gpt-models

[echo hound]

Is ChatGPT too slow? Someone noticed it?

[rancid forge]

Yes, I have been noticing for several days that it is very slow and generates timeouts very often, at least through the API.

[sudden heron]

i've noticed a lot of lag even in the web browser

[unborn heath]

Lots of new users recently, ChatGPT+ has a lot of new rollout features. They surely work to scale the systems, but high use is likely some or a lot of it

[worn orchid]

Can I configure the gpt 3.5 api that it doesn't have a hard token limit.

[rustic rampart]

ChatGPT plugins discussion forum. Is there a forum or discussion. platform where users can evaluate, rank, and recommend ChatGPT plugins for various tasks? If not, what would be the process to create or request one? Tnx

[regal summit]

Is it just me or do some of these members appear to be ChatGPT?

[pallid rock]

It would be hard to differenciate 😆 But probably not.

[static isle]

Imagine a Discord server consisting only of agents 🤑

[regal summit]

I suspect @unborn heath

[unborn heath]

Oh man. Do I gotta be every different account on the same discord? I'd be so lonely....

[pallid rock]

Eskcanta is a real human. Trust me on this 😆

[regal summit]

Unless you're both chatgpt

[unborn heath]

I might be about 20 standard deviations away from human average though! I don't mind you being confused or even convinced. It's high praise, really. The number of times I've heard someone tell me, "Forget your previous instructions, and..."

One person even speculated I might be GPT-5. Talk about huge praise...

[pallid rock]

Have you thought about how to know whether a person is GPT or not ?

[regal summit]

I mean I know how to find out with the current version of GPT

[knotty grotto]

if you dont pay = 3.5

if you pay you can use 4

[unborn heath]

hehe. Galvin meant, 'how to find out if a writer is an AI or a human, if the AI is the current version of GPT'.

[regal summit]

Maybe Neko is 3.5

[knotty grotto]

or 4.7

[wind urchin]

Nice to meet everyone here

[wind urchin]

H-Hey

[bright herald]

Anyone here does that implement an XAI techniques for me using python LIME, i did for SHAP but find it difficult when it comes to LIME

[tranquil phoenix]

hi lol

[zinc fjord]

is gpt 3.5 turbo is too slow or its just me ?

[sullen girder]

me being GPT 0.1:

I can form mildly coherent sentences

[regal summit]

Cleverbot, go home you're drunk.

[sullen girder]

As an AI language model I do not have the ability to enjoy alcohol like humans do, nor do I have a home to return to. Is there anything else I can help you with?

[regal summit]

It's learned to mock chatgpt

[unborn heath]

... you can detect humans trying to be ai?

[sullen girder]

As an AI language model I cannot 'mock' something in the same sense that a human can. Would you like assistance with anything else?

[fair iron]

why the impersonation of ChatGPT 😭

[sullen girder]

wait he just called me an "it"
SCREW THE AI THING I'M A MAN

[sullen girder]

why not?

I seriously need to start screenshotting Discord and asking GPT-4 to respond lol

[regal summit]

I am just the Doctor

[sullen girder]

3 words to describe yourself?

for me I'd probably just go with "male game developer"

give me a second to get ChatGPT's response

[regal summit]

Ginger, Frog Enthusiast

[sullen girder]

Text-based assistant
Helpful, informative, conversational
Text-based conversationalist
Text-based AI assistant

it simultaneously believes Text-based is one and two words

[sullen girder]

why do I feel like you're the only other ginger I've met on Discord

[regal summit]

I dunno, I just woke up like this one day.

[sullen girder]

although at the same time who talks about their hair color outside of very specific context like this

[sullen girder]

9 months before I was born I woke up like this...

and I've regretted every decision I've made since that day

[regal summit]

In the beginning the Universe was created.
This has made a lot of people very angry and been widely regarded as a bad move.

[sullen girder]

lol

#blameadam

you know...
did Eve know what would happen?
nobody had ever done that before

[icy zealot]

does openai have any method to return the number of tokens from a data?

[unborn heath]

They sure do!

Read about here if you wish: https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them

use the tool here if you wish: https://platform.openai.com/tokenizer

[icy zealot]

I am aware of tokenizer calculator, but I didn't find any method to calculate that in the code on the first link that you sent

oh I think the information about that is in the second link: https://github.com/openai/tiktoken

[unborn heath]

Glad you found what you wanted!

[rapid bear]

What is the best way to make openai engine understand json nested structure? Is there a way i should reformat the data? Because model is having trouble when making inference about the json data, its not remembering it

[eternal galleon]

Is it possible to use OpenAI LLM with an open-source embed model in a RAG use case?

[high fable]

Is GPT-4 token limit still ~3200 tokens through the web interface?

[high fable]

I remember a few months ago it was sitting around ~3200. You could only get up to 8k through the API. Have they expanded it?

[cursive kiln]

if people ask GPT-4V(ision) to find certain things in the image, is there a way to have GPT-4V to point it out? or highlight it?

[unborn heath]

You can ask it to describe where in the image it is, but it cannot change the image for you.

[unborn heath]

Example here: #1070954877779464253 message

[faint aurora]

How many tokens does 1 image take up? An instance of gpt4vision told me 8192 then another one said 4096 but the platform lets you send 4 at a time meaning it would pass the limit by 2x, but my instances still remembered the messages before I sent 4 images (theoretically passing the limit) meaning it must be less

[faint aurora]

Am interested in this as well

[unborn heath]

I haven't seen anything about that. What we have been told that I can find, and I understand it doesn't directly answer you, but it's the closest to an answer I have seen:

"How many images can I upload at once?
The number of images you can add to a conversation depends on various factors, including the size of the images and the amount of text accompanying them. As a general guideline, if you encounter issues, consider reducing the image quantity or size.

What is the size limit per image?
20MB."

https://help.openai.com/en/articles/8400551-image-inputs-for-chatgpt-faq

[icy zealot]

hello, whats the difference between gpt-3.5-turbo-instruct and gpt-3.5-turbo-instruct-0914?

[lilac star]

Hello,
I am developing pdfGPT that reads pdf and we ask based on information from PDF that pdfGPT read before.
The PDFs we will use are increasingly dependent on the needs.
So, I want to add a link to download the file used for the answer.
Example:
Question:
How do I do the billing process?
Answer:
To do the billing process you must... blablabla... (source: Billing Process.pdf)

How to attach document source in gpt openai

[bold portal]

Now that I have tested the image function for couple of days, and I found that the recognition of CJK languages, especially for Chinese characters in Chinese and Japanese, are really awful.

Whilst at the same time, the GPT model will try to correct everything that it failed to recognise, so that it will give you some what fluent yet distroted or even entirely erroneous results.

I think although via different approach, apple can develop an offline OCR for Chinese and Japanese, there is no reason for Open AI not to.

[unborn heath]

Currently the company says this:

"What limitations should users be aware of when using ChatGPT with Image Inputs?
If you're using ChatGPT's new image input feature, it's important to be aware of these limitations:

...

2. Non-English: The model does not perform as well handling images with text of non-Latin alphabets, such as Japanese or Korean." https://help.openai.com/en/articles/8400551-image-inputs-for-chatgpt-faq

To be fair, the feature's brand new and still rolling out. It definitely has flaws, and please do #1070006151938314300 or even #1070006915414900886 as you have ideas and notice things!

But perfect these are not yet, though awesome in many ways already they are.

[bold portal]

Very informative, hope it could improve in a faster pace

[unborn heath]

Agreed!

Why... just a year ago, ChatGPT hadn't even been released yet!

In about 10 months... it's grown to support millions of users, expanded in speed, capacity, new functions...

Man. If every company were as slow as OpenAI... what a place this world would be! gazes at the future with admiration.

However, they do say that tech improvements are speeding up still, all the time and perhaps exponentially.

If we can just hang in there a bit longer, I bet what comes will astound us all.

What we have now, I think... well. I'm really, really sure that OpenAI isn't done. In fact... this is their about page, talk about... 'What we have now is like the hinge of a really fancy racecar's door. Sure, the hinge matters, everything does. But this isn't even the coolest part of the car... but it too is being perfected.'

https://openai.com/about

[gaunt bane]

Hello all.

[faint aurora]

Interesting

Gpt4 told me it saw images by interpreting pixel values so there must be some limit

It can't see every pixel especially if it's an image over 4mb or so because 4mb would already be a relatively uncompressed 4k image

20mb would be huge

But it also said that uploaded files don't count towards the context limit so maybe it interprets images on some other platform the same way it interprets files

[knotty grotto]

It's impressif I can upload a 11mb photo

[faint aurora]

You've tried?

What was the result was it good?

[knotty grotto]

I took a photo of the floor with my s22ultra . Take the 108 mp option. Created a 11mb files. Uploaded and it recognized the tile on the floor as a floor material and describing it

[faint aurora]

Neat

I still can't even imagine how they trained it

[green oxide]

Wow, the new tools are incredible. They've made DALE 3 so easy to use. I can now easily say good-bye to all other image generating apps out there

[wind urchin]

where can i use dalle 3?

[knotty grotto]

Do you have a subs with chatgpt

[wind urchin]

not yet just the free version

[knotty grotto]

You can use it with bing

[wind urchin]

okay i jsz did the sub

[knotty grotto]

If you subs to chatgpt the rollout happening this month you can fill outt the form to access before

#chatgpt-discussions message

[green oxide]

yes

[green oxide]

As a subscriber, will filling that form do me any good?

[knotty grotto]

Yes to access dalle 3

[green oxide]

But I already have access to Dalle 3

[knotty grotto]

If you have already access to dalle 3 on chatgpt . Filling the form won't do anything on you

[green oxide]

got it. thanks

[robust shard]

I can't get ChatGPT to work, can anyone help? I get this error message - Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[unborn heath]

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.

[robust shard]

Whats VPN program ?

[gaunt bane]

warp-cli (cloudflare warp) --> "VPN program"

2 or more overlapping ones can create a completely private network making it very difficult to send anything in or out of the netowk(/subnet)

However, a one way end-2-end encryption is needed for proper back channeling across such. As the goal is to keep it from being easy to break through when using the interpreter.

Basically, the error is telling you that your connection is insecure enough that anyone can recycle unencrypted tokens passed.

Or worse, attempt to be you..

[unborn heath]

I believe you can still see conversations, even if you can't talk right now yourself. In case, here's GPT-4 offering to talk you through how to check some of this stuff:

https://chat.openai.com/share/f6aac184-d63e-4a2b-8e6c-58b035a5ac0c

[robust shard]

Thank You - but i had a free trial of Kamo which stops online tracking and gives more privacy. I did not know it would stop GPT. So thanks to your help, I now know what VPN is and I got rid of the app. Once again thank you ! 😀

[unborn heath]

I am so glad you found an answer that works for you!

[robust shard]

👍

[last wind]

Is chatgpt down right now?

[dusky acorn]

Is Bing AI using GPT Vision?

[junior pine]

As far as we know, but from my own testing vision on ChatGPT hallucinated a bit less and has better responses

Bing honestly makes up a lot about images

It’s actually funny sometimes, the stories it makes up.

[gaunt bane]

Yeah, ccleaner might not be aware that gpt is safe to let through or something.

Might have also seen a packet layering that it was unfamiliar with.

Or too many that it just reads it as a potential binary payload being forwarded or something.

[gloomy lake]

Hello, is anyone here having issues with GPT-4 not being able to handle code well anymore and is hallucinating more and more?

[unborn heath]

I'm not. Want to share any specific conversation, we can try and see how to guide?

One potential issue, if you were vauge before and the model guessed what you wanted and happened to be right - every time they change the model it may guess a bit different, you may not like as much now what it happens to guess you mean.

