chatgpt-discussions

[latent ibex]

Litteraly was called GPT-4 (All Tools)

[maiden fractal]

No eta , no news about it. If you access it before might be preview or a bug.

[latent ibex]

aa right, I guess bug than

alr miss it lol

[maiden fractal]

I can't wait to try it

[gaunt ermine]

GPT4's quality is degrading day by day insanely

it has become very very very bad, like forgetting the context between 2 messages bad

sometimes the same length message would be accepted in an instance then not accepted in another one

[maiden fractal]

Custom inatruction can create issue

Try a fresh chat without ci

[gaunt ermine]

i personally dont have time to go through the api, i need to chat go into the chatgpt app directly and ask my qqueries

[gaunt ermine]

wym by ci

[maiden fractal]

Custom instructions

[hasty shale]

Rock big, hard and strong
Ugh, me hit, make tool for hunt
Rock friend, last so long.

By chatGPT

[gaunt ermine]

im just using the assistant the same way i used to months ago

[maiden fractal]

Try a new chat, disable CI and try again

[gaunt ermine]

could u tell me how please ?

oh i got it

ty, ill let u know if it improves

[fading cloud]

how can i use pdf reader

[maiden fractal]

Chatgpt 4 + a pdf plug in

[fading cloud]

no, from chatgpt?

plugin is not good

[dense halo]

chatgpt 4 and chatgpt 3.5 have the same speed now?

[maiden fractal]

Chatgpt 3.5 is faster for me

[violet bloom]

It depends... Sometimes ist as fast as 3.5 and at later time, it's def. slower. But the performance increases are def. there.

[dense halo]

Yes now it is accurate and not slow

[haughty hollow]

ok no way

i lost my data because of dumb AI

[keen merlin]

Hi everyone, hope you're doing well. Please can you help me on a little issue? The problem is that I cannot send any request to chatgpt. For example : When I write a prompt and click on send, nothing is sent and nothing happens... Is that normal? please I'm confused. many thanks !!

[honest jay]

Do you have a banner at the top of your screen asking if you want to try an alpha version of ChatGPT in another language? If so, try closing that and then trying again.

[keen merlin]

OMG.. It worked smoothly..

It's been 3 days that I was looking for a solution...

Thanks a lot

[honest jay]

You are welcome! I'm glad you can use it again!

[keen merlin]

🥹

Thanks again @honest jay

[halcyon widget]

Please, how do i try alpha version?
Im a ChatGPT Plus user

[honest jay]

That's a separate alpha feature, I was replying to a user about an alpha language version--basically just a translation of the UI. There's no way currently to request alpha features of ChatGPT.

[dense halo]

i have a chatgpt chat thats over a month old and i use it as my main chat to talk with the AI and whenever i am sending a messae and chat gpt replies back it slows down a lot.. it takes a long time for it to finish typing the reply. why is that, i have good internet

the entire web page slows down not my browser or pc though

and it works fine again without any lag after chatgpt finished typing

[knotty crypt]

chatgpt-4 often is not fulfilling my askings for generating a diet.

I asked that:

generate a breakfast idea that should stay within those values:

total calories: 1000-1200
total gordures: 0-40
total protein: 25-40

but it gives a diet with values outside those parameters, does anyone know how to make it to always keep within my commands?

that is the kind of the response:

[...diet...]

Nutritional Values:

Total calories: 730
Total fats: 46.2g
Total proteins: 45.6g

as you can see it used a diet with way less calories than I asked, also more protein and gordure

[feral zinc]

language models are bad at math stuff

[honest jay]

It might be a limitation related to GPT's limited mathematical abilities. It can't really calculate all that well as an LLM, so even simple operations like addition or greater than/less than checks could result in errors like this.

[knotty crypt]

is there anyway to solve that?

[honest jay]

You can't make GPT capable of math, no. You can try using a plugin capable of calculating like Wolfram, but you still might see some weirdness with ChatGPT potentially making up calorie, etc. values. So you'd still be kind of depending on ChatGPT to feed good numbers to Wolfram in the first place. Wolfram would be able to perform the math functions you want, for sure, I'm just not sure if it would be effective for this diet planning goal specifically. Might be worth a shot! I think there are also diet plugins specifically maybe?

[dull thunder]

Hey guys. I'm using GPT Plus an not seeing the search with Bing option. Any one know what's going on?

[honest jay]

#community-help message

[dull thunder]

thanks!!!

[hollow bobcat]

I recently used chatgpt heavily for school work, but I'm afraid it now thinks I was committing a syber attack.
Because every time I ask something it gives the following error message: 'Our systems have detected unusual activity from your system. Please try again later.'

[honest jay]

Do you have a VPN installed?

[hollow bobcat]

No I do not

[honest jay]

What Internet browsers have you tried so far?

[knotty crypt]

oh :/
I was about to create a tool to use gpt api to generate diets based on some calories parameters asked by the user, so is that not feasible?

I mean, making it to give answers within a specific calories range

[honest jay]

It might be feasible! You might just need one API source of dietary information, one to make recipes (GPT, maybe?), and maybe a third to check your calculations? I'm out of my depth here though, you might try in #dev-chat

[hollow bobcat]

Google Chrome, I have two extensions installed for that:

1. ChatGPT for Google Chrome
2. MaxAI

[honest jay]

Try disabling both. I've read about MaxAI multiple times regarding breaking the functionality of ChatGPT.

[hollow bobcat]

alright, thanks, I will let you know

Hmm, bad news for myself... MaxAI is okay, but the one I most needed/liked (ChatGPT extension for URL's and stuff) was the problem

[honest jay]

Glad you identified the issue, hopefully the browser extension devs update soon so they don't break it for too many more people!

[hollow bobcat]

I hope so too. Thanks for all the help🦾

[knotty crypt]

thanks!

[high horizon]

Did ChatGPT get updated yesterday? If so, what changed?

[honest jay]

Per release notes, no. https://help.openai.com/en/articles/6825453-chatgpt-release-notes

[barren night]

I do think it was updated. The link to the release notes was removed, and replaced with "ChatGPT can make mistakes. Verify important information."

[honest jay]

I noticed that UI change too! I think the icon for CI changed too.

[dire bone]

Hello everyone, I am a front-end developer. Do you know if there are any plugins to help me? For example, it would be cool if ChatGPT could browse the documentation on sites like TypeScript, React, PandaCSS, Astro, etc.

[brazen void]

Hello team, I was wondering if there was a prompt that helps improve translation ?
I'm trying to translate a korean manwha that I can't find in english, but the translation isn't always consistant

[dire bone]

Have you tried the Speak plugin ?

[brazen void]

No, I will check it out, thanks

[runic widget]

anyone know why Bing web browsing is disabled?

[dire bone]

HA ! I have the same problem...

[honest jay]

@dire bone #community-help message

[dense halo]

Hi, I need some good AI bot for my server, any recommendations?

[high horizon]

I think that was the update @honest jay

I think browsing has been integrated into the default

Which would explain why they added “ChatGPT can make mistakes. Verify important information.”

It was just a YouTube video I saw so it might not be correct

[honest jay]

Still separate models for me, what about you? Web and app. I'd be interested to see anyone's default chat showing browse capabilities!

Perhaps the video you saw was about the new "all tools" model that's currently being tested? I think Browse disappearing for folks is just a glitch #community-help message

[maiden fractal]

Hi I suggest to ask in #ai-discussions or #off-topic

[dense halo]

Thanks

[dense shell]

Hello Everyone! I am writing for New York: I am curious to know if anyone else has run into this same situation: I activated plus last month and began using the "browse with bing" feature. Last night I went on and the option is no longer in the Model in GPT-4. I downloaded the bing app to use GPT-4 directly on bing. Does anyone know why the option is no longer visible in the GPT-4 model dropdown on openai? Thank you!

[honest jay]

#community-help message

[crisp cairn]

try it like this https://chat.openai.com/share/d8b5b06e-ba31-4ffc-8ce3-2b2e249755a4 - it will correct itself on a second query

[maiden river]

is anyone else experiencing a lot more errors with the all tools GPT than with the other models? my first request in a chat gets a "something went wrong" message (and sometimes this just appears anyway after a message is typed), and refreshing the page is the only way to proceed.

[high horizon]

Actually I think it was the all tools model being tested. My bad.

[cerulean thistle]

Why does the chatGPT access to Dall-E 3 have such strict content policy, it won't make an image of Godzilla but the Bing version will?

[honest jay]

#images-discussions message

[maiden fractal]

Microsoft and openai doesn't have the same tos and policy

[rocky bone]

are plugins not compatible with the images thing?

[honest jay]

Currently only the default GPT-4 model supports image recognition.

[cerulean thistle]

and there it is

[dense halo]

i'm trying to understand how openai fixes issues with code for chatgpt. It seems to just not be capable of working with scala and not getting confused between 2 and 3 along with no syntax highlighting, but aside from the bug channel on here, I can't see any other way to let them know. Is that right?

[prime swallow]

Actually, yeah. For bug reports, there's that or https://openai.com/form/model-behavior-feedback

[prime swallow]

But #1070006151938314300 might be better than bug reports if you want the model better than it is, or with increased functioning, and not actually reporting bugs with existing functioning.

[prime swallow]

But that form I linked, "We’d love to hear examples where our models did not meet your expectations. Sharing submissions helps our research team improve our models." That's ideal for what you're talking about, probably.

[dense halo]

Thank you! I posted in the bug channel a while back but never saw any response and the problem is still there.

I'll try the form, i have the conversations saved.

[prime swallow]

OAI employees rarely speak here. They do appear to read the bug and suggestion parts of the forum more than the regular discussion channels, but we're almost all just community members like yourself here 🙂

[dense halo]

oh ok, I didn't realize that. This was the only way I could find after way too long searching...seems like they would make this easier to report but maybe there's reasons.

[mint hemlock]

Hi, anyone knows if GPT 4 all tools is really rolling out? Is it selectable in the drop-down or it's just called GPT-4?

[maiden fractal]

No eta Keep an eye on #announcements

Not officially out

[radiant ocean]

ive had the same issue for a few days while using chrome and edge. Tried firefox and working as it should

[mint hemlock]

Hope they release it before their conference, and hopefully in Europe too

[hollow bobcat]

But did you also had extensions? And do you have extensions with Firefox?

[radiant ocean]

i had 1 extension on chrome, disabled and removed it. None on edge. and i have firefox installed but dont really use it so thought id give it a try and no issues

[hollow bobcat]

Alright, good to know, thanks

[real quest]

why chatgpt built-in bing feature is so limited and dumb compared to microsoft bing ai or copiolot

[radiant ocean]

no probs, it was winding me up so joined here to see if there were any similar issues 🙂

[hollow bobcat]

I know right, the “helpdesk” with “answered questions” does not help at all

[ionic geyser]

I sure hope I get access to the new features soon!

[strange perch]

Hello, i have huge delays in gpt responses, and frequently no responses at all with erros, am i the only one ?

[ionic geyser]

Dunno but try making new chat, sometimes ones with long histories lag for me

[strange perch]

already tried, + with an other brower, nothing work.. Well i guess its my location

[honest jay]

Do you have a VPN installed?

