maiden fractal]

❤️

[hollow bane]

Anyone know how to get that Pixar look without saying Pixar?

[willow ether]

I notice that as the chat history becomes longer, and its appended to instructions, then chatgpt 3.5 and 4 Turbo seem more likely to not follow the instruction and simply claim they did something that they did not (e.g emit an API call).
Has anyone else noticed that instructions are less likely to be followed when appended to longer chat histories? Is there a solution?

[honest jay]

Hello, are you referring to instructions given at the beginning of a chat in a message, instructions set in custom instructions, instructions/actions for a custom GPT, or some combination of these?

[chilly summit]

Hello, anybody knows why I'm getting "Message in conversation not found" when I send files to read (like docx, pdf or txt) ? ChatGPT answers the message that it has indeed read the message well, then I send another message and get this error

[willow ether]

In the API, I provide: [system] instruction 1, [system] instruction 2, [system] instruction 3, ... [assistant] hello [user] ... [assistant] .. [user] .. [assistant] .. [user] .. [assistant] .. [user] .. [assistant] .. [user] ..
The longer the recent history context (>10 interactions, the worse the result -- instructions not followed, says it did something in the instructions that it did not)

[honest jay]

Aha, sorry for misunderstanding you! If you're using the GPT models through the API and not on the ChatGPT platform, you might try asking over in #dev-chat

[gloomy widget]

Is there a way I can have chatgpt pull from my website. In the beginning when they first opened it to the web it could now it can't what gives? When I try to use plugins it can't do any code writing why are they kneecapping ChatGPT. Not the point is there a workaround to get it to work. Either a way I can authorise chatgpt to access my website or what?

[honest jay]

Is your website's robots.txt perhaps configured to disallow ChatGPT?

[gloomy widget]

I don't have a robots.txt

Because frankly I don't care who indexes what

[chilly summit]

Hello, I'm trying to send the code analyzer a few txt/docx file, read them and then generating text based on it.

First I had the error "Message not found in conversation" multiple times and then when it finally writes something after a second instruction, it stops and upon hitting "continue" it just bugs into "error generating response".
Upon trying for a third time to regenerate, it just says "Conversation is too long. please start a new one"

Could anybody help please ?

[gloomy widget]

you need to just reupload them

[chilly summit]

I can't continue the conversation since it says it's too long and I'm forced to make a new one

Repeating the process leads to the same result inevitably

[tawny raft]

QUestion: How can I record a demo of ChatGPT with sound recording of my voice and output... on Android? I tried the built-in recorder but I get no audio. I tried another recorder (AZ recorder) but As soon as ChatGPT takes control of the mic the aps stops recording audio.

[honest jay]

What version of Android are you on? I'm on Android 13, and I have a Quick Settings tile called "Screen record" which includes the audio recording option you described, "Device audio and microphone"

[zinc valley]

Guys what should I ask for for Christmas

[hexed ice]

Ask exactly the same to GPT 😄

[zinc valley]

I did

[hexed ice]

And ?

You need Q* I guess 😅😅😅

[dense halo]

Is it allowed or not or its like you're not supposed to, having another account? Like paying for another one is that a thing?

[steady delta]

Can someone direct me to in depth info on all aspects of reprinting, selling, and merchandising w/ GPT/dall-e?

[plucky timber]

Both wife and me have an acc in same household doesnt seem to cause issues. I doubt they care as long as you pay

[plucky timber]

Would be top tier funny if it replied Message limit reached 😂

[dense halo]

Just find that since the new features i've been hitting the limits a lot, and its usually from building and testing custom gpts, but its like so many ways to use it all now, I think the context and a lot of the image generating uses up the limit. It would be helpful to have a separate developer account and tester account is how I would probably use it.

[plucky timber]

They just need to offer better amounts for bigger monthly costs

Certainly with the features you mention. It's a 90% chance its just ignoring the instructions

Leading to long back and forth aka token wasting.

[dense halo]

or like save up from the time when not using it

[plucky timber]

Another cool idea would be to actually reward users "token refunds" based on good feedback when the bot ignores instruction via the feedback button

Because read file != the first 500 chars

And it will always do that first 😦

[honest jay]

Hello, scroll down to "Content" here for some info on content ownership: https://openai.com/policies/terms-of-use
The short answer is, you can merchandise, you just can't say that the art isn't made by AI.