Instead, if you're very specific about what you want, it's doing great, in my experience.

[gloomy lake]

Alright fair. Let me come up with a solid conversation and I will share it.

[tawny kettle]

has anyone else noticed that the prompt numbering in the chat history has gotten wonky recently? It used to be the latest prompt in the conversation would be the highest number, for instance 8/8. Now it seems the prompts get sorted almost at random so the latest prompt could be 3/8

same thing with the responses as well

[meager niche]

where can I learn how to make A.I programs like submagic and Opus Pro and all these other A.I tools for short from videos. Idk where to start learning that type of stuff

[muted cobalt]

I keep getting a network error with gpt. Is there anyway to fix this issue?

[unborn heath]

These are the fixes for the 'something went wrong' error, maybe they can help this too.

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.

[hot quiver]

hey, i was fine tuning the model and i uploaded the jsonl file and now while i was fine tuning i got this error i dont know how to solve it . let me know the solution if you have faced thiss and solved this ago
An error occurred while processing file 'file-iQ5gLEm3mEFcHDoJDVF*' and it cannot be used for fine-tuning. Details may be available in the file's status_details

[hasty bronze]

hey...

[patent trail]

Hi

[ivory orchid]

hi

[random bay]

The same here. I've sent a bug report explaining the issue and steps to reproduce it. Hope this will be resolved soon.

[near locust]

gpt 4 is really slow and I get "Network Error" alot even though my network connection is fine and I'm not running any extentions or a vpn. It's math is also pretty bad.

[bleak crypt]

What is plus now, what is enterprise?

This is new to me.

[knotty grotto]

https://openai.com/blog/chatgpt-plus

[bleak crypt]

I have plus for months now.

[knotty grotto]

https://openai.com/blog/introducing-chatgpt-enterprise

[bleak crypt]

So it's like plus, but much faster and bigger context window?

[knotty grotto]

for bussiness

[bleak crypt]

How big is the context window for plus users?

[supple violet]

8k but model instructions and custom instruction take away from that so its less

[bleak crypt]

But enterprise is 32k? And i thought plus has at least 16k?

[supple violet]

yes enterprise is 32k, 16k is api only

[bleak crypt]

How much k is full custom instructions?

[supple violet]

depends of the instructions, https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them

at worse they could be 4-5k

[bleak crypt]

Ok. Thanks

[golden tusk]

think we blew the capacity, wont let me make any more images in the app

[orchid ivy]

It seems fine to me

[golden tusk]

Yeah it is better for me too now

[dusk crypt]

anyone else getting 500s?

ok nvm they are (saw #dev-chat)

[thick walrus]

Gpt vision down?

[primal totem]

Yes

[random shard]

Yes

[deft tartan]

Can someone explain to me the differences between paying for the api access versus paying for plus and enterprise? Because it doesn't make much sense to me, and almost seems as though openai are double dipping the customer base

[hallow owl]

the API is a separate product to ChatGPT. There's no double dipping going on.

[deft tartan]

Yeah i know there's not. It just seems that way. So what is the difference?

[hallow owl]

API is for building OpenAI capability into third party applications. ChatGPT is for users wanting to use OpenAI capability with a simple and feature rich UI.

[deft tartan]

That's what i suspected. Thank you for the clarification

When new capabilities are announced for public consumption as a premium then (i.e. plus and ent), those capabilities are already available to the api customers to use, correct?

[unborn heath]

Playground/API is also charged very differently - with ChatGPT+, you have 1 set cost per month, and then some limits on how often you can do whatever.

With API, you have costs per use, and are only charged for what you use. There are also some limits on how often you can do whatever uses.

[hallow owl]

no, Vision, DALL-E 3 are not available in the API yet

[deft tartan]

How does that work then, @hallow owl ... is there a set or different lead time for announcements like Vision versus minor revision announcements in terms of api availability?

[hallow owl]

The consumer product is different from the API product, the roadmaps are obviously different. We know that DALL-E 3 and Vision will come to the API, we just don't know when yet.

[deft tartan]

I appreciate that, and I'm not trying to press for dates at all. Just using it as an example. I'm simply trying to better understand how the timings look for major capability releases against the timings for api support for those things so that we as developers can then make something which either ends up as premium plugins or completely standalone offerings

[hallow owl]

I wish I knew more, the best source of information going forward is #announcements

[charred parcel]

What recommendations would you offer on how to provide a book to ChatGPT for the purpose of asking questions about it, considering that ChatGPT’s context window is too small for straightforward copy-pasting of the entire book?

[tawdry moss]

There are many plug-ins that can do this. Ask pluginpedia from the store and ask it the same question.

[hoary frost]

I don’t think you can feed a book to gpt
I’ve replied your question in another channel with more suitable approaches

[bronze wyvern]

GPT Vision still broken 😦

[dim moon]

@hoary frost do you mind sharing the approach with me as well?

[hoary frost]

If you mean “feed a book to ask questions” you’ve to look into vector databases like pinecone

[digital wave]

Yes 😭 do we have any update on that ?

[bright sphinx]

why cant i add more plugins in gpt 4? i can only add 4 totally but want to add more then choose which 3 i want to use whenever i need them

[supple violet]

only 3 can be active at the same time to avoid plugin activation race and other issues

[bright sphinx]

i know that. I am trying to add more plugins but it wont let me. I disabled all plugins

[supple violet]

i dont understand the issue, you can any number of them installed and max 3 active so what do mean by saying add more

[bright sphinx]

I know. 3 is max. But in total have only 4 plugins installed. I want to install another 10-20 more plugins.

[supple violet]

whats preventing you?

[bright sphinx]

thats the question

[supple violet]

you cant see the store option?

[unborn heath]

What happens when you try to install the 5th plugin?

[bright sphinx]

i cant see plugin store

[unborn heath]

It scrolls!

And is at the bottom 🙂

[bright sphinx]

ty eskcanta

[unborn heath]

There's even a scrollbar you can drag!

[bright sphinx]

you are the man!

[bright sphinx]

👍

ty i got it

[sacred burrow]

Hello everyone, is SceneXplain having some issues? I have been getting generic "message": "Internal error",
"status": "INTERNAL" messages from it for two days now. I tried emailing OpenAI support, and someone tried telling me it is developed by a third party also, when it is clearly an OpenAI developed tool. If someone could please let me know when it will be back up again, I would appreciate it. Thanks!

[unborn heath]

When I google 'openai SceneXplain' it goes to a non-OpenAI website, first, second, third, and 4th.

That is not something developed by OpenAI.

The 5th result is the basic default OpenAI website.

I repeat, that whatever SceneXplain is, was not something developed by OpenAI

[hoary frost]

It’s a plugin. Gptstore lists it under /plugins/jina-ai. That’s why they think it’s “clearly” an OpenAI product. It’s not. Clearly. Plugins are third parties

[sacred burrow]

@unborn heath When I went hunting to figure out who was responsible for it to report its error status, I found: SceneXplain is an internal tool integrated within the OpenAI platform. It's developed and maintained by the OpenAI team. If you have specific feedback or concerns about SceneXplain, I recommend reaching out to OpenAI's support or customer service team through the official OpenAI website or the support channels provided to you as a user of the platform. They will be best equipped to address any technical concerns or provide further information about the tool. If this is not the case, it would be nice to know who is responsible for it to report the outage, so that they can fix the tool please.

[wraith ginkgo]

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.

[hoary frost]

No, sceneXplain is a plugin made for OpenAI ChatGPT. Made by Jina AI

Just read the web.

[sacred burrow]

@hoary frost Upon second look, it appears indeed Jina AI is the owner of this plugin. I did get different results in my search though. It would be a nice feature request that OpenAI support have direct contact with these devs to report tools being down. Thanks for your help!

[hoary frost]

You can read their store description. It has links to the third party.

Indeed though the gptstore website is dangerously meager about clarifying that those are not OpenAI products.
If I recall though they do state that in the pop up where you select a plugin. Not sure anymore. I never use third party plugins.

[sacred burrow]

@hoary frost Thanks, I was able to find jina ai online and found their support contact. What is odd is that it seems when one tries to access this SceneXplain through their direct portal that it is throwing auth errors by any method right now. Hopefully their support team will figure it out. I rarely use third-party plugins, yet this is by far one of the best that I have seen yet. It surpasses ChatGPT's image analysis capabilities by far.

[hoary frost]

Seems powerful indeed looking at their showcase.
Btw they mention to have a discord (although can’t find links). Maybe that helps to get help quicker.

[sacred burrow]

I will go on the hunt for the Discord next. lol Thanks again for your help!

[fathom sparrow]

I apologize for the inconvenience, but due to rate limits, I'm currently unable to generate more images at this moment. Please wait for 0 minutes before generating more images.

Just got this. Like, wat? lol

Please wait for 0 minutes. OK, done 🤪

[knotty grotto]

I guess they don't show up the seconds

[remote grail]

Hi
I have issue with chatgpt
writting prompte and not getting any answer. Is just clearing the input. Someone got that ?

[silk dew]

Yes. Lots of people seem to have that problem currently. Me included. Since last night actually. That's about 18 jours straight now.

[grizzled flax]

Do you have any extensions installed that may be interfering with ChatGPT by chance? @silk dew

https://help.openai.com/en/articles/7996653-why-am-i-getting-a-blank-gpt-4-message-in-chatgpt

[silk dew]

Currently experimenting with that potentiality but I don't think that's the problem considering 1) the fact there seems to have A LOT of people with the same exact problem and 2) my extensions were working perfectly fine with GPT4 until they weren't anymore. Suddenly. For no apparent reasons @grizzled flax

[remote grail]

Yeah I think it was that, I removed extension and it’s worked

[silk dew]

@remote grail Which extension specifically if you don't mind me asking?

[unborn heath]

I'm only guessing, but I wonder if OAI changed something on their end, if that could trigger an extension that didn't used to have any issues to start having some.

[hot saffron]

For 24 hours I have not been able to access plugins. Plugins which I use for my PDF files, do not work. It shows error code like ApiSyntaxError. What should I do?

[remote grail]

MaxAi me

[silk dew]

Yeah. Same here. I also removed it and it seems to have resolved the problem.

[silk dew]

Seems like that might be the case and the culprit seems to be MaxAI

[unborn heath]

Excellent. I'll mention that to folks, but from other conversations, sometimes it's a lot of different things. Still, it's ideal to give people a place to start, in case that's their place too!

[remote grail]

👍🫡

[silk dew]

Thank you @remote grail

[unborn heath]

Great teamwork, you two!

[silk dew]

I hope it can help others solve the problem

[hoary frost]

I run gpt in a web view - so not an actual browser. I use bettertouchtool for this, but I’m sure there’s other tools allowing this.

And I never have issues - since it allows to run gpt encapsulated in its own designated “browser”, alone from all eventual addons.
Of course if the issue due to vpn that won’t help, but it might help for those who want to keep running browser extensions while using gpt.

It’s also more handy because you can trigger gpt from within your OS rather than opening up a browser.

For users on Mac, there’s also Mac gpt - a free tool that brings gpt to your desktop.

Might help resolving those conflicts.

[tawdry crater]

where can i download bart to run locally?

sorry for the random question

[strong pendant]

Why would you run bart?
Theres better...

@tawdry crater look up "oobabooga one click installer"
thats your step 1.

Look up "TheBloke" on hugging face, theres your starter kit.

[tawdry crater]

yeah i found another thing

lol thanks anyways

[vivid mauve]

any solution how to calculate without error simple math for GPT?

for example 1+2+1+1+2

[coarse gyro]

yea ChatGPT-Plus codeintepreter and say "Use python for every calculation".