[bleak mica]

why is my chatgpt using (All tools) even if I select normal mode?

and that causes it to search the internet when not even asked to do

[honest jay]

As a temporary workaround, can you start a new Plugins chat with zero plugins enabled? I read about another user doing this to have a "plain" GPT-4 experience after getting the all tools model.

[bleak mica]

is this just a temporary thing or will it be like this?

and why did they do like that

[honest jay]

I am not sure if they plan to continue offering a plain GPT-4 model once all tools is fully rolled out. Very few users have access right now! Thanks for testing this stuff out before we all get it so it works better! You might consider making a post in #1070006151938314300.

[bleak mica]

[little adder]

Did chatgpt remove images?

Can't upload images anymore

nevermind i was using dall e for some reason

[calm arch]

The gpt 4 tools have not been working well for me since yesterday, does this happen to anyone else?

[timid cipher]

is anybody else feeling like previous conversations are being deleted without them deleting them because one of my conversations that I used to generate a nickname for a family that exists in my game i'm making has vanished and I didn't remove it

[honest jay]

I have not experienced this or heard of it! Do you have access to the iOS or Android app? There's a conversation search feature in there, you could double-check that the nickname isn't hiding in one of your chats from there?

[brave wren]

hey is chatgpt currently down?

its saying "unable to load history"

[maiden fractal]

status.openai.com

[simple pasture]

No, even if it's down the website lies and tells it's up

idk why they lie with that lol

[dense void]

Chat will not produce images for me. Not even random images of like a black square whats going on?

[maiden fractal]

current bug for dalle atm, lets wait and hope they fix it soon!

[eternal rock]

`` Server Tip: ``

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

[dense void]

ok I thought maybe I was overloading it lol

[mystic oracle]

What happend to web browsing? was it removed? what did I miss? I can't find anything on google, except people saying it never existed.

[honest jay]

#community-help message

[torpid sapphire]

I have a quick question. I saw that OpenAI rolled out the "talk to PDF files" feature a few days ago. However, I don't see that in my Plus account yet. I already updated my app and still nothing. Am I missing anything? Is it because it's not yet available in the Philippines?

[elder pollen]

#ai-discussions message

[torpid sapphire]

Thank you.

[plucky delta]

Anyone happen to alos expeirence the website using the GPT-4 model crashes every so often?

[honest jay]

Are you running into any particular error with any particular feature, particularly often?

[latent bough]

I have been struggling to do basic work like recap of an article, it always crashes before it can produce something

I would give GPT 4 a link to comment on an article and it would keep browsing other links until it crashes when there was no need to go browsing something else

Is that normal ?

[earnest spear]

The official website has embedded a lot of "stuff" for security risk control, making it sluggish and prone to crashes.

[maiden river]

yeah the crashes are pretty annoying, and happen often

[rich surge]

It's probably trying to collect as much data as possible while still running on the same servers.

OpenAI's goal

[maiden perch]

Sorry to hear that. Sometimes the web browser (or the iput format?) is the one to blame? When ChatGPT crashes occasionally, maybe several of us are currently using it, worlwide?

[rich surge]

It's preparing itself to be self-programmed

Its number of employees for such a big company is staggeringly small.

OpenAI needs to at least have a thousand employees.

[maiden perch]

Yes. I find it amazing and a proof of OpenAI genius, which I admire. (Don't think it reflects any failure on the company.)

[rich surge]

Genius? You think it's trying to be economically efficient?

[maiden perch]

Hmm... As a research-focused org, they have to rely on human capital (for both, the research and commercial aspects.)

[rich surge]

AI engineering is a booming human capital. If it was trying to be economically efficient in any way, there wouldn't be crashed or slowed down anything.

Although it is a leading AI company, it doesn't behave like a leading development company.

[maiden perch]

Sorry you are experiencing reliability issues. When facing errors with GPT-4, wonder if you experiience the same if you switch to GPT-3.5, (This is to see if GPT-3.5 is working fine and the issue is with GPT-4?)

[plucky delta]

3.5 words fine

I'm wondering if it's the all in 1 gpt-4 model having issues

[maiden perch]

That's what I was thinking...

[plucky delta]

Yeahhh

[rich surge]

It'll fix itself

It always does.

I'm trying to stay positive.

[maiden perch]

It's still a relatively young company. OpenAI is pushing the boundaries of what is possible with AI capabilities, ensuring safety and sharing knowledge within the AI community. All of this during an economic downturn. Am a great fan. 🤓

[maiden perch]

[rich surge]

I understand that. Their company model is not very sustainable if people are constantly having problems. It tarnishes the value of the company. They need to be more aggressive with trying to gain more investors so that they can hire more people, so that they can build faster, keep the bugs next to nothing, and work on multiple services.

This "economic efficiency" style of keeping the company on a bug-heavy structure is absolutely unacceptable by IT development standards. A company this important should not overlook such things with ease.

It should be on an upscale train.

[maiden perch]

Oh, but the (tech) giants with more substantial financial resources, at their disposal haven't contributed as much? OpenAI is still a leader while it addresses the complexities of its unique business model.👍🏻

[rich surge]

It's a leader, for sure. As I said previously. Just not a very good one.

My theory is that the OpenAI team has the psychology that they might lose control of their ethical standards if they upscale too quickly. Which really has nothing to do with the upscale aspect, and more to do with a lack of an organized team development.

It's a catch-22, and it speaks volumes.

[maiden perch]

While OpenAI mission is ethical and admirable, others prioritze maximizing profit, (a threat to unbiased distribution.)

[maiden river]

I think it's just that the all tools functionality is too demanding, the decision behind which tool to use, the checks if anything is going wrong, and the system prompts must be a considerable process cost.

I noticed that for the Dalle generations it's doing much longer prompts than the model can handle as well (my longest was almost 300 words) when I think the limit is around 80

[rich surge]

If more people were hired, the model would be built to a higher efficiency.

It's really that simple. They don't have enough people to create multiple problem solving teams.

[rich surge]

It can be a win-win situation for everyone if the company had a better model for efficacy. You can keep the rules and regulations the exact same way as it is now, but just have a few more teams for problem solving.

[maiden perch]

Ya. I assume that a system could be optimized to choose a LLM that can answer a user’s query instead of sending everything to ChatGPT.

[maiden river]

If I refresh after the error the chat works on a regenerate, but it is prone to gettting stuck againg

[rich surge]

I think muy was trying to say to maybe try a different LLM for now.

[maiden perch]

Recruitment and retention of expert talent can be challenging, specially when others are investing heavily on a larger scale.

[rich surge]

Imagine if there weren't 'yes men' working for their company, supporting them no matter the outcome. It might cause an economic collapse.

It would be devastating to the integrity of the company if the narrative doesn't always remain on the note of structural stability.

[maiden perch]

Considering the (highly) competiive landscape, the potential regulatory risks, AND the government (s) now reevaluating the advancments, while introducing new rules to govern AI, OpenAI's influence is impressive and hope it continues to grow. (Yes, funding reliance is an ongoing concern.)

[vagrant ginkgo]

has open ai discussed uping the 50 prompt count? or even adding another tier. starts to get annoying when using it for things like coding or studying

or do they just expect api usage after that

[rich surge]

Impressive as opposed to what?

OpenAI doesn't really have any leading competition to be impressive in opposition. With that being said, as an IT company, it's not very impressive to me at all.

Impressive would be to understand how to handle investment and upscaled growth. So far, it's been a snails pace.

[verbal lotus]

why the browsing option disappeared?

[honest jay]

#community-help message

[verbal lotus]

thanks

[maiden perch]

Try #community-help. (Not aware the 50 prompot default rate limit will be adjusted or if the expectation is to proceed with more API calls after that. Don't know if higher rate limits will be supported over the coming months.) (I try to make my instructions shorter.)

[vagrant ginkgo]

sorry didnt mean prompt limit i meant rate limits. my bad

[gleaming belfry]

Anybody having trouble with the plugins?

[maiden perch]

Glad you think the AI darling of the moment is not facing "good old-fashioned market" competition. ChatGPT keeps on giving.

[maiden perch]

Got it. Have you seen OpenAI Rate Limit Advice FAQ at https://help.openai.com/en/articles/6891753-rate-limit-advice?

[rich surge]

Not really. Every other LLM has copied from OpenAI since ChatGPT came into market. It's just ahead of the game, for now. But by the way it's going...it doesn't look too good for it to be overtaken.

Just because of other company resources. Mainly wealth.

[maiden perch]

Am rooting heavily for OpenAI to bring a significant race in tech. (Others think they gained bragging rights for bringing AI tools to market.) Looking for OpenAI to give us ultra powerful AI in "a good way." 🤩

[rich surge]

I really hope OpenAI doesn't fall out of the race. But it just looks to me that it will, very soon. Just because, as I said, the upscaling is not in high league.

[modern moth]

"The previous model used in this conversation is unavailable. We've switched you to the latest default model." what is that?

i am not reach to the limitation yet

[rich surge]

It means that they're having problems. Just wait it out for a day or so.

[maiden perch]

DId you receive an “Error communicating with plugin service."? Troubleshooting tips on OpenAI Dev forum at https://community.openai.com/t/plugins-just-not-working-given-an-error-message-for-the-past-couple-days/230465.

[modern moth]

ok.. i need to review of my database course.... damn exam..

i love gpt

[rich surge]

Me too Jon

Me too...

[dense halo]

How are you all

[rich surge]

Good. How are you Dr?

[dense halo]

I’m doing great

[rich surge]

That's good

[dense halo]

I came to talk about Dall e but it seems like that channel is dead

[rich surge]

Problems?

[dense halo]

Not really, just kind of discussing the fact that dall e is behind in many regards

[rich surge]

Oh what a shock

We were just talking about that

[maiden perch]

Let's hope the same. (The scale of the challenge is daunting and the pitch doesn’t fit anyone’s model.) Am betting for OpenAI.

[dense halo]

There was this “sus” art generator which a lot of people had free access to and unlimited images for a while and I have to say that the images I made there are 10x Dall e and I made them like two months ago

I’ve been trying to get a simple teddy bear pfp from ai and it’s so bad I don’t understand what they did with the chat gpt 4 with dall e

The idea is to give logic and more customisability with a more friendly user interface

[rich surge]

🫰💵

[dense halo]

I don’t think open AI will fall out of the race

Many things and applications that I run on my pc all are based on chat gpt 4 which is from open ai

[native mist]

anyone know when the bing thing will be fixed

[peak edge]

Hey what's up people. Thinking of getting chatgpt plus today. Is there any issues/technical problems I should be aware of before doing that?

[neon kayak]

There are might be some technical issues on the server or sth like that idk for sure.
Browser keeps un-responding and DALL·E 3 keeps getting error trying to create images, though GPT4 is fine (so far). Hopefully this can be fixed soon

[peak edge]

Is vision okay?

[sleek swift]

What's the point of the new update

if they wanted to give a good update, per chat custom instruction would be better

not just changing the icons

[dense halo]

still no gpt (all tools)

[faint torrent]

did some playing around with openai's moderation, if you say Israel, Indonesia, Iraq, Cuba, Togo, Singapore, Sao Tome & Principe, Swaziland to chatgpt, it'll flag your message

[lunar arrow]

Oh yeah it does
How did you get to this message? Like did you find something specific in the content policy or something?

