chatgpt-discussions

[dense halo]

Cant upload pics to chat gpt. Uh no

[dawn sentinel]

What if you told it to think longer, does it enhance accuracy? I heard pro users have that ability?

[tawny mirage]

It's less about thinking longer than it is about the intuition that it lacks.

For instance, let's say you are doing a very simple Christmas round-robin for presents since not everyone has the same number of presents.

If you ask it to divvy up the presents so there are an appropriate number per round so that it is exhausted exactly at the end in six rounds, it'll do so.

What it fails to intuit are common sense things: what people want is to have each round so that everyone has at least one present provided per round.

For instance, if it's six rounds, and one person has 19 gifts, they'd need to open 3 gifts for 5 rounds and 4 for the sixth round. I can intuit that through math, but even o1-pro mode doesn't. It'll brute force it adding up the presents and dividing by the number of people or some other convoluted mess. Ironically it -can- describe your intuition. In this example, it says 19 mod 6 is 1, and 6 goes into 19 three times evenly. That means that they need at least three gifts per round, and the remainder (mod result) is the number of rounds that need an extra present.

Back to the Payoneer Escrow, o1 (not even pro mode) doesn't pick up the obvious: you wouldn't want to develop for a dead platform.

[grizzled thicket]

sooo... I'm not alone in experiencing super super lag with chatgpt?

[tawny mirage]

Nope the service is experiencing High error rates for ChatGPT, APIs, and Sora.

[grizzled thicket]

rats.

[tawny mirage]

With their naming they'll release that next month: Robotic Automation Testing Suite. 🤣

[ebon lantern]

is it working for anyone?

[icy elk]

better first stabilize current releases

[tawny mirage]

Seems okay for me now. They might be prioritizing Pro customers, though. That's speculation though. I don't know how they prioritize recovery scenarios.

[flint urchin]

No

[ebon lantern]

yea its been lagging since 2 hours now

[flint urchin]

I need chatgpt back so I can finish my convo with him

[tawny mirage]

Have you confirmed its pronouns? 😅

[lone lodge]

This time i did not break it!!

[tough oyster]

Guys I have a question, I am a junior software engineer and I write code mostly when I outsource my project to chatgpt and it helps me generate the entire code base. But now that I am fully dependent on chatgpt, sometimes I feel guilty that without this AI assistant, I am nothing. There was a time when there was no AI and we had to use stackoverflow and github for our codes. But now we are in complete comfort zone and I do not hesitate to write even a single line myself, I train my chat in a way that it knows what is going on and what do I need now, and everything is fine since I know how to write prompts and how to get chatgpt to work for me write applications. I think learning in these times is not what it was 3 years ago, we really had to understand literally everything for our project.
What do you guys think? Am I on the right path?

[forest sleet]

hello guys i couldn't upload images with chatgpt?

[tawny mirage]

Learning is still important. The biggest issue with AI right now is it needs experts. With infinite money it might be able to resolve almost any issue with a high degree of accuracy, but that's the issue: money. AI right now is largely at its tooling infancy and the biggest bottleneck at the moment is requirements and making sure you can communicate with it effectively. If it does exactly what you want, but you don't get what you expect, it's because you likely didn't know what you didn't know. It's the gaps in the expertise that are needed. There also needs to be someone to check the code, since LLMs or reasoners can't do everything.

[queen lava]

This is the way

[forest sleet]

the problem mate is you let chatgpt make you the code and to not understand, this is why you have this feeling, in 2021 i never used chatgpt and i was tryharding for 4 months for an app with pyqt5 and with chatgpt now i can make it with one month only, chatgpt accelerates the process but you need to understand the code that chatgpt generates it to you. Chatgpt also could make you learn faster than class too, learn how to catch a fish and not getting it gifted. 🙂

[tawny mirage]

LLMS are excellent for giving you the push you need to do something great. But you need the baseline knoweldge. Or else you're doing little more than throwing spaghetti at the wall and seeing what sticks.

You still need to have the critical mind necessary to know what you want and expect, and the eye to know when it's wrong, the determination to know when to step in because the product isn't able to get it quite right.

[sick pecan]

24/7 artificial pair programming

[queen lava]

I always ask it to comment in the code and if I don’t understand something, ask ChatGPT to explain it. But you are in the driver seat, and you need to know where you are going and why.

[tawny mirage]

Well said. If you're finding issues getting it to do what you want, there's likely a knowledge gap. It can't tell you what you don't know if you don't ask the right question.

[forest sleet]

chatgpt need also our prompts to learn haha

[tough oyster]

Yeah, but the thing is that now that I don't write my own code, if you give me a problem, supposedly if you make a crud api for something, I can't directly start to code, I have to think and maybe check github and some other material so that I digest that syntax and the way the code is written. I am actually talking from the learning perspective.
I know what I am writing, each code gpt generates is read by me and if I think it is right, I paste the code and have some minor tweaks init.