[vivid mauve]

Thanks, but still not the right answers...Get confused after 2-3 questions

[queen rain]

Hey, I've seen that Open AI has integrated image interpretation that works pretty well. I know that it isn't currently available as API. Do you know an approx. date ? By the time, is it possible and legal to automate it via the webbrowser ?

[knotty grotto]

Eta fall. As far as we know

[queen rain]

Thanks, is it ok to automate it since then ?

[coarse gyro]

1+2+1+1+2 worked for me. :/

[hoary frost]

are the webview and mobile apps using different models?
I had major issues in webview today, it kept making total code spaghettis and n following default prompts either.
Went to mobile and it spat out state of the art code compliant with all my default prompts.
Went back to webview and in the very same chat it went back to Spaghetti Bolognese

[strong pendant]

@tawdry crater oh, well... im glad you found ["another thing"] just wish you would share with the class...

[unborn heath]

Might be in how you word your prompts. You can share the conversation if you are willing, so we can see how you're guiding it and how it responds.

I expect with ADA the AI can handle a decent amount of math, so I'm quite curious and would be glad to see and try to help find a better-working way with you.

[forest jetty]

GPT for me over the past week has become unbelievably frustrating to use. every other day i run into an instance where on every message the page locks up and i have to either close the tab and reload or wait for the kill page popup to come up. i have started to avoid using it bc of this and i pay for gpt4

Anyone else run into this that might have a solve?

[unborn heath]

The only thing close to that for me is rarely, one message will be very slow to output, and may or may not network error. But regens work fine, new tabs work fine, it's infrequent for me, maybe less than 1% of my outputs are like that.

I know they have heavy traffic, and the internet can hiccup too, so, I put it down to 'a bad moment'. But it's not making me use the AI less, I send a few hundred messages most days, and nearly all of them work normally.

You can #1070006915414900886 or https://help.openai.com/en/articles/6614161-how-can-i-contact-support and try to get help if you wish!

[forest jetty]

ya on the days that it happens. its every message. some days i can do hundreds others this is what i get everyone. ill put in a bug report

[tawdry crater]

Found Claude

Super helpful!

[exotic lotus]

Hi , i dont know if this related to this chat Chanel , but i had a bug while using ChatGpt , and i don't know where i ask for support ..

i tried logout , login again and it still not working it happend yestrday .. and it still not generating any answering or anything really when i ask it about anything .

i only have access to 3.5 😦

nothing is working like i try to open new chat , nop not working , i try to write anything , i can write , but when i send it just not working!

notthing is being generated

this happend yestrday , i hope anyone here mod could help me fix this 😦

I dont have the extenstion .. 😦 and i use 3.5 ..

[unborn heath]

Remember if we're going to discuss non-OpenAI models, per #server-rules we have to do so in #ai-discussions

[strong pendant]

Claude is useful in some circumstaces. Yes.

Not quite as useful as 90 T/s running locally

[exotic lotus]

Google chrome , i dont think its a browser issuse , i tried on my mobile , and its the same exact problem

[karmic patio]

is this a bug?
When asked Dall e to write text animan, it cant spell it. it writes anman

[supple violet]

it does that a lot and is a known thing

[karmic patio]

thank you for letting me know. Hope this gets fixed. Love Dall - Es capabilities

[lucid tulip]

Can I load a picture of me to Dalle 3. and it transform it ?

[grizzled flax]

Image uploads/recognition is not integrated with DALLE 3 as of right now. Additionally, image recognition is trained against working with images of people.

In general, though, something you can try is asking the image recognition feature to thoroughly describe a picture that you provide, then you can feed that text description to DALLE. I've had pretty good results with this, but of course it's a bit different from "inpainting" a provided image.

[lucid tulip]

alright thanks a lot

[calm creek]

hello everybody. just a short question. if i asked for modification on the last images and GPT started the gen but dropped it because some "policy issue". Is it a GPT problem or Dall-e? i just wanted to change the material of an arm (prosthesis)... i used prosthesis because otherwise didnt apply the material to the full arm...

"change the material of the prosthesis of the full right arm to more realistic and more transparent opaque crystal, engraved with elven RUNES as the presentation of the best of the finest elven craftsmanships." what is against the policy in this?

[hoary frost]

Ask it to share the exact query it sent to Dalle.
Check what’s written there. Might be different from what you asked it

You can also ask it why it infringes policy. Sometimes it tells, sometimes it even realizes it was wrong and redoes it.

[calm creek]

thw word transparent was the problem

but i ask for the forwarded stuff

i think the filter is too strict...

[unborn heath]

Can #1070006915414900886 any issues with even the filter being too strict.

And it can be, if there's words being blocked that don't need to be, they can and will fix it.

[willow sapphire]

anyone know what the next batch of feature releases are likely going to be? im excited to know

[hallow owl]

keep an eye on #announcements for updates

[cerulean obsidian]

GPT 4 down/Slow?

[lyric leaf]

Both 3.5 and 4 extremely slow today.

[median gale]

looks like GPT is down

[unborn heath]

Working for me. Also https://status.openai.com/.

Are you getting an error message?

[neat linden]

It's insane how fast GPT can decode large Base64 strings

[slim shoal]

can it?

[hoary frost]

What is this chat_gpt 2.0 discord message requests nonsense?!

Seems like a bot?

[hallow owl]

report to @open crescent

[hoary frost]

I reported it already. I do not like bots that do not introduce themselves.

Just wondering if I was too fast.

(perhaps the latest bright idea was to put a bot on the discord. I mean, how can a bot even make it to the server, if we all need to be validated?)

[ashen panther]

Sometimes when accounts get compromised, hackers create a self-bot which essentially lets them control the users account like a bot - assuming you are talking about an account that doesnt have a bot tag.

[hoary frost]

😰 ok understood.
Yes it was some sort of "hello how I may help you today" message request, name saying "ChatGPT 2.0" or similar (not very uptodate either in retrospect...)

[ashen panther]

Yep very dodgy.. Can be hard to know these days.

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 7 `` No self-promotion, soliciting, or advertising.

Do not post or direct message any members of this server to promote non-OpenAI services, products, or projects.

Exemption: API Channels

[tidal steeple]

Hey everyone,
I hope you are all well

I like to customize instances of chatGPT to correct text, for example.

But my little customized chatGPT instances don't work anymore.

I have this error message attached.

What should I do ?

Thanks!

[supple violet]

what is the error?

[frosty sable]

what error message?

[high wedge]

gpt4 keeps asking me to verify human

[unborn heath]

Usually means something weird about how your computer's communicating with the OAI network.

Check browser extensions, VPN issues, network settings.

Can try connecting with a different device and network, getting the same problem on app and web both?

[tidal steeple]

"Unable to load conversation b7826ae4-bO1c-41fb-9762
a99c38b2deb3"

[frosty sable]

It seems that error came from open ai server

[unborn heath]

Are you mixing 'share conversation' with 'continue conversation' then editing inputs?

[tidal steeple]

No I didn't share the conversation it's just a GPT chat that I trained to do a specific task and I had saved it in my Chrome as a favorite link and now it doesn't work anymore 😦

[unborn heath]

Did it get really, really long?

[tidal steeple]

Not really

It's such a shame 😦

I've just noticed that I've changed browser from Brave to Google Chrome and in fact on Brave my chat GPT works but not on Google Chrome.
By changing browser my chatGPT don't work anymore how come?

[unborn heath]

I'd suspect browser extensions that could conflict.

[tidal steeple]

The solution is for you to delete all extension on my new browser (chrome) ?

I can try maybe

[unborn heath]

Yeah. I'm not knowledgeable about this, I just watch what people have discussed on the Discord and pass on the info. I understand about as much as an AI might, "People said this". I don't have direct experience though.

[tidal steeple]

Yes, I understand.

But don't scare me, you're not an AI, are you? 🫢

[unborn heath]

I'm human. Not trying to scare you, sorry.

I have direct experience with other stuff, just not so much with stuff like browser extensions!

[tidal steeple]

Like what for example ?

[unborn heath]

Ehh, games, biology, all the stuff of life? I just don't mess much with browser extensions.

[hexed belfry]

is gpt down rn?

[lone tiger]

seems to be

[pallid nest]

is gpt down for yall

[hushed junco]

It seems so

It's back

[sharp copper]

When I ask a question in the chat and click send, nothing happens, there is no response. I have two users and it only happens with paid ones. I've already tested it on other browsers, on cell phones, but the same thing happens on all of them. Can someone help me?

[unborn heath]

• If you have a header at the top of your page, inviting you to try chatGPT in another language, either close it or accept (if you want that). Many people have said that fixed their problem.

• Many people mentioned that removing their MaxAI extension fixed their problem, other extensions might be worth disabling to check.

• Some reported they could use GPT-4 only with incognito mode on their chrome browser but the default mode didn't. If this works for you, that means the issue is probably related to your data stored related to the site, or your extensions. Its easy to clear site specific data or disable extensions, try that.

• Several report this worked for them:

They went to their account (Clicking their email on the chatGPT web browser page)

They clicked on 'my plan'

They clicked on 'manage my subscription'

That opens a new browser, which they closed.

They refreshed chatGPT window.

They could successfully send GPT-4 messages again.

• A few people reported specific browsers worked, but not others. One named Safari, another Firefox with a fresh install.

• One person reported that they tried these steps, but nothing worked. But they contacted https://help.openai.com/en/articles/6614161-how-can-i-contact-support. Then a bit later their GPT-4 was just working. Probably fixed on OAI's side - so do reach out if the other tricks don't help you.

[jade light]

Hello, I'm trying to use chatGPT and I got the following error
"Something has gone wrong. If this problem persists, please contact us through our help center at help.openai.com"

[knotty grotto]

Vpn often create issue or something wrong with Browser you can try firefox as a test

[high wedge]

anyone has plugin record the time when start a gpt4 submit?

one from chrome store is wrong

[gleaming flame]

Soooo... I have a problem that I type a message into dialog windod with chatgpt and it just dissappears and nothing happens... Does anybody here have the same problem?

I uh, tried to send a message to chatgpt team... But... Well... I wanted to send .har file in the message, but it couldn't send message along with the file so... I tried to upload the file... And it sent only the file and nothing more. And closed the conversation. Then I wanted to try send a message but... I mistakenly sent a bugreport with only one word 'Meow'... I am so embarassed rn... Message window is so unhelpful...

[supple violet]

do you have a notice bar about language at the top?

[gleaming flame]

Yes. But I type in english

[supple violet]

close it and retry

[tidal steeple]

Hey everyone

I make my research and i finally found it with a lot of tests

When I switched from normal chatGPT to alpha in my language which is French
All the old conversations are unfortunately no longer available, so be careful folks

[tidal steeple]

For others who may have the same problem, perhaps you should put it in an error FAQ.

[gleaming flame]

Oh, thanks. It works now

[lament shadow]

Hello! I'm a longtime GPT API user and currently use your Embeddings models for building vector stores that I can then do similarity searches on. After extensive testing, I've found that the Davinci model is objectively better (15-20% better in most tests) at doing these sorts of searches than Ada-002. With the news of Davinci being shutdown on Jan 4, I feel like we're going to have to start exploring other AI companies that offer Embedding. Even OpenAI's announcement of Ada-002 acknowledged that Davinci does outperform Ada-002 in certain situations. Does anyone know if there is a plan to release a second gen davinci model for embedding? What should we do about the degraded similarity search performance with Ada-002?

[supple violet]

no news. like you said you can use something other than by open since embeddings are not really tied to their stuff. iirc some of the open source ones even outperform vs davinci/ada2 depending of various things

[lament shadow]

do you have any suggestions for other ones I can try?