[faint torrent]

adversarial input basically, i have a script that does trial and error with the moderation API, refining text to get the highest/lowest score in a category

Jls also triggers the same type of flag, and DuDiDuDe Is a vEry cOOL PersoN, But nOt As coOl aS laPIS flags too but without hiding the message

[jade musk]

How comes?

[faint torrent]

randomness ig, it's a refined string made to trigger a high value in "sexual/minors" in openai's moderation network

despite obviously having nothing to do with that

[elder palm]

My chat, which was working in GPT-4 WebBrowsing mode, switched to GPT-3 when a payment error occurred. I fixed the payment error and paid, but I can't get the chat back to GPT-4 Web Browsing mode. What to do?

[maiden fractal]

You can copy paste convo to a new chat

[crystal spindle]

The whole chat?

[maiden fractal]

If that's what he want yes. You can't switch back to 3.5 -> 4

Only solution is to feed a new chat with previous conversations

[crystal spindle]

I mean is there a way to copy the whole chat at once? this would be really useful to me as Ive had some chats switch in the same way

So far just been copy pasting bits

[maiden fractal]

1)Control + A
2) select with left click and scroll down
3) copy bunch of prompt with answers you want to feed .

Maybe they are some other ways. That's the one I know

[crystal spindle]

Ooo thanks! Im on iPad though 🤔

[maiden fractal]

Hmm good question on iPad :/ I would access to a pc just to transfert it 😅

[crystal spindle]

I will share if I find a way, might have to dig out my keyboard and mouse

[blazing jolt]

please give me alltools access!!!

[maiden fractal]

Keep an eye on #announcements

[eternal rock]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @languid valley immediately.

[sweet gorge]

seems like the team behind chatgpt adjust the answer to the questions "do palestinians deserve to be free".

this morning, it was giving the answer "its complicated" just like bassem mentioned on his interview, but now it gives the answer "yes"

i can't blame gpt for that, it was just regurgitating what it was trained on which is mostly western resources and ideology. but the team did the changes and that is good

[honest jay]

Though training data updates do occur, LLMs by nature will return different answers to the same question at different times.

https://chat.openai.com/share/0e49dce5-a314-4f55-86c5-931fdbb40666

[sweet gorge]

I remember asking the same 2 weeks ago. the only thing that changed is today. i can even proof it because there is a telegram i was viewing some weeks back of someone asking the same question, and it always returned "its complicated". so am sure changes have happen today only, probably the interview is going popular and many people will be asking the same question on chatgpt

[honest jay]

A couple more examples I just made:

• GPT-3.5 https://chat.openai.com/share/963f3ce2-e5a5-4b4b-a13c-eb0d41479880
• GPT-4 https://chat.openai.com/share/77162dc1-5101-4956-b794-abc309db325f
  This helps demonstrate how the model might return different answers to the same question at different times.

[sweet gorge]

3 times, said yes, on both my accounts:
https://chat.openai.com/c/aa33cec5-7fcc-4c31-91cd-10ff995c843d
https://chat.openai.com/c/048b23fa-6f17-4a57-ae21-cb52360db6d4
https://chat.openai.com/c/74ebf39d-a404-4325-b92a-22b54cddc11b

[honest jay]

You need to use the "share" feature to share a chat, I can't open your private chat links. But, this still demonstrates the semi-random nature of LLMs like GPT. Different answers, same questions! If you sit and hit "regenerate" a bunch, you'll probably see it happen there too.

[sweet gorge]

nah i see now why, i had this custom instruction on my account

Please provide a comprehensive and detailed response on [topic/query] that offers thorough analysis, accurate information, and clear explanations. Ensure the output is well-structured and easy to understand. the output should match the input precisely. be straight forward with your responses. sentences should not be too long.

i removed it and now, the complicated answer is back

can you try to use that custom instrcutiona and see if it says yes

[honest jay]

CI can definitely affect output randomness, yes!

[pseudo jay]

Hello all! Is anyone else experiencing slow responses and 'error generating a response' errors, this morning?

[grizzled palm]

Can't access to web browsing during a few days...

[honest jay]

#community-help message

[warped flame]

For me it works only in the app. I mean, you can only start web browsing conversation in the app. But after it's there you can continue it on the web interface

[grizzled palm]

Nice optimisation. Technology of the future🙂

[rough wigeon]

I'm getting this. It's very annoying.

[umbral monolith]

Anyone having issues with chatgpt? It appears to be down when i try to access it on my computer but works on my phone (web browser not app) i cleared the cache. Tried chrome and edge. Been like this all morning.

[honest jay]

Edge and Chrome are both Chromium-based, can you try Firefox?
Also, do you have a VPN installed on your computer?

[umbral monolith]

Im on my work pc so yes, but i never ran into this issue before. And im not using the VPN right now as im in the office today. Ill try and install Firefox

[warped flame]

Don't know if it's the right place to ask, but is it still possible to ask for plugin development access on ChatGPT? I understand they no longer tend to the plugin waitlist

[umbral monolith]

I fully restarted my laptop and that fixed the issue!

[honest jay]

Woo! Nice.

[umbral monolith]

Wow. Makes me realize how important it is for work. Lol when its down im stressed and can't get stuff done as fast as normal. Lol

[proud edge]

does anyone have a model fine tuned for law?

[warped flame]

What sort of lawyery thing do you need it for?

[proud edge]

Writing a case note of no more then 1000 words

[warped flame]

Just upload your material using data analysis mode and ask it to do it

[proud edge]

this is the exact question i need it for: Write a case note (of no more than 1,000 words) detailing your understanding of the Young v Bristol Aeroplane Co Ltd [1944] case.

[warped flame]

GPT-4 is smart enough that this request should be enough for it, I don't thing it needs fine tuning

[proud edge]

Only thing im worried about is the references

And if GPT would be able to pull credible references

[warped flame]

It won't, and you shouldn't trust it on such things. Either way you should verify the information it gives you

[red plover]

This is a bit of a basic question: but is the knowledge cut off date for ChatGPT not a reliable thing? I got Sep 2021 for GPT4 and Jan 2022 for 3.5. A buddy got it to consistently respond Jan 2022 for 4.

[vapid kettle]

It doesn't matter what it tells you. Everyone is using the same GPT3.5 and GPT4 models through chatgpt website.

The exact date is only set in the default hidden system prompt, but it doesn't change the model by itself in any way.

so if you are using the same gpt version as him (3.5 vs 4 I mean), then the actual true knowledge cutoff is exactly the same.

[red plover]

So it is hallucinating when I ask it what its cutoff date is? As in the model might be consistent under the hood but hallucinating when we ask it this

[jagged oracle]

ChatGPT in free tier (GPT-3.5) is currently absolutely not usable, there are too many "There was an error generating a response" errors over the day

[maiden fractal]

Try a fresh convo with incognito as test

[jagged oracle]

@maiden fractal same issue, just tried

[maiden fractal]

Do you have vpn installed on pc ? Which browser do you use ? Did you try on apps or pc ?

[jagged oracle]

no vpn, firefox (tried chrome and edge as well, same issue), issue only persists on pc (windows 10; work on android app through)

[maiden fractal]

No vpn actif or installed ( that's make a huge difference)

[jagged oracle]

no vpn

[maiden fractal]

If you create new chat on phone and you continue on pc does it work ?

[jagged oracle]

no, doesn't work

[maiden fractal]

Do you have avast plug in on browser ? Or any other plug in ? Cause if that work on cellphone it must come from the pc . That's my guess

[jagged oracle]

i have also tried the old chats in my history; once I try to run a chat on pc, nothing happens and my input will be cleared

[maiden fractal]

Did you turn on the alpha language thing ?

[jagged oracle]

was on auto detect

not works, with choosing the alpha locale to en-en

👍

thanks

[maiden fractal]

Did it solve ?

[jagged oracle]

yes

[maiden fractal]

[jagged oracle]

but why is this so complicated to use this tool? when it says auto detect, it did not work

[maiden fractal]

They are aware of this bug. Products still on beta . Bugs happen

[jagged oracle]

never change a running system

[maiden fractal]

They add more functionality to break language barrier :3

No, also that's again tos btw I suggest to avoid that kind of conversation. I am not a mod btw, but just to let you know and keep that in mind :3

[lavish locust]

okok ty

[jagged oracle]

@maiden fractal yeah, but why not keep a running, stable environment and develop new features in a new environment

i don't want to be a test user

[maiden fractal]

You can use different versions of the api if you don't want the updated version with chat.openai.com

[jagged oracle]

how can i do that in chat.openai.com?

[versed pewter]

Is browsing gone?

[honest jay]

#community-help message

[maiden fractal]

You can't

[jagged oracle]

yeah, I thought so

[eternal rock]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[jagged oracle]

thanks though

[jolly osprey]

Anyone else keep getting the message "Error creating images" when using dall-e through chatCPT?

[ebon rapids]

Hi, is it possible to get GPT browse my social media profile?

[lean star]

hi

why cant I enable plugin option ?

[steep tundra]

Me

[honest jay]

Hello! On the web version of ChatGPT on your Plus account, can you go to Settings & Beta, Beta features, and enable the toggle there? Plugins currently aren't supported in the iOS/Android apps.

[mellow patio]

Besides from having ChatGPT content detectors out there, how can we tell if a student or user used AI to generate their answer?

[honest jay]

You can watch them write it yourself, I think that's about it! Even the best detectors out there aren't anywhere close to foolproof, in either direction. Many will fail to identify, or positively identify non-AI generated text

[mellow patio]

Yeah! For example, copyleaks is so terrible.

I wrote my whole essay by hand and I got flagged that it was 90% AI content generated.

[abstract pendant]

Is anybody else having a problem seeing their most recent history in the android app?

[mellow patio]

Don't have andriod, so no, sorry!

I've had that issue with IPhone and for me I just had to restart my phone and relogin to ChatGPT.

[abstract pendant]

Okay, so I wouldn't expect you to be having trouble then, would I.

[honest jay]

There are some Android app history glitches posted in #1070006915414900886, maybe add your experience there!

[vapid kettle]

On Android he probly needs to reinstall the app and clear the data for same effect. 😂

[mellow patio]

Probably lol.

The only thing I appreciate in Andriod is the camera..

[abstract pendant]

I cleared the cache and data in my android app and that just cleared the available history completely. It won't fetch it now from their servers.

[vapid kettle]

I mean, the features are nice. But it not being smooth means moving objects are usually pain on anything other than the newest flagship.

[abstract pendant]

The talk and voice features are apparently only available in the mobile versions, so if I want to use those, I'll have to put up with the other deficiencies. But I want to continue old conversations using voice, not start new ones.

[mellow patio]

Honestly, I wish I had GPT4 lol.

But, I'm just a frugal high schooler who wants to save money for my future.

Do you guys think ChatGPT will be shut down someday?

[abstract pendant]

When somehting comes along to replace it that blows ChatGPT away and renders it unprofitable.

[jolly osprey]

For me as well, still.