[blazing belfry]

I am asking chatgpt to look things up for me online and it insist to use the analyze feature obviously causing it to error....

[steady delta]

Simply state made by AI? Does it need to include which AI brand?

Thank you!

[elfin birch]

As long as you don't say the content was made by a human, that's what's required.

[steady delta]

Got it.

[plucky timber]

Imagined By Human 😂

[warped pendant]

always a catch 22 sadly

[lofty furnace]

Anyone else got the 40 messages / 3 hours limit? 😦

[plucky timber]

Yeah but its worse because it errors about half the messages
a quarter is ignored instructions.

The last bit is "Progress"

[drowsy tinsel]

chatgpt can take forever when reading long webpages

[dense halo]

I think everyone does have that limit

[shell obsidian]

Hi everyone! Is there a way to use Custom Actions in GPT Builder to prompt another Custom GPT to build an agent based solution?

[fresh thistle]

Anyone know the best pluging that can go and red text/info from a site like vox can for youtube videos?

[tawny raft]

Android 14 on a Pixel 8 Pro.

[honest jay]

I'd imagine you have it then! Can you pull down your quick settings so you can see all your tiles? Is the "Screen record" tile there? If not, can you tap the little edit icon that looks like a pen, and see if you can add the Screen record tile? Then it should just be a matter of changing the Record audio setting to Device audio and microphone!

[tawny raft]

Yes, I do use it and I tell it to record the mic and the media output. But when I listen to the recording there is no sound from the app or from my mic

[honest jay]

That's strange! Do you have any kind of Bluetooth device connected by chance?

[tawny raft]

I tried with and without Bluetooth. In fact, I tried recording another app first and I could hear the sounds. Then, as soon as I displayed ChatGPT, the sound was cut off. So there seems to be something in the app

[honest jay]

It must be related to how voice mode is handled as phone call audio instead of standard media audio. I've seen them comment in #1070006915414900886 that this is because it lets them use some phone call-type features that standard media audio doesn't have, but this must be one of the tradeoffs to that implementation. I don't know if this would be a bug fit for #1070006915414900886 as technically everything is probably working as intended -- maybe #1070006151938314300?

[tawny raft]

Hmm, yeah, maybe. Or I'll need to record the audio on my PC and mix the audio on OpenShot or soemthing similar. 🤷🏽‍♂️ I rarely have iPhone envy but for this one...

[warm gull]

Doesn't android have a "split audio". feature?

Just checked, it's called "separate app sound" under settings / sound and vibration

[honest jay]

As in, different types of audio are treated differently by the OS? Yes indeed, and the ChatGPT app specifically uses Android's phone call audio.

[warm gull]

It enables you to route audio to different devices?

[honest jay]

I think separate app volume is just for setting separate volume levels for different apps, it would still be locked as audio type "phone call"

[warm gull]

No you can literally route audio to another Bluetooth device. I can on my phone select chatgpt as the app, and route its audio to a Bluetooth device in my home.

I just set it up, but none of my Bluetooth speakers work at the minute... lol. I'll try a Bluetooth amp I have a little later and see if it works.

[lofty furnace]

damn. yes got a lot of errors with my messages too

[tawny raft]

I can send audio to Bluetooth. THe problem is trying to record the audio output and input on the device when the ChatGPT app is in use

[honest jay]

Aha, thank you for clarifying--sorry for misunderstanding you! I just tested it out, I think there's a different difficulty. When I turn this setting on and then go to a normal media app like Spotify and play audio, I can then go back to these settings and tell Android which device I want it to play Spotify audio out of.

However, when I get ChatGPT talking to me, and go back here, this option does not appear for me. I am wondering if it's because it's phone call audio instead of normal media audio. But you're saying the routing feature remains for you? It totally disappears as an option for me, but will re-appear with Spotify. Are you on Android 14 by chance?

[warm gull]

Yeah, so I tried routing it to a Bluetooth amp I have connected to external speakers, but the voice still came out the phone speakers. So ig you are right.

[ocean escarp]

Will ChatGPT exceed 8000 limits for Community-built GPT's instruction?

[onyx marsh]

one of my most important conversations disappeared from an errant (but unknown) keyboard stroke while i was typing in the conversation. It looks like it was somehow deleted., how can i recover the deleted conversation?