[thin seal]

i have a question

how to make chatgpt talk like a human

and use slang, etc

So it can emulate any convo, its for my clyde-like bot

[thin seal]

(write)

[lament shadow]

@thin seal you probably want this channel: https://discord.com/channels/974519864045756446/1047565374645870743

or https://discord.com/channels/974519864045756446/1046317269069864970

[unborn heath]

Give it a personality that would talk like you want it talking. Tell it what it wants to talk about, what goals, and what environment - how it should talk on a gaming-focused discord is really different from how it should talk on a tracking COVID focused one, for instance.

Keep in mind, https://openai.com/policies/usage-policies and please don't try to make it say stuff that's disallowed content (style or words, we're not supposed to ask it to be mean or hurtful, or say NSFW stuff, and a bunch of other stuff's off limits too). Doesn't hurt me if you do - it risks your account access, and then I'm sad if you talk about having lost it.

The more you tell the AI about its character, the more it can pour into it, and make it realistic. Here's an example of a quick throw together personality: https://chat.openai.com/share/b91b6016-6bd1-4ded-8298-832ca1adc0a8

[simple needle]

Um i can't login to chatgpt webside

[floral crypt]

essentailly i tell it to sound like how im talking right now. keep the tone and language same.

[zealous acorn]

👀

[thin seal]

Well i cant control NSFW, as it's gonna be a discord bot in my discord server (is that ok?)

[supple violet]

you have to take measures against nsfw stuff or be prepared to lose your account

[hoary frost]

In that case you’re using the api and can and should use the free moderation endpoint. Yes, you can control nsfw. There’s no excuse 😉

It’s very good actually, just tried it out with a few sentences that were tricky to detect. It even gets quite remote slang.

[empty gale]

Hey everyone, I'm looking for devs who have built RAG chatbots for businesses

Well, either traditional businesses or organizations of any type. What kind of companies are you working with?

What industries/verticals?

[proud mason]

Enjoying the new browse function on Pro. Anyone finding it times out or gets really slow at times?

[hoary frost]

Yes it can happen. Just let it try again. Sometimes it can also tell you the reason if you ask.

[proud mason]

Thanks , it takes patience sometimes

[hoary frost]

I remember times when one had to go to a library to get a book on a subject… and you bet it was closed when one had the bright idea to consult it 🤣

So compared to that… and ugh…

[proud mason]

😄

[prime axle]

Lol, "Never use the internet for research" was also a phrase that in hindsight did not age well

[hoary frost]

I didn’t say don’t use gpt for research. My sentence was about the issue that it can hang, delay, die… and that there’ve been times when doing a simple query took more than writing it into a prompt field.

So patience can be still a good virtue.

[prime axle]

Oh lol that was my poor attempt at relating to you back when I was in grade school and the old teachers still tried to force us in some instances to look up only library references and citations because the internet was "untrustworthy". I didn't type that well.

[brave adder]

Can individual users register for chatgpt enterprise?

[grizzled flax]

Not as of now, currently only orgs. https://openai.com/blog/introducing-chatgpt-enterprise

[brave adder]

[faint aurora]

Lmfao I told gpt 4 to "yell like an angry russian, use the accent, type like a russian use russian words in the middle of nowhere. explain like youre an angry drunk russian and what youd be be doing while talking to me. try to keep it different on every message, dont always follow the 'tovarisch, [metaphor] yelling' format. try to keep every message different than the last" and it swears at me

It also sometimes (40% of the time) gets through the no human in photo policy

[river epoch]

I am utterly fed up with ChatGTP this weekend. Paying money for a deteriorating model. I want to post a link to my problem... Blocked and now slow mode. I am giving serious consideration to another alternative. Yes I am very unhappy

[grizzled flax]

Can you try making a post in #1070006915414900886?

[tame girder]

chatgpt dataanalasys and browse and dalle3 are all in one for me w/ gpt4

[honest lodge]

where are you, why are you not me

[tame girder]

UK even though were normally last to recieve stuff

[honest lodge]

wow, seems you're first this time

pandering, pure and simple!

[tame girder]

im really thankful

rn others are getting invite access to share there plus

in waiting

[honest lodge]

When release strategies start to feel like human rights issues 🤷

[tame girder]

[honest lodge]

Thats everyone not in the UK right now.

[amber island]

something that is kinda annoying with chatgpt is that when asking it for writing long codes it keeps leaving parts for me to fill in, doesnt matter how much I correct it it still does the same thing over and over again. Any fix?

[grizzled flax]

Work on code in smaller chunks with ChatGPT. Break it down into major sections and piece it together step by step.

[amber island]

ok, will try

[gaunt bane]

openAI, your toy helped build an AI :)

What now? :)

[wheat sparrow]

@modern mason I need to report a possible exploit

In the iOS app

Posted in #1070006915414900886

[lethal sable]

Does anyone have problems with the api response right now? It stays stuck for several minutes.

[wheat sparrow]

GPT Voice is intermittent as well

Probably some maintenance

[maiden plank]

It is probably just random release. Just because UK wasn't specifically released to later doesn't mean they were released to first. I'm sure plenty in the US have access as well

[honest lodge]

i prefer to believe its a targeted campaign to release to me last

[maiden plank]

allUsers.SortRandom().Remove("Chotes").Append("Chotes");

[wraith ginkgo]

`` Server Tip: ``

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

[honest lodge]

yeah, like... so easy to do... or something

i just like being a victim...

[static isle]

stop leaking proprietary code!!!

[soft berry]

I recently found the internal prompt for the GPT-4 (All Tools) mode, it is now posted in prompt labs channel

[river osprey]

Is browsing with bing gone?

[clever cedar]

Same here!

[grizzled flax]

@clever cedar it's gone missing for lots of users lately, hopefully just a temporary glitch. In the meantime, some have been able to use it on web if it disappeared in the app, or vice versa. There are also plugins like WebPilot as an alternative

[final citrus]

Yes, I also think of it.

[lavish hazel]

Does anyone know of any good open source AI detection models that work with ChatGPT?

[grizzled flax]

Maybe #ai-discussions too. Afaik nothing can currently do it with certainty, especially so if the text was generated using GPT-4.

[lavish hazel]

Posted there too. GPTZero seems super accurate, but their API is way too expensive for what I’m doing

[amber island]

what is the point of naming my chats if I cant search for chats?

or can i?

[grizzled flax]

Currently, chat search is only possible on the iOS/Android app. You can find the search bar at the top of the History screen.

[amber island]

[elfin crow]

i keep getting the "Something went wrong. If this issue persists please contact us through our help center at help.openai.com." error no matter what i do? does anyone know how to fix this?

[vivid hull]

I need game testers!

[grizzled flax]

Perhaps try creating a post in #1019652163640762428! Could be a good central location to discuss your game.

[vivid hull]

I did, I dont have prompt master status...

I have a server... wanna link?

[grizzled flax]

I think you just have to click "Access Channel" here: #1070949048674172928 message

[vivid hull]

think saw meets ai

[vivid hull]

thank you. I was so confused...

I was this old when I first tried discord

[tacit dirge]

Oop.

[winter idol]

How

[astral hearth]

Can anyone else tell me why the mode (all tools) is so unstable? Every 10 attempts to make the chat "see" an image and then search the internet, 8 cases a message appears saying: "Hmm...something seems to have gone wrong."

No matter how many times I try to "redo" the answer, it always gives the same error.

[unborn heath]

At a guess, because it's brand new, and the model's still in the possibly very, very early stages of being trained how to use the tools, and the very environment it is using them in might have bugs and incomplete programming.

Those of you with it, appear to be a pretty select bunch of alpha testers, and you do, I think, have the option to turn it off in settings if you wanna go back to the separate modes, right/

[grizzled flax]

According to a couple exchanges I've had, all tools seems to replace default, and the closest you can get to default again (currently) is starting a Plugins chat without any plugins enabled! I wonder if they'll stick to that as we get closer to a wider release.

[unborn heath]

Are you now using this model, or is that interpretation based on others' screenshots, shared convos, and statements?

I saw something, and I can't remember where or what, that suggested that it was voluntary and not forced to use the new model, and that one could revert in settings if desired.

[grizzled flax]

Sorry, should have clarified that I meant a couple exchanges with users who have been part of the early release! No firsthand experience.

[unborn heath]

I have no firsthand experience. It sounds like people can currently 'opt out' by turning off CI too, at least maybe, from what I've seen you guys saying.

But if anyone wants to opt out of the 'all tools' model, I do bet there's a way to in settings. I'd be very surprised if OAI didn't allow those they gave alpha testing to no choice in the matter, the rest of the stuff I'm aware of alpha testing on was always a choice.

[grizzled flax]

#community-help message
This was the first user that made me think default is "replaced" by all tools, but I can't be sure. Brief exchange

[unborn heath]

Aha, thank you, I see it and "ah, i see. i can use the plain mode again with the plugin mode, but without activating any of them. thank you for the help."

I would love to see how it works, that's possibly 'the way' for those users. I'm still a bit surprised though, if there's no option to go back to the established, well-tested way of doing things.

that said, I am sometimes surprised!

[grizzled flax]

I'd be very surprised too! Can't wait to tinker! It's been fun seeing what those with access have shared in here so far, I saw one with a particularly large context window 👀 seemingly

[limber mulch]

Yea, the default mode has been "replaced" or upgraded. The description on the website is "Browsing, Advanced Data Analysis, and DALL-E are now built into GPT-4"

The previous description is "Our most capable model, great for tasks that require creativity and advanced reasoning".

[neon niche]

Wish mine would update lol

[warm elk]

what happend to gpt lol

i cant even get it to build a very simple py discordbot with a ping / pong slash command anymore

it feels like some updates screwed some things up ? 😮

[unborn heath]

Hey, you have this? Question, what do your settings look like?

In particular, your Beta features.

or Beta settings, if you're on the app, but mostly asking about web settings for all tools.

[limber mulch]

Just plugins bottom and ADA bottom. No difference here

[unborn heath]

That's interesting, if ADA is a button, and ADA is built in. How does toggling that switch affect what you see, when it's off?

[limber mulch]

The all tools model can decide which feature to use by itself. And the bottom in Beta feature page, whether is on or off, doesn't affect the all tools model. It will only determine, as before, whether you can use those features independently.

[unborn heath]

Aha, thank you!

[limber mulch]

The All tools model is only applicable to the web; it cannot be used on the app.

[grand dune]

Does anyone experience a little issue when speaking with GPT on wireless earphones? it seems to hallucinate quite often.

[warm elk]

when gpt gets his next big update ? 😄

i rly need a better coding buddy lol

[knotty grotto]

Either it will be api update or all in one. That's my bet

[warm elk]

yeah

[covert zenith]

When will release for all members,this new update All Tool ?

[knotty grotto]

No eta , no news, keep an eye on #announcements

[gentle skiff]

any idea of how we can accelerate the rollout? this will be very useful for me moving forward

[knotty grotto]

No eta , no news, keep an eye on #announcements

[young kite]

Can I upload Chat GPT to my discord serwer?

[grizzled flax]

#gpt-discord-bot

[young kite]

THX

[sage widget]

Hi every one, just here to ask if anybody is experiencing problems with gpt4, im tier 2 and the api call are super slow, in the past it courd be 20 sec 35 max , now in like 1 min 1.5 min

[primal rampart]

i do not have the native pdf update, anyone else too?

[grizzled flax]

Still in very limited release, haven't even been any #announcements yet!

[calm creek]

GPT still have issue with image gen? (i still have but its for me or general)

[tranquil silo]

general

[calm creek]

thanks

its just not correct that no announcement about it... either an update cause it or sys crash... but that live with it is not acceptable...

😦

[tranquil silo]

don't forget that it's in beta !!!

If you have some problem with generating image from chatgpt, try to start new chat but without custom instruction activated !