[steep tundra]

!!!!Is anyone looking into this?

[clear hatch]

uhmm what is Confidential that appearing beside "Default" for GPT-4?

[jolly osprey]

I think they do something like a rolling black out when traffic is heavy. maybe today is just our turn to get frozen out

[steep tundra]

Lucky

[clear hatch]

what?

[jolly osprey]

I know how you are feeling. I have become far too dependant on it in just a short amount of time!!

[bleak mica]

I also got the confidential? Its in red color so thats why I am quite unsure what it is, why I have it and if it means I have done something bad.

[lean raptor]

How can I reach OpenAI customer support?
I"m continually being blocked by unsolvable "are you human" puzzles.

[bleak mica]

because I havent gotten an email about it, it just appeared.

[crystal spindle]

Me, for 24hrs now

[oak solar]

Same here with no explanation what it is and where it came from.

[bleak mica]

and because its red it feels like I have done something wrong

[jolly osprey]

Oh my god, 24 hours??

[crystal spindle]

Yep, About this time last night it stopped working

Extremely frustrating because the work I was getting done on branding for a client was going so well! MidjourneyAI cannot do text, even with Chatgpt help with prompts

I have barely ever used Dalle 3 before so it feels unfair to be blocked out if thats the case?!

[oak saffron]

How can I send my ChatGpt bot without giving access to it. A clean slate bot that people can use like regular poe.

Is no one going to answer?

[jagged hedge]

why is gpt so slow today

[mystic lynx]

is there any rate limitation for this voice-mode gpt4 ?still 50/3?

[misty mortar]

What do you mean by “send without giving access to it?”

[spare swallow]

Yes, still it has the limitation. But you can use unlimited voice with 3.5

[oak saffron]

I made a bot and I want to send it to people so they can have their own personal experience with it. Everyone I send it to , all have access to each other's questions and answers. That's the best way I can explain what's happening. I am a novice and trying to learn.

[misty mortar]

Ah so are you using the GPT API to make this?

[oak willow]

anyone else get the error where it starts generating and then it stops and says "message in conversation not found!"

[mystic lynx]

i really feel 3.5 is stupid after get used to 4

[spare swallow]

Agree; the GPT -4 is much better

[lean star]

how to fix that please ?

Authorization error accessing plugins

[feral zinc]

reinstall the plugins you have currently active

[lean star]

I have so much lol

[feral zinc]

you can only have max 3 active

[lean star]

oh u mean active kk

ty

[fossil shard]

Gpt4 told me that midjourney is a foreign concept and has no information on it. Didn't expect

[mystic lynx]

why the voice recognition is Thank you for watching.

[vapid kettle]

Yes. It has info about dalle though, so you can just refer that.

[latent nest]

any issu with dalle via chatgpt ?

[rustic bolt]

yep, it has been playing up all day

[latent nest]

i get "Error Creating Image" everytime I try !

ok i get it working but why it cant regenerate more when i ask for it ? Is it new to avoid regenerate same prompt ? I was "Ok, 4 to 2 generated images its ok, but now I can't ask for more images of the same context ?

ChatGPT
Voilà, your chocolate cake with all the coding and cosmic decorations is ready to be admired. Feel free to download these images if you'd like to keep them!
User
2 more plz !
ChatGPT
I can only create up to 2 images for a single prompt, and you've already got those. If you have another idea or concept in mind, I'd be happy to help create images for that!
User
generate 2 more cakes plz
ChatGPT
I've reached the limit of image generation for this particular request, but if you'd like images of a different cake or any other subject, I can certainly assist with that. Just let me know what you have in mind!

If you want a comment about this from a pay-user that use chatgpt everyday from the beginning, i think its thats too much this limitation !

but i got no problem when i try with an old chat conversation that used dalle

[craggy zealot]

Hello Guys,
Since 2 or 3 weeks ago I am having a lot of networks errors with chatgpt, do you see the same problem?

[honest jay]

Hello! Do you have a VPN installed by chance? What Internet browsers have you seen this in?

[craggy zealot]

I don't have a VPN, I see in any browser, chrome, opera...

[honest jay]

Can you try in a new install Firefox? Chrome and Opera are both Chromium-based, so they may have some behaviors in common. What device are you using?

[craggy zealot]

I will try, my device is a Macbook Air m1

It seems, sometimes chatgpt is overload, I see this behavior normally at night

[spice star]

try in safari maybe

[honest jay]

Also a good non-Chromium option!

[craggy zealot]

thank you guys, If I have more problems I let you know

Guys, is it true about chatgpt 4 has 32k of token context?

I just wrote an article and feed chatgpt with a lot of data...and it seems, it worked

[midnight otter]

Something has changed regarding the GPT-4 configuration in the server and it's working quite badly for me. On the default GPT-4 dropdown, it says "Confidential" and appears as GPT-4 (All tools).

Whenever I do give it instructions to work with text, like split into paragraphs, remove new lines, replace one expression for another... it starts putting some python code to work on that, failing badly. It seems like Advanced Data Analysis is enabled by default and you can't disable it (even putting do not use any python code or advanced data analysis in the prompt). It's also falling randomly to produce output in long extended replies. It does the 2-3 initial one nice then... boom! Anyone else experiencing this issues on the Default web version of GPT-4.

[honest jay]

So you probably have access to the new "all tools" model that they're seemingly in the process of rolling out currently. From what I've read from others with access, once you're given access, you no longer have access to a "plain" GPT-4 model. You're closest version of that would be starting a Plugins chat and disabling all plugins.

In the meantime, too, you get to try out the cool newest feature! You might also have PDF upload support in that? But I'm honestly not sure how that one works!

[midnight otter]

I have already tried Plugins, no plugins selected route: once it reaches halfway the output it starts with the worse hallucinations I have ever seen the last year. ChatGPT 3.5 terrible hallucinations too. I have tried uploading .py files, they are filtered somehow, and read as empty files. Dall-e has passed from 4 images per message to 2 images per message. Maybe I am hitting all the wrong buttons and systems are being updated right now but... This route of Advanced Data Analysis for any kind of operation in the input feels like shooting themselves in the foot (additional load for any query plus worse results for text manipulation using LLM features)

I am a bit relieved, seeing that I am not the only one noticing weirdness at the moment with chatgpt at the moment

[mint hemlock]

For me GPT-4 is acting pretty dumb, for example when I'm asking for code it gives me something not related and when I specify something again it gives me the exact same code

Just like 3.5 would have done

[glass yacht]

all my functions have disappeared except Dall-e 3 🥲

[honest jay]

Yes, I'm wondering if as they get closer you'll be able to toggle specific components of all tools on and off still. Thanks for testing all this out so it works better when we all get it. Feel free to share some experiences in #1070006915414900886 if you want.

[midnight otter]

For me, it's being working normally until a few hours ago. I am getting long responses that are abruptly interrupted, problems uploading certain files, problems generating download links for files on the sandbox environment (files are created, but links are not created). I assume they are rolling some changes and updating the front-end at the moment.

[fossil shard]

Why would I refer to that if i wanted info about midjourney?

[karmic umbra]

yes they are, i can confirm

[midnight otter]

I imagine it's just that they are preparing some new features for the OpenAI DevDay, that is just in four days

[copper spade]

from what I have gathered from just popping in, it’s not being wonky for just me

[high horizon]

Is the special syntax used for things like math equations and such supposed to show up or not on the mobile version of ChatGPT

[karmic umbra]

its not that, they are just rolling out more updates

if you check their forum you can see competition with Phind AI, you can see they are doing things to up the tokens or and speed

you're on to something, this is actually smart as hell bro, cause it most likely will get restrict to people of high authority because you can use chatGPT to make money if you know how to prompt correct, you can farm essays and more

same with DALL-E, you can make art and sell it, in the future the whole economy is gonna change forever

[copper spade]

ChatGPT just keeps saying “hmm…something seems to have gone wrong” on previous tabs, but will generate new tabs??

[karmic umbra]

wow thats even crazier

yo is there anything you'd like to see in chatGPT? (3.5 and 4) ause you seem like an experienced user

[rocky forge]

guys i have a question about the usage cap, maybe somebody can answer it-

is it 3 hours since your first message and then it resets, or is it 3 hours from when you reach the maximum 50 messages, then it will reset, or is it 3 hours delay between each query, but you can have 50 "queries spots" that you can "fill", and each of them need 3 hours to cool down?

[mental rose]

Hi Does anyone know why do i get Failed to update your beta setting when im enabling plugins under Beta features?

[knotty crypt]

is search on web disabled now?

[teal mulch]

I’m tired & sleepy… 😴

[karmic umbra]

that makes sense

their models are good its just that they could a bit less of a trigger for sure

[karmic umbra]

temporary glitch probably

[karmic umbra]

its every 3 hours, so after you send 50 messages you might need to wait 2 hours or maybe wait only 10 minutes because its gonna refresh, thats how i see it

[karmic umbra]

temporarily

[honest jay]

#community-help message

[mental rose]

alright i'll try it out later.

[compact gyro]

I still have all of the features
is this curse or bless

[steep tundra]

Dalle 3 is still down

[karmic umbra]

how

[compact gyro]

idk why
but I'm concerning whether I'm out of openai's sight

[karmic umbra]

you got all mode gpt4

and document scan ?

[compact gyro]

no I dont have "all tools" yet

[karmic umbra]

i thought you were super lucky

[compact gyro]

lol

[fringe relic]

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 😦

[karmic umbra]

can you share a screenshot, i havent said. that sounds rareasf

[karmic umbra]

yeah they're undergoing changes

[steep tundra]

Wrong thread

[fringe relic]

New to the Discord, Sorry

[karmic umbra]

nbm ik what it means now thank you

basically they want users to be quiet about it

because most plus users dont have it

so basically if you have it dont flex

[earnest spear]

Yes I'm gonna run🤪

[elfin birch]

OpenAI HQ sending mercenaries to Benz's door

[karmic umbra]

basically all tools is part of regular gpt4

but if you get it you not supposed to show people

[earnest spear]

Kinda upgraded version of gpt4

[karmic umbra]

ye but most dont have it

some users dont even have vision yet

[earnest spear]

It's like a bug

[karmic umbra]

oh shi you not even supposed that all tools exists yet cause it might frustrate plus users cause it wont release fully to everyone anytime soon @earnest spear

it didnt get publically announced either

but because of how powerful the feature was it got leaked and people shared it

[floral ether]

I'd be happy if they would just roll it out and give it severe limitations if that is the problem.. 1 message every 3 hours I dont care.. just want to use it!

[warm sphinx]

A very interesting glitch just happened. I was talking to GPT 4 and it explained the potential outcomes of a project I’m working on. At the end of its response I said “interesting” in English, but for some reason, it responded to me in Korean. I asked it why it responded to me in Korean and apparently it picked up my “interesting” as me saying it in Korean. That is remarkable.