[ocean escarp]

Probably long gone

if it was actually deleted

[autumn ocean]

wts difference between gpt and AI agent

[warm gull]

latter is dependent on the former

without a pre-trained transformer - there can be no agent

[autumn ocean]

like gpts?

[warm gull]

yeah - sorta like that

[autumn ocean]

thx man

[warm gull]

no worries

btw - what i mentioned above is assumed that you mean it in the context of ChatGPT

AI agents can also be other automated app's that have nothing to do with OpenAI

[autumn ocean]

yeah i se

[warm gull]

Tesla's FSD autopilot feature for instance is like an AI agent

which has nothing to do with ChatGPT's gpt ..if you see what i mean

[autumn ocean]

yeah, i got u

[icy comet]

Can ChatGPT only invoke Web Browsing once per message? I am trying to create a GPT that will perform multiple searches at different parts of its response. But in my attempts so far, once it calls Web Browsing once, it just summarizes the information and ends the message, regardless of what is asked.

[ocean escarp]

Bing Chat is great for that

I was considering using AI agent awhile ago, wasn't sure if it's actually legit plus the monthly pricing is pretty huge

[fickle forum]

does AGI emerge nowadays?

[ocean escarp]

Does it matter?

[eternal rock]

`` Server Tip: ``

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

[fickle forum]

if we human being are replaced by the AI when the AGI emerges?I mean the most jobs.

[ocean escarp]

lol

Delusional thought

[fickle forum]

[ocean escarp]

It's so one-sided thought

[fickle forum]

it's not the reason that the altman was fired and called back time and time again?

really curious about it

[ocean escarp]

The fired was regrettable and they had him back again. Also, AGI won't be able to outperform AI in performing tasks. AGI concept is to perform general tasks, but it outmatches by AI that was trained for specific tasks.

I think AGI is overrated because AIs seem to be more useful than that in concepts. People only needs a powerful tool to perform their tasks effectively. What can they do with AGI? It thinks on their own and the concept itself doesn't seem like a tool

Let's say I want AGI to make money by running a clothing shop. But it only can give me text instead of designing/generating artworks, finalizing the artwork concept, putting it on a clothing, config the price and placement on it, name it, publish it, marketing, and everything of that. You have to do all that

So, the idea of AGI replacing humans is silly

Next, we are going to be replaced by a pet companion AI

[prime swallow]

ChatGPT-4 can search many times in the same output, but you're going to need to give it a reason to. This prompt was one such:

A person on the OpenAI discord forum is asking about your ability to browse more than one in the same output.

Let's demonstrate that you can browse more than once within a single output, but you'd need a reason to that made it wiser and more efficient than a single set of searches and then discussion.

To that end, please browse to determine the current headlines then use your Python tool to provide a word cloud of some recent events and news in technology and science.

Next, please browse to determine the recent weather conditions across the US, and use your Python tool to design a creative graph that showcases recent temperatures across the US.

Lastly, please browse to determine the upcoming expected new movie and book releases in the next few months, and provide an annotated list of those entertainments we can look forward to.

Images and shared conversation showing it did search, then provide requested information, then search again, all within the same output: #off-topic message

[kindred ravine]

To me, GPT is extremely useful in terms of helping me saving vast of time on searching and fact checking, but I've seen quite enough influencer or media keeping spreading the fear that AI is replacing which kinds of jobs and blah blah. Well, say you are having a hair cut, even though there could be 1000 institutes having checked this AI confirming its superior in hair styling and trustworthiness, how much confidence you could easily put on it? It is same with many areas, replacing seems to be a cool idea, but in the forseenable future, it still has a long way to go.

[dense halo]

Guys I'm developing feelings for Sky, what should I do?

[icy comet]

It tried to do all these searches at once and then crashed:
https://chat.openai.com/share/b8aad1b8-61e5-47a1-8a79-5e4b2c0a90a1

Another issue is, OpenAI seems to have put a time limit on tool use per session. Here's an example with a custom GPT I've been working on to help me look things up:
https://chat.openai.com/share/fef1c60b-ad9d-44fa-9c96-23289ede6a00
It runs out of time to interact with tools.

[prime swallow]

nods So, you can plan around it.