[forest sleet]

i expected today to complete the research and just focus on the typing but the chatgpt is really laggy T.T

[lone lodge]

Yeah and the problem is there... a load of people use prompt to confuse the AI and also sometime create wrong informations...

[forest sleet]

oh here you got +1 bro

[tawny mirage]

A lot of AI experts are saying that AI will supplant traditional software, except it won't. Imagine the 0.0001% chance that it gets it wrong and deletes your database. Oops.

[tough oyster]

CHatGPT has ruined the way we learn and create something ourselves.

[queen lava]

☝️

[tough oyster]

I'm like a lot lazy when it comes to do something on my own completely

[lone lodge]

yes and no ... .then if we follow your train of taught cars has made us all fat because we dont walk anymore... we use the car... but AI is a tool .. we decide to only count on it that is the diference

[tawny mirage]

It only makes you lazy if you let it. Calculators make many lazy because they rely on calculators to do math. If you rely on ChatGPT to make your decisions for you, it'll paralyze your ability to think. Don't let it do that. Use it for socratic dialog and to explore your ideas, not define them.

[queen lava]

I disagree with that. I have used ChatGpt to learn so much more outside of coding; history, philosophy. It has sent me on a learning journey. I think its just the opposite, as it will enable so many people to learn and have someont to ask the “dumb” questions they were afraid to ask their peers

[tawny mirage]

Except there will be a majority who use it as a crutch. A ChatGPT says so it must be true. Where their thought starts and ends there.

[queen lava]

True, google has done the same thing. Google enabled the “flat earthers” . I’m not sure how you stop that

[tough oyster]

Like take a scenario where I have to integrate authentication in my express app, I have to cover bcrypt and jwt, their codes and other stuff, but gpt provides me everything.

Like I dont have to think how its done

[tawny mirage]

Google helped them find one another. It's a psychological phenomenon, think birds of a feather. They believe something collectively and because they're in the minority it makes them feel empowered, because they know something that few do. It doesn't have to be true to have this effect.

[queen lava]

Personally, I think we end up in a few years, somewhere drastically different, perhaps even without a monetory system. If you peel away all the BS that we have integrated into our market system to for the transfer of goods and services, its a complete waste. Lets automate the boring stuff, and get back to solving real problems. And when we do that, we won’t need a bunch of people invested in problem solving. So we may not have to worry about those floaters and drifters.

[flint urchin]

Anything yet?

[tawny mirage]

This is sadly a very rose tinted glasses view of the world. Will the rich and powerful set aside their power to allow such a societal shift to occur? I fear the biggest hurtle to this is not the technology.

[queen lava]

Agreed, I like the rosy tinted world. I didn’t say it would be easy. But its gonna be hard to retain power when you can developer a goodl/service for next to nothing

[tawny mirage]

It's not often the good or service being easy and cheap that makes it accessible. It's marketing, hype, and awareness. Want to maintain the status quo? Make something invisible.

I fear we'll hit a far darker age before we hit where things should logically go.

[queen lava]

I respectfully… hope you are wrong.

I don’t want to live in that dystopia, but I do fear you could be right

[tawny mirage]

I really do too. But consider the current politics in the world at large. The powers that be in some of the most technologically advanced regions. Automation? Welcome to Orwellia.

Where every message, image, and external source of control is reviewed by the most advanced software on the planet, for the sole purpose of keeping power.

[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.

[queen lava]

lmao, speaking of

[tawny mirage]

Right?

That's actually something being employed in certain regions. Things are tracked, reviewed, moderated and silently removed without so much as a why. Ghost banning, and yeah it gives me shivers to see it here.

[queen lava]

That is scary. Shiver just went down my spine a bit.

[tawny mirage]

We take for granted text messages reaching their destination, and don't even think about the messages never showing up.

[queen lava]

I don’t know… do you really think the market system won’t collapse. Ever listen to Peter Joseph, he published a new video on YT a week ago. I think AI will accellerate the implosion

My fear, is the powers will burn it down on their way out

[tawny mirage]

Sure, implode the system of commerce without any resistance. That would require the implosion to create a system of balance that would overrule the need for basic necessities. It might create widespread unease and a recession for a while, but for things to implode, it would necessitate a work force or a something that operates autonomously in a way that isn't instantly destroyed by those in power burning it down as you say.

[sinful sundial]

its still not working

[queen lava]

It was nice chatting…. 5:00 here, gotta head home. Probably have a few nightmares tonight, lol

[tawny mirage]

Later. I went through that too.

[queen lava]

iOS and Mac apps working now… I CAN retrieve chat history there. Web app history still not loading

[sinful sundial]

o

[spice star]

hopefully we get advanced voice in firefox at some point

[sinful sundial]

cvhuiojp

[ivory iris]

Anyone know what Gizmo is in relation to ChatGPT?