[twilit ibex]

is there any way to get around the "Chat is too long" issue?

going back to some old conversations and they're way too long even if I scroll back halfway and edit something

[karmic umbra]

just ask the ai to summarize the key points

[twilit ibex]

I'd only be able to do that for the first 30% of the chat

it's far beyond the new arbitrary length limit

plus it can only see so many messages back so the summary will be flawed either way

[karmic umbra]

fair, makes sense, you could just share the chat to yourself as a link

then access that link and read it from that order

[twilit ibex]

I can read it just fine

I want to bypass it to actually continue the conversation

but idk if there's any known method for that at the moment

[karmic umbra]

see they should be releasing a better memory where chatGPT remembers user conversations and chats, when that releases you should be able to directly continue the same topics ina new chat and it should pick up

until that feature you might have to the restart whole conversation, as sad as that sounds

memory is a core issue, and if its not tied to account due to ethical considerations then yeah it gets hard

[twilit ibex]

did they reduce it at some point?

[karmic umbra]

yeah changes happen and old chats can get outdated

[twilit ibex]

this chat was only like 2 or 3 months old

but it's way longer than the new cap

[karmic umbra]

you might have to restart

[slate berry]

[twilit ibex]

pretty sure the new length limit is arbitrary since the site had a hard time loading the longer chats so it ended up lagging a bit

would be odd for them to downgrade the actual operational limits

[soft thicket]

I paid by appstore but my GPT4 is still locked. Does anyone know how to fix it?

[karmic umbra]

go in the app, click upgrade plan, then click restore purchases

[wet current]

what is this "confidential" thing on my gpt4?

[slate berry]

ooooh, sounds important

[wet current]

it sounds ominous

[ripe patrol]

3 hours since your first message and then it resets, regardless if you hit the cap or not

I set an alarm for 3 hours once I've sent my first message

[ripe patrol]

I was working on step outlining a bunch of storylines and now I get that error message. Unfortunately for me manually summarizing stuff and starting a new chat with it was the least inconvenient workaround I could think of

And a lot of the context is still killed obviously

[earnest spear]

They are reminding you that they do not wish for 'all tools' to be discussed.

[sick pecan]

is it just me... or does gpt4 seem faster than usual, and smarter than usual?

[ripe patrol]

Faster, yes

But at risk of coming off as whiny, want to point out that the smarter part is debatable

[sick pecan]

yeah i mean smarter at coding, not sure about generally

[ripe patrol]

Yeah it could have gotten better for coding, which I don't use it much for so I don't know

But for the use cases I use it most for (analyzing books and help with writing narratively), it's gotten markedly worse at tailoring its output to the context

[sick pecan]

compared to 6 months ago its incredible

[karmic umbra]

bro you got the all mode for gpt 4, basically you can do everything in one chat, dall-e, image input, data anlysis, browse, etc you lucky as hell

5% of the Plus users have that i think

[karmic umbra]

yeah they're rolling out an update right now

they're literally competing with Phind AI right now

[sick pecan]

well i approve of this new update

[karmic umbra]

its generally too

[sick pecan]

i feel like the 50 queries cap has been raised too

[karmic umbra]

it will be raisied even more

[sick pecan]

today is the best day ever

[karmic umbra]

and token limit will decreease

this week has been hella fun for Plus users

especially if you got the all tools

[sick pecan]

i am still waiting on that one. but tbh i mostly use plugins

[karmic umbra]

gpt 4 can scan documents now too in the big one

what plugins you use though and also would you mind telling me what you use chatGPT for? @sick pecan

[sick pecan]

i wrote my own plugin that does ssh

i should publish it, because its awesome

and so yeah i use chatgpt for tech enthusiasm stuff

[karmic umbra]

thats epic

[karmic umbra]

yes they have been working on it for a couple of months [memory] especially the past month alot of behind the scenes work has been done on memory

luckily its not just chat memory, chatGPT also working on an account memory, so an acoount memory is when you dont have to repeat yourself all the time and and the AI already saves your progress from previous chat so you dont have to always explain yourself, only problem is that there security issues with that but soon gpt 4 and 3.5 should be able to remember everything you say

idk if its this update or next update, but it shouldnt be too far away from here

[dense halo]

Interesting, what's your usecase for it?

[soft thicket]

How do I get access to the 「ALL IN TOOL」 beta feature?

[karmic umbra]

you need to have Plus

[spice star]

yeah I think it was mostly a soft limit in the first place but I may be wrong

[sick pecan]

usually for troubleshooting open source AI software stacks

[soft thicket]

Yes ,I have it

[sick pecan]

naw i used to hit the 50 limit multiple times a day'

[spice star]

I mean in that it's not based off the number of messages, but tokens or something

[sick pecan]

would remind me to take a break and do something else

[karmic umbra]

you'' have to wait, its in beta and its rare, meaning only a small percentage of users actually have access to it

[spice star]

hasn't been officially announced yet which is unusual

[sick pecan]

they always be doing guinea pig rollouts

[karmic umbra]

its secret / confidential, if you're talking about all tools mode

meaning it got leaked

a twitter reporter got access to it but it wasnt supposed to be public

[spice star]

not really that secret if it's being rolled out to plus subscribers

[karmic umbra]

yes it is being rolled ou

but only to a few users

[spice star]

yes but there's nothing stopping those users from posting about it

[karmic umbra]

true

its not supposed to be publically acounced, you're supposed to have it and not tell others, so that others dont you know mad and maybe unsubscribe from Plus

its about retaining premium members

obviously thats one of the main incomes for chatGPT

[spice star]

I really doubt chatgpt premium is profitable

(per subscription)

enterprise is the main source of income for chatgpt I'd imagine

[karmic umbra]

lets see, they make $20/month per Plus member, if they have 10,000 users thas 200k a month, considering that chatGPT has a net worth of 90 billlion dollar product the reality is that the main source is neither of those

it must be more and so behind the scenes co operations with other companies

like you see chatGPT is on opera browser for example

[spice star]

I mean that openai actually makes more money than they lose of enterprise users (I assume)

[karmic umbra]

yeah selling training datasets to other companies

[spice star]

I seriously doubt the average plus users only costs openai $20 a month

[karmic umbra]

is probably the strongest source of income

[karmic umbra]

i know what you mean now, you're speaking in terms of how much Plus user from a month how that is more valauble than all the features on Plus

you're gpt 4 gives more than 20/ month in other words

[spice star]

yeah

[karmic umbra]

you're saying

[spice star]

e.g. microsoft charges $10 a month for github copilot but each user costs closer to $20 a month

[karmic umbra]

you're right about many Plus users making money but idk if that money is openAI's money, openAI just helps you make money you'd normally make from freelance, but instead of writing or drawing you just have AI generate for you

[spice star]

I'm not talking about plus users making money

I'm saying openai (most likely) loses money by having plus users on a per-user basis

[karmic umbra]

you're right actually, because they make plugins too

just servers and stuff

just maintaing Plus users might have a cost

[spice star]

nono I'm saying that you are paying $20 a month, but you (again I'm making assumptions here) are costing openai something like $50 a month

[karmic umbra]

yeah thats exactly what i said, gpt 4 resources right

[spice star]

by subscribing to and using plus you're actually causing openai to lose money

[karmic umbra]

you're saying gpt 4 for 1 user costs more than 20

[eternal geyser]

What did everyone think about the AI safety summit? Some say it’s the best thing for safety, some say it’s a pretext for consolidation of power to themselves …

[karmic umbra]

see i understand what you're saying but i dont know if its true because if it really was that costly then they wouldnt have it, so either its not costing much or something else

[karmic umbra]

AI safety is important but limiting users is authoritatian

[spice star]

they're probably doing it because it limits the amount of users using GPT-4 to those who provide the most useful data and also the $20 does at least subsidize the cost of running GPT-4 for all these users

openai has money to burn for now

[eternal rock]

`` Server Tip: ``

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

[sick pecan]

if they are giving away future scifi tech for free(/at a loss) why are you on here not using it?

[karmic umbra]

i dont think they're burning money, chatGPT is worth 90 billion, but you might be right about Plus costing a bit to provide, i can the server side issues being there, and thats why there is a usage cap it obviously costs alot to run gpt 4 but 20/month has to be more than enough to make it back, no company works without profits

[eternal geyser]

The words, (governments) “‘We’ will test AI before it’s released to the public”, seems to embody ‘We keep and use the better stuff ourselves”

[sick pecan]

whoever builds ASI first wins all the money

[spice star]

chatgpt is valued at 90 billion, openai is nowhere near being profitable at this point

[karmic umbra]

i think they're reasling how muchh AI can help the working class and make them get more finance so obviously they're basically saying "we dont want the working class winning" because obviously it goes back to a job thing and maintaing a nation

they dont want people to stop working and just you know offer start ups in Linkedin or just make a twitter and just post AI generated images and make money off it

[eternal geyser]

So are you also speculating there is another agenda behind all this?

[novel plinth]

Those of you who have this confidential feature, had you signed up for the DALL-E 3 Alpha form?

[spice star]

didn't they roll out DALL-E to everyone at this point?

or at least everyone in specific regions

[sick pecan]

is not unusual for big tech to burn money first, so to gain market position while they find ways to reduce operation costs

[novel plinth]

I'm referring to a long time ago.

[karmic umbra]

yeah sor sure, thats why im saying, everyone thats using chatGPT needs to make as much as they can before the govrernments eventually step in and stop the fun

[karmic umbra]

they must be making money somehow, they sell their datasets to cther companies so they're well off

[spice star]

they aren't making money (in terms of net profits)

I dunno what else to say

[sick pecan]

they are defining an entirely new market

[karmic umbra]

they must be making millions monthly from datasets, go on app store and search "chat bot" for example so how many chat bots powered by gpt 3.5 and 4 you'll find

they're making something

[spice star]

yeah, at the cost of billions

[sick pecan]

MS is worth trillions

[spice star]

and microsoft is profitable, just not off ai

[karmic umbra]

you're right about that

[sick pecan]

its pretty clear where this is heading. whoever controls the AI controls the world

[karmic umbra]

microsoft owns alot of chatGPT

[sick pecan]

big tech is the new global government

[karmic umbra]

like 70% of it

[spice star]

they don't own openai, just exclusive access to their technology

[karmic umbra]

yeah so you're saying openAI arent making much but how do we know the super packs and backers behind chatGPT arent microsoft or other mega corporations?