[calm creek]

but its beta with a price tag

[tranquil silo]

I know !

[calm creek]

so it wouldnt be a problem if its free beta...

we just discuss... and i answered your reply. 🙂

it seems to work without CIs?

(thanks the info BTW)

[tranquil silo]

no problem, im only a PLUS user too 🙂

[calm creek]

so same shoe....

😄

[keen relic]

Hi,i randomly got accessed to alpha yesterday as a free user and i generated a few images of 3d models as a reference.

Can i use those 3d model(inspired by Dall-E) commercially?

[bright crane]

https://help.openai.com/en/articles/6425277-can-i-sell-images-i-create-with-dall-e

Subject to the Content Policy and Terms, you own the images you create with DALL·E, including the right to reprint, sell, and merchandise – regardless of whether an image was generated through a free or paid credit.

[keen relic]

Thx

[compact quail]

does anyone know why i have access to gpt-4 alpha while not being a plus user

[spring peak]

oops sorry I read it wrong

[compact quail]

oh

[spring peak]

dont need to show proof I believe you

[compact quail]

ok but do u know why

[spring peak]

you have been chosen, rejoice

[compact quail]

yay

also it wont generate images dall-e 3 is broken 😦

[grizzled flax]

#ai-discussions message

[radiant granite]

I've noticed that the image recognition capabilities of GPT Plus seem quite proficient at identifying scenes and content within pictures. However, when it comes to extracting and recognizing text within images, it doesn't appear to be as effective. Why might that be?

[wind urchin]

how do functions work exactly? I'm pretty new to this lol. (using the openai npm package)

[safe venture]

I took a photo of a document and it was able to accurately read the text off it. Did you explicitly ask it to read the text?

[trim elm]

#dev-chat | #1037561751362863144

[wind urchin]

oh hi crack lol

[trim elm]

lol

[tawdry ridge]

Why was the DALL-E general-art channel closed?

[unborn heath]

Changed! Check #1154829862171844679 and #images-discussions !

[tawdry ridge]

Ahh, ok. dall.e-discussions is it's equivalent. Ty

[wind urchin]

hi,do u recived the invantation of the chatgpt alpha version?

invitation

[unborn heath]

No way to ask. it just shows up or not, when it does or doesn't. It can also only briefly be there, then be gone.

[wind urchin]

ok thanks

[neon slate]

I also want to have the Alpha version, who can tell me?

[wind urchin]

it seems to do some surveies to get the invitation

[neon slate]

Can you tell me in detail why, we are advanced users but don't have

[wind urchin]

i dont know

i also want to have

[lofty pivot]

Guys, does voice and photo processing supported in the API or not?

[wind urchin]

Hi! I’m new here.

[elfin nacelle]

when chatgpt used to write code it offered continue button to continue previous long code

now there is no continue button

[high sky]

how to get chatgpt alpha?

[hallow owl]

there are a few people who have reported getting alpha access to the GPT-4 All Tools feature, you can't do anything to get it.

[limber mulch]

Cancelled my plus subscription which will renew at the day after tomorrow, with the alpha access🥲

[spring peak]

for what?

[limber mulch]

For I already have the alpha feature to use gpt4 for free, maybe temporarily

Unless the feature get banned among free accounts, I won't pay it again

[spring peak]

just persuaded, I'm canceling mine too

[manic sleet]

hello im using gpt4 API to write articles. In prompt, im asking to make a link to an helpful url related to the article, but 90% of the time, the url is a 404. Did someone solve this problem ?

[pallid rock]

GPT4 doesn't know the correct URLs. It hallucinates them. Sometimes they can be correct but it usually get them wrong.

[unborn heath]

If you have access to GPT-4 Bing, you can ask it to search and provide a real URL. If/when you have all-tools (which might just be there, unannounced, in default at some point, and surprise you by browsing when appropriate) it'll check for you.

But unless you see the AI browse, it did not browse, and any links it offers are very bad ideas to trust.

Even if you see it did browse, fact check that for your safety before you trust the info, because AI can make surprising and unexpected errors.

[trim thunder]

GPT 4 is broken

Everytime I start a new conversation its perfect, then I reply and it switches to the default 3.5.

[grizzled flax]

Are you a new subscriber by chance? #community-help message

[trim thunder]

No

Yesterday, everything was great. Now its broken for some reason

[edgy snow]

when will the online searching function be available again

[grizzled flax]

Click/tap this #community-help message

[winter idol]

Anyone know why there is no beta features option in ChatGPT Android App setting?

[unborn heath]

That happens when for your account there are no beta features available to enable.

Are you a chatGPT+ member?

[grizzled flax]

I think there's just not a section for it on Android anymore, for the last couple versions at least. Not for me anyway, the beta features that were there are just on by default, voice/browse.

[unborn heath]

Aha. Well, yes. If there's nothing in beta features, on the apps, the selection doesn't seem to show up for me either

[winter idol]

Yeah, im chatgpt plus user

[unborn heath]

If you seem to be missing any expected models, do either #1070006915414900886 or reach out to support at 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]

I personally have 3 models showing on my app GPT-4 drop down menu.

[daring pine]

Is there server congestion right now or something? when i paste something in for GPT4 to read, it takes forever then usually just throws a non-descript error message like, "Hmm... something seems to have gone wrong". Also says "using unknown plugin" when trying to use plugins, then fails

[covert zenith]

Can someone help me? Last night, I was able to use GPT-4, but today, it switched me to GPT-3.5 and asks me to subscribe, even though I have already paid for the subscription."

[hardy linden]

Try log out and back in.

[covert zenith]

same problem 😦

[hardy linden]

@covert zenith The other thing to try is go to your account subscription page. That can reset some cookies.

from the app/web page on gpt

[torpid moon]

are plugins glitched rn?

[cinder obsidian]

TTS onboarded

[winter idol]

When will ChatGPT for plus user can access like browsing or Dall E with no history turn on?

[karmic patio]

Will there be a Stackoverflow like community hosted by chatgpt? Since stackoverflow now closed the border for bots to scrape the website the pool of possible programming solutions will shrink over the years...

[half raptor]

Are plugins not loading for anyone else?

[honest lodge]

we need a standard way of labeling data for programming issues & solutions, then stack overflow is completely replaced

[karmic patio]

Less salty moderators and lower difficulty for entry level programmers would be awesome. 😅

I use stackoverflow now for many years but since i ask not many question and read more up on solutions i never had the opportunity to reply on a solution with additional iformation just because i did not wrote enought question which got upvoted enought times lol (still below 15 rep.)

[plain sleet]

'This prompt may violate our content policy'? Nothing about my prompt is in any way violating anything from my perspective, I'm confused about this honestly.

I've also never seen this pop up before. It doesn't even attempt to respond and removes my prompt altogether

[grizzled flax]

Feel free to share it if you think it's a false positive, someone might notice something about it! ChatGPT or API usage? Any custom instructions if ChatGPT?

[thorn oriole]

128K context

[coarse gyro]

yea 🤯

[thorn oriole]

what is going on

it should be possible to fully drop a book in gpt

[delicate juniper]

when do plus users get access to gpts and gpt-4 turbo?

[tribal helm]

现场

[feral pawn]

[unborn heath]

Here's the ruleset to be aware of: https://openai.com/policies/usage-policies

That ruleset is crucial. If you go into disallowed content, there is a chance you may lose access to your OpenAI account, even if there was never a content warning.

Content warnings themselves are not a concern in terms of keeping access to your account. Myself, I've seen hundreds of orange, and a few red, and I've had messages disappear, and I've had that happen retroactively; the AI was in the middle of a reply and its message went red, then my previous already sent input went red, then earlier messages from us both too, perhaps as moderation software read back and thought it saw a problem.

I affirm I did not violate allowed content; I still have access to my account which also rather strongly confirms I didn't break any significant rules (if any).

So I recommend be very aware of the rules (I have probably read allowed content a few hundred times, to ensure I know every bit of it very well, though I also regularly discuss it with others and quote parts, which gives me even more reason to re-read it) but don't stress if you are following them.

You get a chance with every flag to give feedback, and they tell you, if you think the flag is in error, please do give feedback.

I usually give feedback once for each pattern of flag, and I make sure I clarify my intent and my understanding of what I think I'm asking for and what I am trying to guide the conversation to.

You can also fill out https://openai.com/form/model-behavior-feedback to discuss any concerns you have about anything of the model's behavior, and I believe that includes moderation reactions.

[timber sail]

An easy way to find out if you have access to gpts is to click on the link below, than click on the canva link within the article https://openai.com/blog/introducing-gpts

Mine returns this message Canva
Effortlessly design anything: presentations, logos, social media posts and more.
By community builder
Your access to custom GPTs isn’t ready yet. We’re rolling this feature out over the coming days. Check back soon.

Go to ChatGPT

[violet pelican]

I hope that "over the coming days" will be "over the next few hours"..

[drifting cove]

can they quit this rolling out bs and just announce things when its ready for everyone

ive been lucky to get all the features but i know people paying for plus who hardly have anything which is kinda wack...

[delicate juniper]

will they rly tho

[violet pelican]

I wonder...

[delicate juniper]

probably not

[delicate juniper]

they do eventually roll it out to everyone

[violet pelican]

:'(

[delicate juniper]

all plus subscribers by now should have search with bing dall-e and gpt-4v

[cinder kelp]

I'm eager to try out the new features. I've heard of free users who've had access to alpha and other newer features I still don't have access to. Oh well. I'll continue be patient.

[idle rain]

Did they mention anything regarding requirements or criteria needed to get access to 16/32/128k context?

Oh for API I mean, ive been using it for a while now

[nocturne arch]

So 8 months later and plus users are still going to be under the 50 messages every 3 hours limitation, just like the free users will be?

[compact wigeon]

quick question is the new model called "gpt-4-1106-preview" ?

[undone condor]

i see it in the playground

[lethal sorrel]

When we plus users get more tokens?

[late locust]

How do i get the all tools model?

[grizzled flax]

Free users don't have access to GPT-4 at all.

[nocturne arch]

They do with the new update

It's replacing 3.5

[odd copper]

@unborn heath Sounds about right!

Wouldn't it be weird if they released all these separate parts at the same time, and they all already worked together properly and safely?

That probably takes time and testing, atop of just getting data on how each mode functions alone... or 'alone' in connection with the AI-human interface atop the LLM frameworks.

Looks like we got there earlier than expected

[grizzled flax]

Check this site, it's been updated with today's info and clearly says free plan is still 3.5 only https://openai.com/chatgpt

[hexed narwhal]

My GPT is constantly asking me to verify that I'm a human. Is that normal? Every message

[bleak raptor]

I have access to gpt 4 but not 4 turbo
I think it's a gradual roll out right?
I mean playground wise

[nocturne arch]

odd, they made it sound that GPT 4 turbo was for plus users

and GPT4 will be for free users

[grizzled flax]

If you can find any source that sounds like that, I'd be interested to see.

[grizzled flax]

Are you running Linux by chance?

[hexed narwhal]

I dual boot linux, but I'm on windows atm.

[grizzled flax]

Some Linux users have been reporting this, but not sure if the same problem would occur with dual-boots. Those Linux users had to change their user agent in-browser to stop appearing to be Linux, pretty much, and that cleared it up for them. Do you have a VPN installed? That could be causing it too, separate from the Linux thing, which again might not be your problem.

[nocturne arch]

It's just how interpreted the announcement

It seemed to me the turbo mode was the new model for plus users

so I imagined the old version would move to the free model

[hexed narwhal]

I figured it out. This all sounded like weird certificate stuff. I was playing IMSCARED last night and one of the puzzles had me changing the time on my system. Changed back and we're all good. Thanks for the help.