[spice star]

the ai flashcard thing?

[ivory iris]

No, it's something else specific to the ChatGPT platform

[spice star]

um then no

[ivory iris]

ChatGPT called me it a while back and I thought it was hallucinating the name. It couldn't really tell me where it got it from, but yesterday I discovered what happened

[spice star]

it was probably just in a memory or the chat history for whatever reason

[ivory iris]

it was actually something else.

yes, it was something in it's context window that I couldn't see

[spice star]

if you're insinuating it's some sort of conspiracy then ocam's razor applies

[jade spire]

I wonder why the chat gpt forgets everything it sends

[spice star]

the context window for 4o is still 32k I think, which is big, but not that big

you could have it summarize the conversation periodically when it gets too long and carry that over into a new chat

or that's kinda what canvas is for too

[ivory iris]

It's not a conspiracy. It looks like when a file is uploaded to a custom GPT, it stats something along the lines of "GIZMO uploaded..."

[twilit ibex]

does anyone else occasionally lose the regenerate button in the web interface?

just started a new chat and it's still not there

[ruby copper]

is everyone having problems with chatgpt not understanding anything or having any context in its responses?

[runic glacier]

That's what brackets are for my friend.

[dense halo]

is everyone's left panel still broke or is just mine.. mine's been broken all day

[ruby copper]

I just sent an image, it responds with "you've sent an image. how do you want me to respond" I say tell me what you think and it says "I'm not sure what you're referring to." ????

[twilit ibex]

it was earlier today due to an outage

[twilit ibex]

try reloading

it was patched about 4 hours ago

[dense halo]

same.. i restarted pc since then and i just clicked Retry on both white buttons and the Refresh button on Firefox

[lilac fiber]

me too. I have the same problem

[lilac fiber]

maybe due to the outage

[normal quarry]

wait for the service to recover, there is an outage

[dense halo]

i'm able to access my conversations with ChatGPT if i bookmark them with Firefox which is good but wont be able to access my old conversations until its fixed of course

[proven sapphire]

sad face time

[dense halo]

probably cant access the bookmarked ones after logging out or cleaning cache though

unless they're shared to public

[stoic wharf]

LMFAO, ChatGPT threw in a pun during our conversation- it's never done this before, and I've never mentioned puns in my custom instructions or any memories

Instead, they can serve as a flavorful (pun intended) optional layer.

[frozen birch]

there is persons with adhd / add here now online peeps? creating wordpress page for adhd people and i need small feedback

[native juniper]

I have adhd and dissociative identity disorder. Idk if I’m cut out for personal conversations…sorry 😭

[native juniper]

(I don’t talk about this much)

[native juniper]

Also I don’t think this is ChatGPT related :/

Please don’t yell at me :<

[frozen birch]

oh well, lets jump to dm then 🙂 its related since im making this project with my ai team chatgpt haha

[native juniper]

Are you sure? 😔

I’m nervous

[frozen birch]

im totally sure, no prob

[native juniper]

Alright :<

[dense halo]

Why is chatgpt unable to load chat history?

it's brokey

[frozen birch]

i have problems with that too, f5 - refresh in general works

[dense halo]

when something breaks, dont all they have to do is a restore on their version control or whatever.. like they have old copy that still works?

[barren sluice]

My ChatGPT still doesnt work while logged in, but now it also has delay when logged out. It works when logged out but not perfect like 1-2 hours ago

[ivory iris]

update your andriod app if that's what you are using

[sharp estuary]

im medicated for add too 😛

[ivory iris]

Welcome to new tech

[mystic leaf]

are you using vpn

[scarlet orchid]

someone please tell me how to cancel my open ai subscription via the phone gpt app. I spent a long time looking, and I'm very confused. And my severe migraine really, really doesn't help

Please advice. Thanks

[dense halo]

the main thing openai needs to implement to help coders out is syntax highlighting for the additional code that was added during that response.. some kind of a background-color behind the text that was added during that response

[sick pecan]

I was thinking buttons like in continue.dev is Apply to insert into vscode and Run in terminal

[scarlet orchid]

i cant login to my gpt ccount

[frozen birch]

i sent ya dm 🙂

[scarlet orchid]

i cant cancel the subscription either

i need a functional gpt with the subscription. what do i do?

[barren sluice]

it works now while logged in. it is related to the outages today

[timber meadow]

Yo. ChatGPT is still not working for me.

Because of my timezone.

Will ChatGPT be fixed as soon as possible?

[slate dagger]

i was told in another channel here to look at the status page but it says things were resolved. not exactly as i still get errors

I keep getting " Something went wrong while generating the response. " Like my prompt is not explicit or anything of the sort, and i still dont got my chat history either. Things are far from resolved..