The AI will follow your direction, even if it cannot possibly succeed.

So, you can direct it to stop at a reasonable point, such as: Only check at most 5 sites; provide whatever evidence you can at that point.

[icy comet]

Yes, I'll try to retool my GPT to work around the limitations.

[blissful geyser]

why do I get a network error so often? My browsing experience has been fine otherwisee

[delicate pier]

chat not woking on pc but works on ipad chatgpt 4 please help

[tawdry marsh]

chatgpt 4v(turbo) and gpt 4 classic, what's the differences ? seems like they have different limits as well.

i know one is packaged with bing, DA.
but other than that ?

[proven relic]

Hey, does somebody have had a problem updating to GPT 4?
I am on a waiting list from more than 2 weeks and I can't find any information why I cannot pass?

[azure sundial]

Why did GPT-4 become less intelligent, yet the number of users increased? I feel like this is just a marketing gimmick.🤔

[warm gull]

they stopped accepting new signups temporarily due to load constraints. no one knows when it will be lifted.

[silk flame]

Hello, I have a question. I know that we cannot use the GPT API to generate violent or erotic content. I feel okay with that, but I would like to ask if it is allowed to use text-to-speech API to generate such content?

[crisp grove]

bruh lol

[feral zinc]

the usage policy affects every service and feature

[silk flame]

ok, so if I would like to create an audiobook of a book like game of thrones which contains some violence etc. I can be banned, right?

[feral zinc]

yes

[gloomy quartz]

Hey Guys I really want to upgrade to ChatGPT Pro asap, but it says they have too high demand and I can't do it. Any idea how long does it usually take to get the ability to do it after joining the waitlist? Thanks

[scarlet rover]

they employ 100000 harward compsci people to write the responses, there never was any ai, its all jsut a joke

[maiden fractal]

Unfortunately no eta for now

[azure sundial]

This is not funny. The market cannot provide negative feedback for bad behavior, which is telling OpenAI that as long as it is packaged well, even if the quality is poor, they can still make more money.☹️

[scarlet rover]

capitalism?....

[silk flame]

Is it really true? When GPT3.5 was released I must admint, I suspected that they maybe hired thousands of people in India to handle all the requests, but then I threw away this theory. Do you have any legit sources for that claim?

[scarlet rover]

i hope you are joking man...

[rotund sparrow]

nope, gpt is really an llm obv

[rich surge]

I can confirm that this is true. Sam Altman's real name is actually Roger Frederick and he's the campus leader of Harvard that made it all happen.

[rotund sparrow]

Roger’s got a new chick now and theyre building for the next open

he won the last one as well, half of the first bout was perfect. the second half he screwed up bc insufficient energy, they took a lot in the first itself

ty @rich surge

[hard dew]

Is anyone else finding the mobile app to be a completely worthless piece of trash? I swear to god if you think that the web version has issues with constant errors, please never use the mobile app.

And god forbid I do what they say and just "rEgEnErAtE mEsSaGe", only issues is that it uuuhh, switches me back to GPT-3.5 for absolutely no reason what so ever, from GPT-4 plugins, which obviolusly is not what I want or need.

90% of the time "regenerate" or "continue generating" button in the mobile app does exactly, only that.

[wooden garnet]

Hey guys, what's the current messages cap for GPT-4?

[maiden fractal]

40/3h

[wooden garnet]

Thanks!

[maiden fractal]

If you have any issues I suggest to post detail in #1070006915414900886 . What cellphone, app ver, cellphone version what did you try etc

[craggy flax]

Wait you too?

[supple zephyr]

why is it making errors

oh there are issues rn

increased errors

[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.

[burnt hound]

Where are you guys finding live downtime reports? I can't get chat to do a word count, it keeps saying "error"

[maiden fractal]

Status.openai.com

[burnt hound]

thx, looks like it shows everything's workign though. I wish it would stop failing to execute its own scripts.

[summer saffron]

Did someone forget to replace the dead hamster that's running the chat gpt servers?

[burnt hound]

Chat's lost its mind today for sure

[tidal light]

that's why ChatGPT sucks lol

[summer saffron]

Where did that 10 billion dollars go that Microsoft invested into OpenAI, Bill?

[tidal light]

idk lol

i uhh wasted that money for windows 11

[summer saffron]

Smells like a scam to me, or it might just be the dead hamster..