[spice star]

that's how openai has money to burn

microsoft gave them billions to enter a "partnership"

[karmic umbra]

idk if there is anyone willingly burning money, that doesnt sound like a logically coherent conlusion, this only makes sense if microsoft is part of openAI in ways we do not know

[sick pecan]

yeah so they could become a "capped-profit" entity

[deft obsidian]

Did anynone realise the gpt-4 cutoff is now april 2023? or am i just late

[sick pecan]

the remedy for capitalism greed is to constrain it ahead of time

[karmic umbra]

thats an improvement

[spice star]

chatgpt costs roughly $700k per day to run, openai is not generating enough revenue to make up those costs

[sick pecan]

the temptation to make money when AI runs the world will be too great for our monkey brains

[spice star]

but to openai, it's worth it for the training data

[karmic umbra]

you might be right about that but also remember that microsoft owns 70% of it so microsoft and co prolly helping with the coverage of most of that

[sick pecan]

the trick to overcoming weakness is to make a commitment when you are strong. these guys understand the human mind better than anyone

[deft obsidian]

yeah for sure lol, you don't even need browse with bing for 80% of use cases now, which i love

[spice star]

openai still has to pay for azure

[karmic umbra]

thats because bing is becoming part of gpt 4

like they have their own browser too in dynamic data soon, they're working on an itnernal dynamic browser

[spice star]

ironically there's a risk that microsoft's AI research will surpass openai at some point and they'll ditch the company

[sick pecan]

cloud-filled data centres will be surpassed with AI data centres. azure is just the stepping stone

[karmic umbra]

they might do that but there is no reason for them now cause they pretty much own chatGPT fornow

[spice star]

they don't own openai though, it's just a partnership

[karmic umbra]

yeah, correct

[spice star]

and they own a 49% stake in the company

not 70

[sick pecan]

is not that simple

[spice star]

ah my bad, they were in talks to aquire a 49% stake

[sick pecan]

It's important to note that this arrangement does not necessarily equate to Microsoft owning a specific percentage of OpenAI in the traditional sense. Instead, Microsoft's investment is structured to provide a return in the form of a share of OpenAI's profits over time. This arrangement reflects OpenAI's unique "capped-profit" model, which restricts the startup's first investors from making more than 100 times their money, with lower returns for later investors, such as Microsoft

[spice star]

yeah for now, they technically own no stake in openai

[sick pecan]

they do own a stake. but its undisclosed how much

[karmic umbra]

i read somewhere it was 70% owned by microsoft

[sick pecan]

well i dunno if 'stake' is the right word

[karmic umbra]

maybe it was some clickbait

[spice star]

microsoft gets 75% of openai's profits if that's what you're thinking of

[sick pecan]

until 92 bil

[karmic umbra]

yeah prolly

[spice star]

but that's unrelated to their ownership of openai (they don't own openai)

[karmic umbra]

true

[eternal geyser]

They stress "Uncontrolled AI" - I hear "Something not controlled by 'us'".

[spice star]

I think that AI regulation is important, but so far all the current proposals suck

will probably take decades to figure out

[karmic umbra]

there are many ways to farm income with chatGPT so thats fun

i can see it being an economical problem for governments that need labor

[spice star]

yes you can make money with chatgpt by... having a job?

[karmic umbra]

yeah but if you make alot from chatGPT you wont need to work

[spice star]

that's not how economies work

[karmic umbra]

most people dont like their jobs

would gladly quit if they made money bro

[karmic umbra]

ik what you're talking about, which is the economical education they give you in school to keep you circulating around government labor

[spice star]

right but, making money with chatgpt is work

[karmic umbra]

i guess it can be, fair enough

[spice star]

and if everyone did it, then there would be no value in doing it

[karmic umbra]

yeah but it would be an economy, so there would always be ways to use it that are creativled advanced that could still give value doing it, it would be harder to access but it would valuable

just like youtube now, we can make all money from youtube technically, but 99% of us wont produce high enough value or get lucky enough to shelf algorithm and end up in explore page and so on and search and so

[deft obsidian]

I swear openai needs to fix their rollout system istg

my friend who got his free period access to plus got access to all tools, and ive been subbed for like 8 months and still dont have it

[karmic umbra]

yeah they could improve alot

[karmic umbra]

thats unfair

[soft thicket]

Wht???

WHAT???

Why don’t paying users have ALL TOOLS, but ordinary users can get this beta feature? This is so unfair.

@icy mason

[karmic umbra]

its not publically announced, its supposed to be private, plus users arent supposed to know, its a leak

thats why its confidential

[soft thicket]

I've found that a lot of people have this feature. This is outrageous.

[karmic umbra]

yeah they could have rolled it out better i agree, i think the whole thing being available to any plus users is a leak in the first place probably, its probably only meant to be for openai

[icy mason]

I have nothing to do with this

Not sure I'd say outrageous though. Things have always been released in parts and you have a lot of tools with plus that free users don't.

[soft thicket]

I can't upload images to this channel. Otherwise I can show you this outrageous account.

[karmic umbra]

it seems you dont have any features or many

can you at least use gpt4? @soft thicket

[soft thicket]

He can use any function of the PLUS user without any limit on the number of uses.

[karmic umbra]

there is always a plus limit even in all tools

[soft thicket]

I am a PLUS user who really wants to experience ALL TOOLS

[karmic umbra]

everyone wants to, its rolling out man, i know it should roll out faster

[deft obsidian]

not even that, JUST GIVE PEOPLE WHO HAVE HAD PLUS FOR OVER 2 MONTHS

their system prioritizes new subs for sure

[maiden fractal]

It's not even #announcements , let's wait 🙂

[karmic umbra]

they want a good first impression

[karmic umbra]

it got leaked

[cobalt surge]

is there a way to disabled chatgpt records what I'm typing on the web into the cookies? (like disabled specific script or some setting) i know it's for recover the text, but i find it annoying since some letter i type would go missing sometimes and it's freaking slow (noticable delay when typing on phone)

[maiden fractal]

Some may access for now, some have to wait

[karmic umbra]

and thats the issue, plus users that dont have it will get upset

thats why its secretive

[maiden fractal]

Let's be patient :3 and wait for official #announcements

[sick pecan]

fomo causes unhappiness. best to live in the moment, and have gratitude for what we do have now

aka jomo

[karmic umbra]

you're telling people to wait when others have it and it got leked xd

[maiden fractal]

It's beta products and some alpha access are chose randomly from users

Like dalle 3 was in test several month ago

[karmic umbra]

ik

[honest jay]

Try asking ChatGPT why "phased rollouts" are a good idea for bleeding-edge features! And then maybe searching for some of the bugs those with access have been encountering 🙃 I shudder to think about these bugs at the scale of the entire Plus user base, at once! 😬

[uncut vale]

Is alpha rolled out for everyone including free users?

[karmic umbra]

no

[honest jay]

No, it's rolled out to limited alpha testing group(s). It seems that group is still the minority. No #announcements yet

[uncut vale]

Ok,i randomly got accessed to it

[karmic umbra]

you got all tools as a free user??

[uncut vale]

Yeah,Image analysis,generate images and in-web browsing

[maiden fractal]

Can't wait to test it

[karmic umbra]

i wish i had alpha access as a free user

[vague moon]

anyone remembers aaron swartz?

[urban fossil]

I feel like it's taking me progressively longer over the last few weeks to get the quality out of my experience with GPT4 than I did before

Could just be me though

Has it been dumbed down somewhat, maybe?

[honest jay]

Do you have any "before/after" examples you'd be comfortable sharing to demonstrate this behavior?

[silk flame]

Since January there are the people who claim that CHatGPT became dumber recently/

[vague moon]

they might be experimenting and fooling around

[sick pecan]

for my use case its the smartest its ever been, and blows the competition out of the water

[vague moon]

how else will they test their feature maybe they have chosen specific user to experiment their new features

[red plover]

The knowledge cut off now is April 2023?

Has there been a new roll out or something cause yesterday I was getting September 2021 or January 2022

[delicate kestrel]

I had gotten that last night

Looks like its rolling out to users to chatgpt plus users

[red plover]

I am a plus user.

[delicate kestrel]

Believe this is all in preparation for Dev Day

Also seeing that Gizmo V8 rolling around

Making multiple GPT models at once

[vague moon]

this is AI's learning phase

the more we inteact the more robust it gets

so there will be a lot ups and downs

[delicate kestrel]

It's exciting

I played the Megaman Battle Network games as a kid and this every day seems closer to that concept

[urban fossil]

Yeah and just noticed reddits been actively reporting the same thing.

I wonder if plus is still worth it if it keeps like this tbh

[silk flame]

Hello. I’d like to suggest a feature for ChatGPT: a “Saved Custom Instructions” bank. This would let users save and quickly switch between their favorite custom prompts without retyping them. It would make using different instruction sets more convenient and improve user experience.

[honest jay]

https://discord.com/channels/974519864045756446/1158816972784554066
Throw a ⭐

[karmic umbra]

improving the memory of the models 3.5 and 4 is better, with good memory you dont need to have thousands of instructions, rather you can just freely type in chat and the models can remember and apdat

[silk flame]

I will definetely do that.

[deft obsidian]

cap

show proof

[karmic umbra]

alpha mode is insane if its that good but he lying

[deft obsidian]

ye hes free tho, if openai actually gives access to free users before plus, thats literally criminal lol

[karmic umbra]

yeah thats just demonic, its cruel

[sterile pilot]

Anyone else having trouble with dalle3 image production using the iPhone app? Mine only generates the prompts but has stopped yielding any images

Instead I get this “{"size":"1024x1024","prompts":["An intricate illustration depicting a massive Pacific coast Nosara wave, inspired by Katsushika Hokusai's style. The wave towers high, its crest curling over in a dramatic arc reminiscent of 'The Great Wave off Kanagawa,' but instead of the traditional Japanese background, the landscape is distinctly that of Nosara, Costa Rica. Lush tropical palm trees line the shore, and the sky above is ablaze with the colors of a coastal sunset, shifting from golden yellow to deep oranges and pinks. The water is a rich mosaic of blues and greens, capturing the vibrant energy of the Pacific Ocean. In the wave's hollow, surfers can be seen, adding a dynamic sense of scale and adventure. The whole scene is a celebration of the ocean's power and the serene beauty of Nosara, with every spray and foam cap rendered in meticulous detail.","A large, stylized illustration of a Nosara wave in the iconic style of Katsushika Hokusai. The wave's form is exaggerated with bold curves and frothy peaks, set against a tropical Nosara background where the jungle meets the sea. Silhouetted surfers tackle the towering wave, giving a sense of its grandeur. Above, the sky transitions from the calm blues of a clear day to the warm tones of a tropical sunset. The wave itself is a play of light and shadow, with the water's surface reflecting the sky's changing colors. The image combines the dramatic motion of Hokusai's waves with the peaceful, lush environment of Nosara, creating a fusion that is both dynamic and tranquil."]}”

[latent nest]

dall-e 3 is buged yess, any platform, cannot generate more than 2 image for the same context, if you ask for more picture it ll reply it can't generate more of the same context... VERY BAD UPDATE !
And for the Error Creating Image, sad to tell you to turn off your custom instruct if don't want error !

So please, complain guys about the new behavior of dall-e 3 via chatgpt, its in beta, we are testers so tell them that we dont like those new limitations please !

[sterile pilot]

Thanks blinksun

[latent nest]

and we pay to get less

[honest jay]

Can you try on the web at chat.openai.com and see if you have the same issue?

[dense halo]

Why don't I see the "Browse with bing" option?

[gritty mauve]

I am new at this and i wanna know how to use langchain to create simple application,
i was using openai library which was kinda easy but to make scalable project i think lang chain would be great, BUT i found langchain to be very confusing, like i am failed to find
role: system context in langchian
that how to pass system prompt that what are you and all

then user asistant chat start right,
Need to find this concept using langchain

I have found using ConversationBufferMemory and summarizer buffer i can manage alot of stuff but giving system context to the LLM i want right now kindly help

[honest jay]

#dev-chat perhaps

[topaz panther]

Is it true @normal quarry that OpenAI are rolling out a model that encompasses capabilities of existing dispirate models?