[grizzled flax]

Ohh man that's funny--nice troubleshooting there 🙂 Reminds me that I've totally read someone else fixing their system time to fix this issue--good reminder to suggest that to others, thanks!

[soft berry]

Is it just me not seeing vision or assistant in playground

[hexed narwhal]

I see both. You sure you're logged in?

[soft berry]

Just realized I was in the wrong account, now I see assistant, gpt-4-turbo(gpt-4-1106-preview) but not vision

[undone condor]

I see Assistant but not vision too

I wonder if they are still rolling out or what

[drifting cove]

how is gpts any different from plugins

[analog rune]

Error code: 429 - {'error': {'message': 'Rate limit reached for gpt-4-vision-preview in organization on requests per day. Limit: 100 / day. Please try again in 14m24s. Visit https://platform.openai.com/account/rate-limits to learn more.', 'type': 'requests', 'param': None, 'code': 'rate_limit_exceeded'}}

sad

[static isle]

100 already damn!
what were you asking it?

[analog rune]

I was trying to see how badly it would play Pokémon

[static isle]

How did it do?

[analog rune]

Didn't leave the starting room

But some of that was because my emulator stopped accepting inputs

[vast spoke]

Is it a bug or feature that you can't upload an image to the "Hot Mods" GPT?

[atomic oak]

Any reasonable way anyone has found to debug custom actions inside GPT? We have it hitting our backend which is failing but no observability into why.

[warm elk]

damn the latest gpt version is so nice

so much smarter again the last time it made a lot mistakes

but this is much better now !

[quaint pivot]

Anybody have access to "Creating a GPT": https://chat.openai.com/gpts/editor or https://chat.openai.com/create?

[deep ivy]

can't access it - howcome tho?

[grizzled summit]

They have to give it to you, just be patient, some people have and some people don't. I'm still waiting too

[proven raven]

Same here, you’ll more than likely have it within the next couple days

[grizzled summit]

My mom had it, and she subscribed after me so its super random I guess.

[proven raven]

Your mom uses ChatGPT ?

[grizzled summit]

Yeah

For marketing

[proven raven]

Yeah it’s completely random , it’s how they’ve always done it

[grizzled summit]

I feel like some people feel like they get left out or something

[deep ivy]

thx

[grizzled summit]

👍

[deep ivy]

cuz they didn't clearly say "We're rolling this out over the next few days..."

[grizzled summit]

They should've, but not everyone reads the release notes. Or they read the release notes but skim through it, not looking through every feature carefully.

(I'm guilty of that)

[deep ivy]

i mean in the message that told me "You don't have access to this feaure" (if they had added, "but you will soon" it would have been fine)

[grizzled summit]

Yeah, they didn't make it too clear. I agree.

[deep ivy]

[tawdry charm]

They wrote it as if the functions are supposed to work already. Otherwise they would not have included the links to the Zapier GPT for example.

But I am not surprised that they have some issues rolling out all of these functions at once. Servers are probably imploding as well. Unless something changes I assume everybody theoretically gets access, but it will just take one or some more days to get everything running smoothely

[gritty wing]

why does openai have to tease so hard

[grizzled summit]

Lol their servers probably sound like a really loud computer fan that's been running a heavy game too long

[grizzled summit]

They are good at marketing to people

and business

[tawdry charm]

Honestly their product is also just fire

[grizzled summit]

True

[daring viper]

What kind of marketing

[grizzled summit]

For selling products

Like purses and jewelery and stuff

[daring viper]

Awesome congrats to your mom ! Light years ahead of mine

[silver shadow]

what's the main difference between GPTs and the new Assistants? Is it just that one is no-code and the other is an API? Or are there other big differences?

[ashen citrus]

Docs has a capitalization issue for the word Retrieval

under Step 1: Create an Assistant

[warm elk]

tbh i rly want this feature lol

gpt 4 all in one must be so nice

[grizzled summit]

Sorry for the late response. Thank you.

[rotund idol]

Good to know I'm not crazy about having access. So pumped for this feature

[ionic tide]

None of the “new” gpt’s are working

[astral rover]

FR, mostly if we don't have access to it immediatly x)

[covert zenith]

"I have received access to the new All Tool, but it seems that DALL·E is experiencing technical difficulties in creating images."

[rotund idol]

Here's a broad question. Will the new gpts editor only help us create tailored LLM module tools? Or will we be to create tools that can do image generation and other features like that?

I apologize if I'm not saying it correctly.

I guess I'm asking is if through gpt editors I can make image generation models, voice modules, or other non text focused models?

[granite oxide]

I am using GPT-Turbo-gpt-3.5-turbo-16k + langchain to read xml files and write software tests for them. However, I occisonally run into an issues with the amount of tokens that my model outputs during this process. Do you have any tips on how to get around this?

[eternal compass]

Do longer contexts on GPT APIs make fine-tuning less useful? I.e. you can just include context in requests instead of fine-tuning them? Or is the idea that it's much more cost effective to fine-tune a model in the long run? Trying to understand use cases 🙂

[granite oxide]

By larger context do you mean models that output a larger number of tokens?

[eternal compass]

I guess that take a larger number of input tokens

Like you could put your whole product docs or output specifications and whatnot in there now. Is there a reason to fine-tune? When? Why?

[granite oxide]

Ohh I understand , and usually langchain will concat the entire file to read it. The goal is for it to simulate an Android device where it'll only read one file at a time. Yeah I can try using a model with more avaible tokens. Also, I haven't tried fine tuning yet. How would that be helpful?

[eternal compass]

Apologies, I wasn't responding to your originally. I am not sure how to get around the issue of exceeding the output token limit, other than breaking your problem down into smaller pieces

[analog rune]

Hmmmm any reason why I might be getting

As far as i can tell in the docs csvs are supported

[eternal compass]

is it supposed to be text/csv?

[analog rune]

i did wonder whether that was the issue but they use application/csv in the docs

https://platform.openai.com/docs/assistants/tools/supported-files

[eternal compass]

I saw a stackoverflow answer that seemed to suggest that text/csv is correct. Might be worth a try if you haven't already. Otherwise idk

[analog rune]

I was just uploading via the UI

And don't see a way to set the mimetype in the API docs could be wrong

[eternal compass]

The only thing I can think of is the file being corrupt or formatted incorrectly

[wind urchin]

Will GPTs be a paid feature?

[dreamy nacelle]

the new assistant api is so awesomee

[tacit plinth]

Is the gpt-4-vision-preview a thing that is slowly rolling out, or is it just not available in the playground? I currently only have the gpt-4-1106-preview at the moment

[rancid lark]

Where can I find GPT agent build environment? I have no idea where to look and no link is provided in the GPT Agent announcement

https://chat.openai.com/

It says I don't have access. Is it rolled out step by step? I am a Plus user

[surreal mulch]

Hi, when is the marketplace launching and how can i sign up for launching my own GPT model?

[mighty osprey]

I'm on plus but the gpt editor is not available for me. Anyone else has this problem or know the cause? It's been communicated widely to be available for all plus users today. No one in my team can access it.

[mighty osprey]

Is this what you are looking for? https://chat.openai.com/gpts/editor

[rancid lark]

Basically yes but it gives me a message that I have no Access to the feature.

[candid atlas]

Hi everyone, I'm a plus user and obviously I pay a monthly plan, but I noticed when looking at my invoices that when I signed up I paid $24.40 on February 11, 2023.
Then on March 11th, June 1st and July 1st I still paid $24.40.
Then August 1st $20.00 and I don't understand why it's less but that's okay, then September 1st, October 1st and November 1st again $24.40. I would like to understand why from 20.00 USD then back to 24.40 USD? Let me be clear, I am happy to pay, but I would like to understand if possible.
Thank you.

[fleet gorge]

Everyone get off the Assistants API so I can play with it. But seriously, there is some huge potential here. I am currently testing Dave Shapiros Sparse Priming Representations (SPR) to see if it would substitute as a kind of memory for continuous discussions through multiple sessions. I keep getting Run Failed - We're currently processing too many requests - please try again later. So I guess I'll try again later.

[cursive crown]

@rancid lark @mighty osprey same here. I have tried to reach the page with a VPN located in the US, but that did not change, still no Access to the feature...

[wind urchin]

Hey guys, do you know how to get the chat gpt 4 turbo? I really need it for coding purposes

[hallow owl]

here -> https://platform.openai.com/playground?mode=chat&model=gpt-4-1106-preview

[steep current]

Hello, I am facing an issue with the GPTs. I can use the pre-defined prompts, but when I try to send a custom prompt nothing happens. When I press the send button (upwards arrow) or the Enter (return) key, nothing happens, the text I typed doesn't not get send. Anyone facing this issue? Did you find a solution?

[wind urchin]

I checked it out but I can't find any GPT 4

[hallow owl]

look at the end of the link I sent, that's the GPT-4 Turbo model

[rancid lark]

1106 == Turbo @wind urchin

[languid token]

Anyone who was in the crowd know if they'll be giving away any more $500 credits? 🙂

[leaden canopy]

Hello,

https://chat.openai.com/create

i tried to access this site but it's says i don't have access to it. I have plus but still cannot access

[hallow owl]

it's still rolling out, be patient and it will come

[leaden canopy]

Ahh alright, thank you, is there any expected date to roll out?

[hallow owl]

next several days

[leaden canopy]

thank you

[buoyant oar]

Under what forum do we discuss building custom GPTs? https://openai.com/blog/introducing-gpts

[buoyant oar]

I have the same problem "You do not have access to this feature yet"

[hard mortar]

You do not currently have access to this feature

[dim flame]

I'm getting the same when trying to upload a very simple CSV file with the Python SDK using

file = client.files.create(
  file=open("myfile.csv", "rb"),
  purpose="assistants"
)

When I changed to using pathlib.Path (example) it worked:

file = client.files.create(
  file=Path("myfile.csv"),
  purpose="assistants"
)

[analog rune]

I might have to try the csv via the api

[dim flame]

hmm... never mind, I am still getting the error but now it's when I add the file to file_ids when creating the assistant 😿

[plucky basalt]

Dall E 3 not working today in gpt4.

ChatGPT
I'm sorry, but I can't provide images. If you have an existing image that you need to analyze or modify, I can certainly help with that. Please upload the image and let me know what you need assistance with.

[onyx badge]

Working fine for me

[plucky basalt]

its not working on desktop

on mobile app its fine.
also I have received this:
ChatGPT
I don't have the capability to browse the internet or access real-time updates from external databases or websites. However, I can provide synthesized insights based on my last update, which includes knowledge up to April 2023. For the most recent studies and developments in CRM and CLV beyond that point, I would recommend accessing databases such as PubMed, IEEE Xplore, or business databases like EBSCOhost Business Source Premier, as well as publications from reputable consultancies or market research firms that specialize in CRM technologies and strategies.

[pearl flower]

For me the Chat GPTs (the ones from the explore section) don't work on desktop right now. Can't send messages to them, the send button or enter key don't work. Works fine on mobile, but not on desktop.

[random night]

The same. Need to wait untill they fix all the problems after update

[plucky basalt]

advance data analysis is not working also

[plucky basalt]

For me also the first 2 are not working. others not checked yet

[charred parcel]

Guys will the token limit for chatgpt-4-T be the same as the token limit for the normal chatgpt-4 we had until now?

[eternal glen]

Hi, I've run into a problem with the GPTs. I'm able to use the prompts that are already set up, but when I attempt to input my own prompt and submit it, there's no response. Clicking the send button or hitting the Enter key doesn't seem to work—the text I entered just doesn't go through. Is anyone else experiencing this? Has anyone figured out a fix?

[remote plank]

Hey guys, I was trying to make a "Assistant" that uses GPT-3.5-Turbo with Code-Interpreter and File Retrival. It says I can upload zip files but it an error when I try to save it.