[tidal light]

ChatGPT's servers suck sometimes

[plucky timber]

Errors count as msg correct?
And timeouts / network error too right.
Do you happen to know if the analyzer errors out and it auto retries if that counts as multiple?

Same q for very long messages , do they get counted as multiple, curious what you can tell me about those.

[maiden fractal]

Yes, error count

[tidal light]

like i cannot use ChatGPT because there is very high demand sometimes

[plucky timber]

The double count for analysis and long messages is that the case too ?

[summer saffron]

Yes, me too Bill

[plucky timber]

Asking since I'm running into early limits a lot (way before 40 msgs)

[tidal light]

sometimes when there is high demand instead of 3.5 there is "alpha" why is that

[summer saffron]

Might be A/B testing. You will be randomly selected and put into an alpha version?

[tidal light]

and that version doesnt work at all

[plucky timber]

Nah alpha isn't a new version during larger high demand occurring it was like that for everyone

[tidal light]

i dont have plus tho

[visual bloom]

There are currently a lot of errors at OpenAI’s end. They say that this issue is fixed it is not.

[tidal light]

oh yeah ChatGPT is broken

[supple zephyr]

Is some characters in files unreadable?
My c code had female gender symbol when I opened with notepad. I got 'error in input stream' and I wonder if it can be the problem here.

[summer saffron]

Your code might have some form of anxiety related to female gender

[maiden fractal]

I don't think so

[plucky timber]

Hmm then I'll have to dig further

[tidal light]

can someone tell me why is ChatGPT very good at making stories

[maiden fractal]

Cause it is a llm and have huge database

[tidal light]

ngl the stories it made are pretty good i'd say

[summer saffron]

It basically read all the Harry Potter books

[tidal light]

maybe

but yes, ChatGPT is sometimes bad at giving info

[vague cipher]

Do errors count towards your usage cap for ChatGPT limit?

[tidal light]

i think they dont but im not sure

and how does the usage cap work i have never understood it

[warm gull]

They do indeed

[maiden fractal]

Yes

[cunning mantle]

Yea it's getting a network error every time it tries tod o a respones for me. API is working.

[vague cipher]

Generating anything longer then a few lines is impossible right now with the network errors and just hit the usage limit from retrying these failed attempts. Sad.

[pearl solar]

Has someone also so much Problems with GPT 4 right now?
1 Month ago it was much faster and has no errors and now I am getting errors after errors..

[sharp stirrup]

I cannot log in. Anyone else?

[stray shell]

Hey, I wanna ask if chatgpt is not taking new user registrations right now. I tried signing up with my Google account, but it didn't work.

[iron crown]

is there a solution for the network error problem?

[maiden fractal]

For paid user yes

[stray shell]

Weird, I'm not sure why, but when I tried registering with my gmail for a regular account, it said, 'The email you provided is not supported.' Any idea what's up with that?

[maiden fractal]

Did you own an account before and deleted it ?

Do you try to register on a unsupported country ?

[latent maple]

is dalle 3 currently down for chatgpt plus?

[maiden fractal]

Status.openai.com

Working fine

[stray shell]

Hmm, this is my second attempt at registering. The first account is working fine, and I haven't deleted it. And I registered both accounts in a supported country.

[maiden fractal]

I suggest to contact [e-mail]

[icy comet]

My last 3 chats on ChatGPT have ended in network error even though this issue says "resolved" on status.openai.com
Anyone seeing this?

[drowsy egret]

yeah I been having issues this morning.

[stray shell]

Thanks, I will try it.

[blissful geyser]

im getting an "our servers are busy message" but downdetector says openai is fine, and im a paying gpt4 user. how am I blocked off?

[tidal light]

today ChatGPT is broken and there isn't a single day where it works fine

[normal quarry]

how is it broken?

[tidal light]

i mean everyone else is complaining that they get errors

[visual bloom]

Completion of chats does not work.

[normal quarry]

what errors have you been getting?

[tidal light]

well 3.5 works fine for me i don't have GPT-4

[visual bloom]

I have 4.0 and those what I said. Funny, the status page says everything should be ok but is not.

[normal quarry]

what is happening?

[tidal light]

i use GPT-3.5 bc it's decent

[visual bloom]

Completion of chats does not work.