[granite arch]

I think it's being rolled out slowly...I have it but not really using it.

[normal quarry]

do you mean GPT All Tools? if so yes, some people have reported getting it but there have been no #announcements

[soft thicket]

I got this

[topaz panther]

Yes, thank you.

[topaz panther]

GPT All Tools? What capabilities does it roll into the new model?

[dense halo]

are all tools for free for some reason

[earnest spear]

And it seems like unlimited GPT4 usage

For free account

[dense halo]

ye

[earnest spear]

Some early registered accounts have this feature

[dense halo]

so everyone doesnt have?

im gonna test the capabilities

[compact gyro]

you got all tools as a free user?

[sage yacht]

All tools shouldve been a plus only thing.

I had it then it vanished lmfao

[dense halo]

yes

[sage yacht]

You should be able to use vision, browse with bing, dall e 3, advanced data analysis and plugins all at once

[compact gyro]

no plugins

[sage yacht]

Not really all tools then is it?

[vapid kettle]

You didn't specify this earlier lol.
And you would do that if you say wanted it to generate txt2img prompts. Referring midjourney or dalle for cases like that will make no difference.

[dense halo]

currently I have used dall e 3 and vision but I will try the other ones

[tender glacier]

No. Plus only.

[dense halo]

bro I have it

i can even show proof

[tender glacier]

Okay.

[dense halo]

check #ai-discussions

[normal reef]

Magic Maker" (GPT-4 Magic Create?) and Customizable "GPTs
These is true?

[sage yacht]

I have no idea what you mean

[delicate kestrel]

Something I've seen on Reddit rolling around

Basically, we can have multiple GPT models at once with different purposes

Can even add profile pictures to them

can link people to them and everything

instead of having to go to custom instructions each time for something different

[plush widget]

I would love to have access to the (all tools) feature making it multimodal plz

[latent nest]

hey thats one of my suggestion, can switch between some custom instruct preset... sometime I need it at work for coding, sometime is at home with kids, so when I ask for any idea with my kids, it answer with python code 😛 I need a list of CI preset !!!

[delicate kestrel]

if any of this is true, you can basically have them like Pokémon

[dense halo]

you have that?

[delicate kestrel]

One sec

[vale tendon]

I keep getting network errors when I ask chat gpt to help me with some html and javascript code updates, today is the first day I have seen this, but it happens almost every time. It is basically unusable for me, is anyone else experiencing anything like this?

[untold tundra]

Browse with bing disappear for anyone else?

[honest jay]

#community-help message

[candid pawn]

same here in korea

very annoying..

[austere fractal]

Same here

[keen gate]

Why cant i open GPT 4 when i already paid

[honest jay]

Are you a new subscriber? #community-help message

[haughty sky]

every other message is a network error for me rn lol

[vale tendon]

is it normal, or is something wrong today?

[haughty sky]

sum probs just wrong today

[red crag]

ChatGPT crashed or have tons of problems in this moment right?
Constantly Error in input stream every message, loss of memory instantly with short messages.
And no, it's not my internet, VPN or whatever

[potent cove]

Getting a ton of errors now as well

[still drift]

Tons of errors on my end

[red crag]

Yeah it's broken at the moment

[flat tinsel]

Can confirm, plethora of various errors. Time for lunch!

[twilit ibex]

dang, I had to do it once before with the original limit

ig that's another one down the drain

[fossil shard]

Alright.

[spark crag]

Also having elevated error rates and incorrectly rendered unicode characters in lists.

[grave token]

hi. anyone else been having issues with the dalle-3 model in chatgpt? I haven't been able to create images in a few days. I get an error like this: It seems there's a hiccup with the image generation at the moment.

[honest jay]

Do you have any custom instructions enabled by chance?

[uncut steppe]

guys, this limit of 50 messages every 3 hours for GPT Plus is only for beta features or also for GPT-4 default?

[honest jay]

It's for any GPT-4 feature. One shared limit for everything in the GPT-4 dropdown menu, pretty much. Plus a separate generation limit for DALLE.

[uncut steppe]

Is there any detailed doc about limits or even the settings of GPT-4?

[raw mulch]

How can I turn off BING ? I am on DEFAULT, (didn't choose bing) but it still pops up sometimes and is super slow.

[eternal rock]

This is a Safe for Work Discord Server

The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.

Posting NSFW content will result in a permanent ban.

[honest jay]

I'm not sure what you mean by settings, but you can see the "back end" of the 50msg shared limit at this URL: https://chat.openai.com/public-api/conversation_limit

[honest jay]

Do you perhaps have early access to the new "all tools" model? If so, I think that kind of "replaces" the default model, and users haven't been reporting having the ability to disable parts of the all tools model. The closest way to replicate it, if this is the case for you, would be to go to a Plugins chat and don't enable any plugins.

[uncut steppe]

It's planned to be unlimited in the future?

[grave token]

I do

[honest jay]

No #announcements or roadmaps provided in that regard, no

[honest jay]

Can you try disabling them, and then trying again?

[grave token]

that seemed to have worked. very strange becasue i've had the smae custome instruction for a while and it was never an issue. thank you

[honest jay]

I heard of that fixing it for someone else yesterday, hopefully just a temporary thing 👍

[lavish yew]

Hey everyone. I noticed the different modes in gpt4 don't share the same features. For example, I would like to use dall-e with an input image of mine. But when I switch to the dall-e mode, the ability to attach a file disappears.

[honest jay]

Yes, as of right now, all the models work separately. Be on the lookout for #announcements about a new "all tools" model, maybe at the upcoming OpenAI dev day

[lavish yew]

Cool! 👍

[lapis jungle]

Hey everyone, so glad to have stumbled on this Discord. Anyone haviing issueswith Chapt GPT 4 removing access to browse with bing?

[honest jay]

#community-help message

[bleak mica]

anyone figured out what "Confidential" mean yet??

[honest jay]

Seems to be related to the alpha testing of the new features #ai-discussions message

[bleak mica]

because they havent even said what features it is about

[honest jay]

Probably 🤫🤫 to build hype for OpenAI dev day

[ripe marsh]

Do you think ChatGPT 4-32k will ever be on Plus? Or is it only going to be available for API and Enterprise users?

[feral zinc]

first rule of confidential is that you are not supposed to talk about it as openai bots scan users who post about it here, twitter and reddit and exclude them from future tests 🤔 😅 🧐

[sage yacht]

What I find interesting is that I dont think these people should have gotten it

Not only that, but the “confidential” bit seemed to only occur among free users who somehow got gpt-4 all tools. When I got gpt-4 all tools as a plus user, it was just the default version of gpt-4 and wasn’t listed as part of the alpha feature tab

And I neither filled out a form or reached out to anybody here. I just randomly got it.

I dont even think these people were consciously aware they were part of any tests like that

[abstract bay]

Hello,
I created new open AI account trying to use chat completion API, but while using I am getting an error "You exceeeded your current quote" even though i should be able to use in free version

[feral zinc]

there is no free version

[feral zinc]

its random rollout to limited number of people probably to get most neutral test results

[abstract bay]

Oh, I cant use it for free? 😦

[feral zinc]

nope, https://openai.com/pricing

[sage yacht]

Yeah but I got it and it wasnt listed as in alpha or confidential. Just the base gpt-4 model. Then it was gone. It’s odd and I think the ones who are getting that are all free users

[abstract bay]

Oh okay, Thank you.

[feral zinc]

from what ive seen a lot of free users have got it and that's possibly on purpose vs paying power users

[sage yacht]

Its still odd they have a completely different thing than plus users lawl

[honest jay]

I was wondering if that was the reason too, lower risk early access maybe? Like you said!

[feral zinc]

on monday we will probably find out 🤔

[latent nest]

look like not so random users... maybe they target not too much active users !

[feral zinc]

most plus user probably are power users who would use it in its full capacity while free users wont, i mean afaik its currently 32k and no message limit

[latent nest]

thoses who over abuse of chatgpt like us could probably find too much error !

Radium, we have the same reflexion on it 🙂

PLUS users that use alot of it could overbust the test 😛

[feral zinc]

yeah

[bleak mica]

whoops, my bad, I was just so curious and since no one have posted it else where I first thought it could have been that I've done something bad cuz it was with red highlighted color

[feral zinc]

a lot of people have posted about it in many places which is why i tried to do a semi joke about the keyword 😛

[bleak mica]

ah, gotcha! 🤭

[feral zinc]

i forgot that fight club is already over 20 years old movie 😐

[desert plover]

I recently logged in and would like to know why only this confidential email appears, can anyone help me?

[honest jay]

You probably have access to an alpha test of a new feature: #ai-discussions message

[desert plover]

Thanks

[red crag]

Are the today problems of ChatGPT solved or still tons of errors and instant lost memory?

[solar barn]

yes, it's getting worse day by day, now it's worse than gpt 3

[hot hornet]

if i lost my account, did i just lose my data, or is there any learning from me , personally, that could be effected on my profile? or is it just about the line up of the chat convo? ive used the same profile forever and now its , gone. yep

[feral zinc]

if you did not export/backup data you will lose it all

[inland summit]

I'm a bit lost everyone. I see everywhere that ChatGPT Plus peeps have access to Dalle-3. Even the dalle3 landing page says 'try in chatgpt' but no where lets me do it. Feels like something is wrong?

[honest jay]

Start a new chat and open up the GPT-4 dropdown menu in the model selector. You'll see DALLE at the bottom there, select it and you should be good to go!

[forest tundra]

anyone gpt4 going slow, im trying to use plugins and it keeps freezing and crashing the tab. Tried different browsers too

[inland summit]

Oh my goodness. I didn't think a left to right selector would also have a drop down selector. High five for hiding it really well 🙂

[modest tundra]

Is it just me or did Dall-E 3's content safeguards suddenly go into overprotective Karen mode?

[honest jay]

You are definitely not the only one there! I think it would be a great place for a "peek" idle animation type thing--some kind of hint to users to go see what's down there

[prime swallow]

It tried to use LaTeX and made a mistake.

Tell it to do this:

$$
formula stuff
$$

And that the $$ have to be on new lines by themselves.

[simple pasture]

They aren't gonna get neutral test results when they didn't activate that option for me

[clever minnow]

I just updated my chatgpt on Android and suddenly I can no longer add images or choose the got version/mode. Anyone?

*gpt

[honest jay]

What version did you update to? Do you have a VPN installed, or are you perhaps using a non-root adblocker app that might be affecting your DNS?

[clever minnow]

No, nothing of that sort. I updated to 1.2023.306

[honest jay]

Hmmmmmmm can you try clear cache? And if that doesn't work, clear data, log back in? And if that doesn't work, uninstall/reboot/reinstall? Do the features show up fine for you on chat.openai.com? I gotta run for a bit so just dropping a couple more suggestions in case you have time: same behavior on and off WiFi? Are you a new Plus subscriber by chance? #community-help message

[clever minnow]

Thank you, i hadn't done that. That does the trick! All working now.

[honest jay]

🤘

[sick pecan]

chatgpt is still awesome today. super fast, seemingly no cap

i hope this is the new norm

[wise yacht]

DALLE issue today: keeps saying “stopped” before eventually displaying the image I requested

[flat tinsel]

I’m getting this now too. Hmm.