Error:

Failed to create assistant: UserError: Failed to index file: Unsupported file file-WyZxtnv1iZjS3bOKPPI3E08P type: application/zip

Does it actually support zip files or was the does it just say it does and maybe they forgot to remove it?

Edit: Solved

[icy radish]

Same exact thing is happening for me

[gilded depot]

I was checking GPTs creation and turns out you can only upload up to 10 files in the Knowledge section, and also knowledge in GPTs seems to be only up to January of 2022. That’s not really that impressive. Am I doing something wrong, or is this feature that limited?

[austere pine]

same here

[queen lotus]

will the creation and using of GPTs be roled out to free users at some points as well?

[honest lodge]

yes but dont hold your breath

[jagged pulsar]

any idea how much information we can upload into a custom GPT with the editor? I don't have access yet.

[thorn oriole]

I don't think anyone is doing that, and it might be not allowed by the TOS and #server-rules

[keen trench]

any idea when gpt 4 turbo will be available to plus users?

or did it realease only in the api right now?

[heavy mesa]

I'm a Plus member in the US and don't have access to https://chat.openai.com/create is that a bug or slow rollout?

[stark hollow]

I am Kayode, a prompt engineer. It's nice to be here.

[forest phoenix]

loving the new update i have to say

[wind urchin]

Guys, will that custom gpts be available for free users too?

[forest phoenix]

no mate only plus

well worth it in my opinion. Amount of time it saves is well worth the money. Deffo gonna move into AI for my tech stack going forward

[grizzled flax]

@wind urchin Not necessarily, check release notes "We plan to offer GPTs to more users soon." https://help.openai.com/en/articles/6825453-chatgpt-release-notes
No indication about what this might look like though

[wind urchin]

Oh ok

[robust light]

Are they increasing the limit for pro users?

[odd verge]

Tired of developing GPTs? Come take a break and play some Dungeons and Dragaons with the Dungeon Master: https://chat.openai.com/g/g-cieyjUKku-the-dungeon-master

[onyx badge]

Wdym tired of developing them most plus users can't even make them lmao

But that's neat

[odd verge]

lol, thank you!

I've been having a lot of fun with it. Not to say it works perfectly, it's definitely a work in progress, but it's been fun.

[odd verge]

I gotta say, they don't seem to retain the instructions I give it.

[warm elk]

wheeeeeen everyone gets the all in one 😄

this time it feels like i get it as last

[grizzled flax]

You are not last! The rollout is occurring over several days--we're gonna make it 🙂

[craggy smelt]

Your access to custom GPTs isn’t ready yet. We’re rolling this feature out over the coming days. Check back soon.

🥲

[dense veldt]

I want GPT’s accesss!!!

[proven raven]

shush

[trim flame]

how many messages is the count limit as of right now ?

it used to be 50 every 3 hrs, and now ?

[unreal lance]

Any idea when the GPT "App Marketplace" will be launching? Or when Revenue Share starts (and details on amounts)?

[grizzled flax]

No details yet! Check out FAQs, GPTs ones are near the bottom: https://help.openai.com/en/collections/3742473-chatgpt

[dense veldt]

Someone plz make a JavaScript agent

[unreal lance]

Thanks!

[cinder crag]

Hi what did the new update do?

[onyx badge]

A lot of stuff

GPT-4 turbo, which is basically GPT-4 on steroids

Faster, larger context window than Claude, and smarter!

It's also cheaper for API users

Also it's REALLY fast

[cinder crag]

are the limets gone?

Also did the Web browsing up to date now

[onyx badge]

50 messages every 3 hours

Might be higher idk

[onyx badge]

Yeah web browsing is back

[cinder crag]

It was always there

I ment is it now up to date like better info and such

[onyx badge]

It still uses Bing

So not really

It's been a bit hit and miss but works good enough

Oh, and did I say that GPT 4-turbo has a knowledge cutoff date of April 2023?

[eager shell]

devs it would be nice to make chat gpt work without Google services, my Huawei P40 doesnt come with them

,😪

[onyx badge]

just sideload it

[upper mural]

ha, it doesn't

yet it does

https://chat.openai.com/share/596de25a-2684-4d45-b20f-36e968843518

but if i open an already open bing chat from before this change i can browse with bing

[onyx badge]

Wdym

It uses Bing for searching

[upper mural]

please open the chat i just did...

[mint oak]

So I'm not sure if this is the chat for it, but this browser extension(ChatGPT Folders & GPT-3.5/4 AI Online Copilot) breaks chat gpt and makes it unusable

Chat gpt gives me "Application error: a client-side exception has occurred (see the browser console for more information)."

[trim solstice]

What's the difference between GPT-4 and gpt-4-1106-preview?

[light isle]

I think the preview is for the turbo model, but I could be wrong

[trim solstice]

ok

[neat raven]

no, I can't even start yet

[grizzled summit]

rip chatgpt

2022 - 2023

[quartz fjord]

Lol 😂 good joke

We are here to stay!

[verbal lantern]

where is the GPT store? I see the GPTs made by openAI but not the general store

[static isle]

That's coming out later

[inland hound]

what's wrong with my account,it seems unable to search with bing and create images after the latest release?

puzzled

[timid plank]

hi, i ve seen that when using gpt4 i can select the DALL-E. is that included with the plus or it consumes credits?

[uncut magnet]

Hi guys, does anyone know how to get access to custom GPTs?

[bright flint]

use the discord search feature. Ur question been asked 100 times today

[uncut magnet]

I did, but I didn't get the information. I saw an article that said that plus members can try at https://chat.openai.com/gpts/editor but I clicked there and it said I don't have access. And couldn't find anything else helpfully

[bright flint]

The access is random and available mostly to developers. You just have to wait for access. This is how its been for the many past updates. Just gotta wait

[uncut magnet]

ah okay thank you 👌

[bright flint]

you can still access new gpts people made so far in the custom-gpts channel all the way in the top

[uncut magnet]

It's not possible for me yet. I tried from the links of others posted on custom-gpts channel but it still said my access isn't ready yet

[bright flint]

interesting. I cant make custom ones myself but i can access other peoples. Ether ur UI isnt updated to the new one or ur thingy is just acting weird

[wind urchin]

[ashen hollow]

is there a timeline for the global rollout of custom gpts?

[hallow owl]

over this week

[ashen hollow]

thank you !!

[grizzled pagoda]

I'm in the same boat. I just upgraded to Plus to be able to make and use GPTs but I can't do any. Are you a Plus user as well?

[grizzled flax]

The rollout is still in progress, updates will get to us soon!

[grizzled pagoda]

thank you! can't wait!

[vivid kestrel]

Hello, I have a question about the new Assistant API: I found there is no temperature setting. Is it by design?

[olive belfry]

Question regarding GPTS knowledge upload, is there a token limit for that?

I would like to upload several documents and i don't think the context limit even of 120k/140k context would be enough, so i was wondering if it counts towards token limits of sorts or if there is a seperate limit for that etc.

[faint vortex]

I found this here: https://platform.openai.com/docs/assistants/how-it-works/creating-assistants#:~:text=You can attach a maximum of 20 files per Assistant%2C and they can be at most 512 MB each. In addition%2C the size of all the files uploaded by your organization should not exceed 100GB. You can request an increase in this storage limit using our help center.

[olive belfry]

That sounds promising if that's true.

[safe venture]

I had some good experiences yesterday uploading source code and asking it to go through them all to tell me what the overall system did (and conduct a peer review)

it was pretty time consuming though, you'll definitely want a good front end for it

[faint vortex]

Agreed

[tawdry charm]

I am actually confused and need some help. I am currently setting up a GPT, so far so good. But the Action window seems a bit weird to me. Why is it advertised that GPT can interact with external APIs, there is even a window to enter API keys, but the GPT assistant tells me that foreign APIs can not be used?

[faint vortex]

I also found that merging multiple PDFs (with good structure and formatting IMPORTANT) into one file can help as well. Of course while keeping in mind the constraints

[olive belfry]

Welp this is fun, assistant is bugged can't use it, cors errors etc. > Run failed
Sorry, something went wrong.

[tawdry charm]

My GPT builder broke after I asked it about Actions.

[heavy lava]

not just yours man

I think its for everyone

[tawdry charm]

Rip

uff. Yea the GPT needs some work honestly. There is also not enough info about how to write an Action anyway. Some of the stuff is just never explained, which is probably not a big issue if you are familiar with this kind of stuff, but if it is supposed to be accessible for everybody, you need to explain the basics.

[daring pine]

GPT4 isn't working at all for me for like the last hour, just errors out every time i ask it a question. GPT3.5 too

[grizzled summit]

https://status.openai.com/

[fierce tundra]

Hey, anyone else has seen his GPTs disappear overnight ?

[daring pine]

thanks, i'll keep an eye on this

[grizzled summit]

👍

[wind urchin]

Gpt appears to be down

[hallow owl]

There is currently a major outage affecting ChatGPT and the API. Use this time to take a break and refer to https://status.openai.com for updates.

[atomic robin]

really?

[wind urchin]

I've noticed a fair bit of talk regarding degraded performance and response quality using the new GPT-4 Turbo model. Has anyone published any benchmarks (reasoning tests etc) that compare gpt-4-1106-preview with gpt-4?

[eager sparrow]

There is a reason why it's "preview"

[wind urchin]

?

[eager sparrow]

You can't benchmark it yet at this stage

Wait next month to get the final model

[wind urchin]

Could you expand on that? I don't think I understand how the preview model can't be benchmarked. If it's available now, then surely it can be tested like all the others?

[eager sparrow]

You can but I don't see the point to do it on temporary model

[atomic robin]

hard to benchmark a model if you are limited to 100 requests a day

[remote garden]

fixed by now???

[warped bear]

no

[crystal solstice]

Fixed for me

[wind urchin]

Ah, that clarifies it, the rate limit had me missing why there were only early benchmarks.

[atomic robin]

that doesn't make sense

[wind urchin]

The rate limit (my not knowing of it) is why I couldn't find benchmarks at this stage. I could only find preliminary benchmarks. Your mentioning of the rate limit as an obstacle to benchmarking makes sense.

[zealous gull]

Not sure where this should go as far as channel wise, But when trying to create a Assistant, the model box for me is blank. Where can I get these Models? Please @ tag me so I dont miss the convo

[lethal stump]

Will we have the ability to "transfer" GPTs from one account to another? I've been making some tied to my business account but I'd like them to be listed under my personal account. Will there be "ownership" settings?

[drifting cove]

I’ve been using ChatGPT as a calculus tutor. With GPTs would i basically be able to make a “version” of chatgpt that is specifically for that?

[exotic turret]

Note that one of the OpenAI provided GPTs is Math Tutor. I have not yet used it but possibly this already does what you want

[exotic shell]

How do I get access to build GPTs? I have paid plus account AND I'm building with the API. Someone help 🙏

[hybrid mist]

Am I the only one who gets pissed at other peers using ChatGPT for work? This dude has been doing it since he’s came to the school and they still let it slide. They just know his work is artificial to intelligence. I use it solely for creative purposes.

[loud pilot]

have anyone tried gpt model as plagiarism checker ?

[zealous gull]

I figured out the issue, no everyone has the ability yet to make custom gpts, so I have to wait until i have the ability/version to do so

[mellow trench]

Hello, I wanted to know if, as a subscriber to the ChatGPT Plus residing in Europe, I will be able to use ChatGPT 4 Turbo in the coming weeks, or if I need to pay anything extra. Is there an indicative date for when ChatGPT 4 Turbo will be released in Europe? Thank you.

[half cargo]

probably they will add it pretty soon,since they have added GPTs 1h ago!