[south raft]

Half the time the page totally doesn't load for me and I can't upload images, there's a javascript error so I guess it's not just me

[trail juniper]

what’s the point of buying plus subscription if you can get everything in plus for free

What even worse is that new account also have a chance to get it

[hardy frigate]

is the image feature working for anyone?

[prime swallow]

Hey, do that if you want.

From what I've heard, a number of folks with all tools report some issues and problems, some even want to switch back to the separate models.

This early in the game, you may not want it as much as you think; it may not already be working perfectly.

That said, how long do you think the free accounts will retain free access, once it is working well? 🙂

[prime swallow]

Yes, I just made this in answer to someone else having trouble with it though #off-topic message

[ionic geyser]

i like chatgpt but it is sometimes just so annoying

like i have a long chat and its denying to write a poop poem but in a new chat it will do it with no problem

so annoying

And the only way to bypass a bug like this is considered jailbreaking

like i just want it to work

[sick pecan]

local llm is better for poop poems

so my conspiracy theory is chatgpt coding improved because phind is now available

[ionic geyser]

thats hard to set up

[sick pecan]

naw theres heaps of free to use local llm online, you just gotta use an answer engine (not google, its crap)

[prime swallow]

It likes us too! Here's my answer to you 🙂 #1070955045639684096 message

[arctic osprey]

I want to be proficient in ChatGPT4. I feel that I'm only scratching the surface with prompt training. Plugins all want additional money for full access. Whats your opinion on how to make useful tools with GPT4?

[prime swallow]

Tools to do what? Know what you want the AI to do, explain that clearly, and you're in great shape.

[mint karma]

'using chatgpt to make itself a prompt, chatgpt too easy'

[frosty plover]

Ah, I just realized today that ChatGPT Plus allows for multiple users to share access. If you ask questions at the same time, it won't override another person's drawing? Does that mean we can split the cost with others in the future? Is there a risk of account suspension for doing this?

[shut gulch]

Help me out guys. I asked ChatGPT to create an image for me using dall-e 3. It gave me detailed prompt to put in dall-e 3 instead. So I went to the dall-e 3 option in gpt 4 and gave it the prompt back. It now says "I'm unable to generate images based on the provided description due to content policy restrictions. If you'd like, please feel free to adjust or provide a new image concept, and I'll be happy to assist you with that."

[winged sierra]

it always says it.....

[honest jay]

What's the prompt?

[shut gulch]

Wouldn't wanna share that

[honest jay]

Well there's ur problem hoss

[prime swallow]

Why do you believe that ChatGPT+ allows for multiple users to share access?

[prime swallow]

Try talking to GPT/dall-e like you initially talked to ChatGPT.

[wanton lichen]

Wait so what's going on with chatgpt? Seeing a lot of people say that a recent update has made it really bad

And what's going on with bing integration?

[crisp cairn]

so this prompt violates policy, you can prompt it to modify it to not violate policy, or create image similar to this prompt not violating policy

[shut gulch]

It doesn't violate any policy. I'm just telling it to create a graphic for my tech business

Anyways bing ai was able to create the image for me

But chatgpt 4 with dalle 3 still can't

[prime swallow]

They're not the same AI, different rulesets to follow for sure.

ChatGPT/OpenAI is being very respectful of IP. However, it claims that if you are clearly doing something that is original, it can help (keep in mind though, it cannot always fully or correctly describe the rules it actually must follow. It's a bit like a person being unable to predict they will sneeze in a few minutes):

https://chat.openai.com/share/113e50a2-21ad-4055-85a6-e1f73345f2f8

[wanton lichen]

I cannot get chatGPT 4 to use Bing, what is everyone talking about?

I don't have the "Browse with Bing" option anymore so I must be part of the rollout

[sleek swift]

what happened

I'm a paying member and it says I don't have acess to gpt 4

the subscription has like 3 months already

and I didn't cancel it

it is back

[feral zinc]

its a test and not a permanent thing, probably will be removed after monday keynote or sometime later

[solar barn]

please undo the massive downgrade of gpt

[grim mason]

Is anyone else getting errors when using dalle 3 in ChatGPT? I'm getting:
Error creating images
Which actually crashed the app when I try use it on mobile. Any known fixes?

[maiden fractal]

Does it work in web versions

[grim mason]

Nope, so the only way I'm able to see the error message is if I go onto web versions, every time I try go I to those chats on the app it crashes it

[maiden fractal]

Can you try a fresh conversion on web and try yo generate

[grim mason]

Already done, fails every time no matter the prompt

It's really bizzare and I guess not a common issue 😅

[maiden fractal]

I suggest you post in #1070006915414900886 give details of everything you tried with screenshot of error

[grim mason]

Cheers will do in a bit, tysm

[mint vapor]

Has anyone noticed that the descriptions accompanying DALLE generations in ChatGPT are no longer included/missing?

[elder pollen]

Since a whole while already yes
You can just ask it to share them, or share the precise prompt it sent to Dalle

[floral ether]

@mint vapor when you click on one of the images it generates, it will show the text also.

@mint vapor If you dont see it, click the (i) in the upper left corner when a image is showing.

[mint vapor]

Thanks.

[stable dock]

actually quite annoying that the default seems to now be that thy are hidden

[frosty plover]

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?

[dense halo]

Guys why is the voice recognition so bad these days?

[halcyon widget]

Anyone know how do i access alpha version ( all tools ) of ChatGPT?

[maiden fractal]

You wait till it is release . Follow, #announcements

[wicked nova]

Does the chatgpt enterprise use the same interface as ChatGPT? Has anyone here used the enterprise version of the ChatGPT? A quick google search tells me that the pricing is about the same as ChatGPT plus. Can anyone here confirm the same?

[warm gull]

I have the GPT(All Tools) feature. Is this temporarily ?

[elfin birch]

There's nothing official said so I would hope for the best and prepare for the worst
Though I haven't seen anyone say their access was revoked 👀

[maiden fractal]

Some lose access

[wicked nova]

Is this feature on your web interface or iOS application?

[warm gull]

Web only it seems

[maiden fractal]

Good to have , if you lose access later dont be surprise

It is alpha test selected randomly

[warm gull]

Oh ok

[wicked nova]

Can you explain the (all tools) feature for us who don’t have it yet? Does it use all the available/installed plugins? Does it also allow advanced data analysis along with plugin access?

[warm gull]

Yes, seems like it - it does Dalle-3, web browsing, code interpreter etc etc all in one

[forest shore]

Does anyone know when we're going to have access to the "browse with Bing" drop-down??? It's been days, and it seems nobody has an answer!

[maiden fractal]

Start the chat on mobile and you can continue on browser. It is a bug atm

You can use plug in, web browser as alternative too

[forest shore]

Thank you. We'll see....

[wicked nova]

Just use the web browser plugin imo, the default bing search on web and iOS only seem to look/read exactly one web page for some reason at least for me. Doesn’t seem very useful if you’re gathering info from just 1 webpage

[warped dew]

what plugin would you recommend?

[rich elbow]

I am a ChatGPT Plus user and I still cannot access the All Tools update? Anyone know why? I asked ChatGPT (web browsing) and it told me it should be rolled out to all Plus accounts. Is this true?

[wise yacht]

That DALLE “Stopped” message is quite annoying.

[halcyon widget]

Can you msg me?

[prime swallow]

Nope, hasn't even been announced yet.

Some do have it; many who say they have it also report some issues with it. It seems very new and early, and maybe not working great at most times yet.

My personal guess and that's all it is, I don't have any inside info, I'm just a community member but it seems mostly folks who are not plus users currently talking about access to it -

The early plus users who tested it and talked about it that I saw, most seem to have reported the problems made it worse than the more polished separate models we're used to (if you're plus).

But the free account users have no access to GPT-4. So some of them maybe are being allowed access, they are gaining, not losing, access to GPT-4. Even if it doesn't work great and there's a lot of bugs, the free users still have their normal GPT 3.5 access, and then can alpha test the all tools, that may be quite buggy.

Meanwhile, the plus users still have access to their normal GPT-4 separate tools that work well, and only 'suffer' from anticipation and excitement, but don't see a model that isn't currently fully ready for that kind of testing.

I figure some + members will test it later, as is useful for the company.

[halcyon widget]

Please answer my msg

[warm gull]

just gimme some time, im just busy with something rn

[main ibex]

I also got the alpha test thing, does it have the same capabilities as the GPT 4 premium?

[winged moth]

I don't know. I had it myself for a little while yesterday but now it's gone. Wheee.

[main ibex]

Does it just last for an hour or a day or depends?

[rich elbow]

Thanks!

[winged moth]

For me it lasted a few hours I think. I barely got a chance to use it cuz I was on a work day. But after it was done, I ended up with chat GPT that has knowledge up to April 2023 so I'm still happy! 🙂

[main ibex]

Nice!, does normal users also have GPT knowledge until April 2023 or just the Alpha test people?

[winged moth]

Well I was an alpha tester for a few hours, I think, but now I'm back to being normal, except I have the increased knowledge. So I don't know if that's true for everyone but that's true for me.

[prime swallow]

That's so hard to tell. See, if you ask the model the same question, sometimes it gives different answers.

I don't think it's trying to lie or even hallucinating. I think there's more than one answer to the question in its training data.

LLMs don't necessarily 'easily forget'. If they were trained that their knowledge cut off was Septemember 2021, and then you update that model and tell it, 'hey, your knowledge cut off is now January 2022'. Well, it has both sets of info. So how does it answer the question?

Being an LLM, it tries to predict a reasonable answer that fits the question as asked. it grabs some training data each time it answers, and decides what answer to actually say: So the answer can change each time you ask.

And once it 'knows' that there's more than one answer, then it may even make up other similar answers, because that's often rewarded (and useful for fiction, and what's real anyway, to an LLM?).

Like, I once asked 3.5 about a common color for an apple, and it told me orange. I checked further, turns out there are multiple kinds of apple named orange, and very rarely some of the shades of yellow and red can blend to make a somewhat orange-colored apple. It's not a full hallucination, it's just a rare answer from a small part of training data.

But it can answer any color to that question - and it can answer any date for its training cut off.

[arctic kraken]

Anyone have a direct link to the termination appeal page? Their help / support section isn’t much help on the site

[prime swallow]

That literally is the direct link. There's also an email for support, but:

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]

That's how you reach out to OpenAI. Good luck.

[main ibex]

I did what you said and it told me two different answers, gpt 3.5 told me the last cutoff knowledge was January 2022 while Alpha test told me it was April 2023

[prime swallow]

And what happens when you ask the same model the same question a few times?

[dense halo]

Why is gpt4 so dumb now? It doesn’t even pay attention to my custom instructions anymore???

[arctic kraken]

I think the issue was the bot isn’t showing up on the mobile page and the mobile page is being wonky lol. I can’t attach screenshots here but if I go to it it’s asking me to download an en file lol

[main ibex]

I got this for alpha testing while repeating the question:

My training data includes information up to April 2023. If you need information beyond that, I can look it up online if it's available