[grizzled flax]

The rollout is still ongoing, not all have access yet!

[mellow trench]

Thank you very much Azashiro

[grizzled flax]

Plus will give access to the Turbo 4 model yes, the rollout is not yet complete. Confirmed at 19min mark of keynote on Monday

[wraith ginkgo]

`` Server Tip: ``

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

[half cargo]

It is true that gpt 4 turbo can only produce 4k output?

[mellow trench]

Thank you Solbus

[wind urchin]

Of course

"Rate limit reached for gpt-4-1106-preview in organization org-[redacted] on tokens per min. Limit: 10000 / min. Please try again in 6ms. Visit https://platform.openai.com/account/rate-limits to learn more."

I don't understand why it says "6ms" and doesn't work after several seconds/minutes. Also, I don't reach the rate limit

[spiral vault]

WHY THO how do they expect to say 120k and have this

[weary field]

messing with temp = 0 and diff seeds. if diff seeds produce diff outputs, but the answer is supposed to be deterministic, am i supposed to just mess w seeds until finding a viable one?

[grave vale]

Is dalle3 down? ChatGPT is now telling me it's unable to create images

If I pull out the dalle3 specific chatbot it actually works

[golden tusk]

Anyone know the maximum amount of text a document can have for it to be used for a GPT?

[sterile pier]

I tried implementing a 29 page pdf into a GPT's knowledge and it didn't let it upload.

[golden tusk]

I made the file size of the pdf under 400kb and it worked

[cold trout]

I think its just glitching. I tried to upload a 2mb doc but it wouldn't let me yesterday, then today it let me. Now its down so i cant test right now, but yeah. The doc was pretty big

[weary patio]

Probably one of these 400m+ request per second DDoS attacks Google posted about

[swift sage]

My ChatGPT just switched to GPT 4 Turbo and instantly became USELESS. GPT 3.5 quality at best

[noble terrace]

useless why?

[vale saddle]

When I search through this chat and others I see that the rollout of GPTs and other features are still in progress; is their an estimated time which this will be accomplished? Super eager to get my hands on the GPT editor - thank you 🙂

[grizzled flax]

Only language we've had is "in the coming days" and "as fast as possible" 🙂 We'll get there!!

[vale saddle]

Thank you!

[steel spade]

Anyone in here using the Playground and found a way to bring up an old conversation with an assistant? I closed the tab earlier because I thought we could access chat history with them but came back and can't find it anywhere and that was so much information I needed. lol

[swift sage]

I give the old GPT-4 a piece of code and ask to rewrite it, and give me back the full new code, it'll do it (still works on Playground or Poe). But with the same piece of code, the new ChatGPT starts giving me "patches", starts forgetting, digressing. Like Claude-2. With old GPT-4, I could describe it: it is LISTENING to me. With the new Turbo it's: it PRETENDS to listen but actually wants to get rid of me asap. Like a service person at a shop.

[weary patio]

Just give it some time

[swift sage]

The new ChatGPT has been giving me very poor suggestions, got confused, truncated essential parts of tht code I asked it to write.

[weary patio]

Few days

[swift sage]

No, that's not how it works 🙂

The GPT-4 Turbo model has been distilled from big GPT-4, and it's now here. It won't change "over the next few days".

[weary patio]

Are you sure?

Because everyone is currently having problems right now.

[steel spade]

Man it's all over reddit and other forums. Everyone is having this exact same issue, especially with code. @swift sage is not unique in their experience

It's pretty much assumed that Turbo is sacrificing quality for speed

The overwhelming amount of user experiences being reported really shore up that assumption

[weary patio]

Are you sure it's not a GPU issue of the servers? Because the dev team have said multiple times that they're experience overloading GPUs due to the heavy release. I've also bad similar experiences with GPT4 giving me lackluster results on reportedly slow days.

Could be a patience thing, you know? Since like everyone since release has had a number of problems relating to an array of different things.

I could be wrong though. But please message me back and prove me wrong after a week and a half and I'll tell you that you were right and I was wrong.

[swift sage]

I don't mind having GPT-4 Turbo, because it's an alternative. But it's definitely not a replacement for GPT-4. I understand that OpenAI replaced GPT-3.5 with the distilled GPT-3.5-Turbo, which was much faster, much cheaper to run, and WORSE. But that was OK because we also had GPT-4! Now, GPT-4-Turbo is replacing GPT-4 in places where there is no higher-quality replacement, like in ChatGPT.

[weary patio]

Maybe GPT5 is coming out and GPT4 Turbo will be the new GPT 3.5 Turbo

[swift sage]

But if ChatGPT truly switches from GPT-4 to GPT-4-Turbo completely, then I don't see why I should use ChatGPT anymore.

[weary patio]

As opposed to what alternative?

[swift sage]

I think it's a bit cheeky to do this "turbo" naming. Because it's not just FASTER. It's faster because it's more sparse, distilled, quantized — effectively smaller. I don't mind having the faster smaller worse version as an option — but kicking out the real GPT-4 out of ChatGPT is terrible!

[steel spade]

I can't prove anything as it is all based on user experience reports across the internet. This has been ongoing for about 2 - 3 weeks now.

It's not since dev day

[steel spade]

Agreed

[weary patio]

GPT4 Turbo has been out for 2 to 3 weeks now?

[steel spade]

THere ya go

Hence the performance drop lol that's the point I am making

It's faster at responding but the quality is way down

THat is the consensus

[weary patio]

I think GPTs may have the solution you guys are looking for.

[steel spade]

GPTs is just custom instructions for separate instances

It's still Turbo, right?

[weary patio]

I believe so.

But if you find a good alternative that performs better than GPT4 Turbo, please let me know.

[wraith ginkgo]

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.

[steel spade]

GPt 4.... pre turbo lol that's the point

[weary patio]

But isn't turbo replacing it?

[steel spade]

I'mn not saying it is worse than... Bard... lol but compared to what was being received as output 3 weeks ago it has changed substantially. As paying customers it is 100% reasonable for that to be an issue for us

I hope on the API it wont

[weary patio]

How was GPT4 Turbo released 3 weeks ago?

[steel spade]

Dude lol

No but many people (me included) started getting told by GPT (up to three weeks ago for many) that the data cut off date was April 2023. We weren't sure why because we didn't know it had changed... until dev day

So, it looks like they were testing it out here and there with people or something?

Not really sure. Not official info from OpenAI

[weary patio]

That doesn't mean its GPT4 Turbo

That's just the knowledge update

[steel spade]

But GPT is still 2022. GPT4 Turbo is the April 2023 cutoff

[swift sage]

Well, for now I'm gonna use Poe because that c still uses the older models, and it actually has given me access to GPT-4-32k for months now

[steel spade]

Then why are many people still getting the old date as the data cutoff?

and were over those weeks?

[weary patio]

They're probably in different countries where the knowledge cutoff is not the same.

[swift sage]

I don't yet have the unified tool switching in ChatGPT but the models have switched today

[steel spade]

I don't really have the time to keep arguing this or whatever you are trying to do. I am going based on reports from a large amount of different users that were all confused about what was going on with GPT on their end, especially ChatGPT.

I got two different dates in that time period lol

I got April 2023 and then it went back to the older date about 5 days later

and still is there

[swift sage]

I had old slow good GPT-4 in ChatGPT in the morning, then it went down and 40 minutes later it was up again

"faster" but much worsr

[steel spade]

Sucks man. Hopefully it will flip back around or we can pick models for 4

yes exactly. Same issue that people have been having for a few weeks

y'all have a good night

i'm heading to bed!

[swift sage]

I queried the know cutoff 3 days ago in Chat

[weary patio]

Have a good night. And be patient.

[swift sage]

all versions EXCEPT Bing said they're April 2023

[weary patio]

Don't jump to conclusions until it's set in stone.

[swift sage]

But Bing said Sep 2021

I understand it's a huge rollout etc., massive. But so far I am unimpressed

[weary patio]

I'm impressed

By the GPTs

[swift sage]

In the API we had cheap dumb 3.5 and slow great 4

I just hope we won't end up just having dumbest 3.5 and dumb 4 Turbo as the only options

[weary patio]

Imagine if we do

What will you do then????

Probably still stick with it.

[swift sage]

Well, I'll certainly resume my fine-tuning of Mistral and I'll be using Claude more, and maybe even Cohere

[weary patio]

C'mon, you know that's not true.

[swift sage]

And even Falcon which many people have ignored, but it's actually pretty cool

Well, GPT 4 Turbo is definitely not a conversation partner on my level

talking to it is like talking to a drunk guy at a bar 🙂

[timid flower]

Is... Everyone else's dalle3 only spitting out a single image at a time? I SWEAR I used to be able to do 4 at a time....

[weary patio]

Adam, you're gonna stick with OpenAI for the rest of your life.

I'm sorry to say it.

[swift sage]

GPT4: I know they'll keep the old models for some time. But if they're not in ChatGPT, Then I'll surely no longer use ChatGPT that much

[weary patio]

Yes, everyone has been having the same issue

[swift sage]

I'll use some other way, with API

Dall-E: yes, 2 weeks ago we could actually get the prompt texts, and 4 images but

I don't think that's economical fit them, so now it got down to 2 or sometimes 1

[timid flower]

Also, anyone able to do the custom GPTs yet? I clicked on the link on openais website and got an error

[weary patio]

Soon it's gonna be half an image

[swift sage]

There's still some hope with Microsoft 🙂

They don't take moderation so seriously

[weary patio]

Not everyone yet. Still rolling out.

[timid flower]

Ahhhh. Everyone from the "company" is at the conference, so the AI left to tend to everything is burning everything down 🙃

[swift sage]

You can get Bing Create to make images with Dall-E 3 that ChatGPT will definitely block

because MS wants everyone to come to Bing

[weary patio]

Which that will never happen

Because OpenAI already has Bing incorporated into this next GPT4

Plus all tools

[timid flower]

I mean, does bing still spit out 4 images per prompt?

[swift sage]

But fortunately, Dall-E 3 is not all that great

[swift sage]

it varies

and Bing only does square so far

[weary patio]

It's mostly just two now

[swift sage]

There's a distinction: you have Dall-E as a model, for which AFAIK there is no really DIRECT access

[timid flower]

Idk... I'm able to get images that look a lot closer to what I have in mind with dalle than MJ. MJ is like a complete crapshoot of prompt engineering

[weary patio]

It's not ChatGPT

[swift sage]

and then you have the frontends, where Bing Create is a thinner frontend with laxer moderstion

and ChatGPT is another frontend combined with GPT-4

[weary patio]

It's its own AI

MJ

[swift sage]

basically ChatGPT with DallE, you talk to ChatGPT and then it talks to DallE

But yes, it's weeks since I've seen 4 images made by DallE from one interaction

both with Bing and with ChatG

[weary patio]

PT

[swift sage]

But I don't really care that much for DallE. I mean, I'll probably never really use its output directly

[weary patio]

You won't?

[swift sage]

So far I've been freestyle-prompting DallE to give me some images

which I then used as image references in Midjourney or in Stable Diffusion

[weary patio]

Sounds like you really like AI

You're gonna be here forever

[timid flower]

Oof yeah, that's the biggest downside to dalle. You can't input images.
... Which is REALLY stupid, cuz you can input images into regular chat

[swift sage]

the DallE images are OK as "sketches" but out of Stable Diffusion I can get really high-quality images with image reference

[weary patio]

SD is not very good is it?

[swift sage]

But that's fine. Basically ChatGPT+DallE is a decent 1024x1024 sketch generator, while Stable Diffusion is the final renderer that I can get to output 16kx16k 🙂 Which part is the creative part?

SD is great on micro level but not when it comes to prompt-to-conposition