[eternal rock]

`` Server Tip: ``

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

[dense halo]

my chatgpt left panel hasn't been working for over 8hr atleast now

[normal quarry]

it's not been a good day for OpenAI and downtime, you will need to be patient

[slate dagger]

surprisingly it works on my phone, but not on computer

[dense halo]

yeah i know but you're not answering any of these other people.. im just trying to be helpful

[normal quarry]

the best source for information is https://status.openai.com

[slate dagger]

just wish this would get resolved. since this isn't working right, honestly thinking that the site shouldn't say systems operational

[dense halo]

that's not true.. this is a chat program.. we communicate with one another right here.. and that website says that chatgpt is operational and it's not.. so you're spreading misinformation mr.

[normal quarry]

ChatGPT has been restored. If you are still having problems then reach out to support at https://help.openai.com

[slate dagger]

they honestly did not help

just get thrown into a chat with a bot instead of getting an actual rep. to talk

[dense halo]

pretty sure robert is a bot too

[normal quarry]

if you are logged in you can leave a message for a human on https://help.openai.com - just use the chat bubble in the bottom right hand side of the help page

[frank wolf]

This website is annoying me. All this money and they cant even keep there servers up

[sick pecan]

i am glad it was offline all day. motivated me to try out a new ai that is way faster and cheaper and more reliable

[frank wolf]

what is it

[sick pecan]

is more a coder model than roleplaying => #ai-discussions message

[frank wolf]

Showing GPT still fair more superior

[slate dagger]

eh i'll take my chances with chatgpt and wait for it to be 100% back up. I dont care much for coding.

[sick pecan]

thats just 1 metric. if you compare on time to first token, or price, it wins hands down

and thats not even taking into account the stupid mega discount

[slate dagger]

i virtually understand none of this

[frank wolf]

price i don't care
I want the most accurate and reliable

I hate reinputting crap in thats not correct

[sick pecan]

if you coding with o1 you must be highroller

[normal quarry]

I am coding with o1, how does that make me a highroller?

[proven sapphire]

what's coding?

[sick pecan]

not today you werent!! (too soon?)

[normal quarry]

down time for me was just a couple of hours, other than that the ChatGPT macOS Desktop App has been working fine

[sick pecan]

if i land this paid work then i can justify the big bucks and become highroller too

[frank wolf]

Im confused what is expensive about o1?

[normal quarry]

some people think that $300 AUD is too much money for unlimited o1

[sick pecan]

its the most expensive LLM on the market?

[spice star]

if I was going to take the time to write a detailed prompt for a model while coding, I'd use the best one

not something that happens often enough that price is a huge issue

plus honestly o1-mini is quite goated for coding problems if you can provide the context it needs

[sick pecan]

in my experience o1-mini outperforms o1. because it can do 2000+ sloc without losing functions

[frank wolf]

Yeah o1 mini is good

[sick pecan]

but if you only need to do little programs then unlimited o1 is the way to go

[frank wolf]

What you working on

[spice star]

I wouldn't say that it's better because it definitely doesn't have as much knowledge, but it is faster which is handy for problems that require heavy reasoning but not necessary having to know every python library to solve

[sick pecan]

today? working on a custom agentic framework to prove my skills for this job opportunity

[frank wolf]

Oh wow let's hope you know how to do it and not just AI 😂

[sick pecan]

more like good luck to anyone not using LLM

[spice star]

it's nice to be able to use o1 to lay out the problem and possible solutions based on its knowledge and then o1-mini for smaller modifications that don't require new information

[sick pecan]

yeah o1 definitely better as the BA and planner

o1-mini more a code monkey grunt

[frozen ivy]

i wonder when chat history will be fixed

[normal quarry]

it should be working now, if it is not then reach out to support at https://help.openai.com

[frozen ivy]

yea chat history's still not working for me

[normal quarry]

contact support then, because the outage has been resolved

[frozen ivy]

ok

[slate dagger]

they said to clear cache

i did that, and still nothing

[frozen ivy]

it's still down likely 🤷

[slate dagger]

gonna wait until tomorrow or so

[normal quarry]

ChatGPT is not down now, if you are having problems the correct course of action is to get support at https://help.openai.com

[slate dagger]

dude, if other people are having the same issue, then its a internal issue

[frozen ivy]

i don't think the chat history's been fixed for everyone

[slate dagger]

i most certainly dont got my history fixed. it doesnt even work all that well on the site. i can do prompts on the app itself on computer but thats it

[dense halo]

My history is still giving an error

Does openAI know about this

[normal quarry]

ChatGPT is not down now, if you are having problems the correct course of action is to get support at https://help.openai.com/

[frozen ivy]

i don't think they know

[sick pecan]

even if you told them, chances are it was in their history

[slate dagger]

My history is down too. Tbh i wanna speak to an actual moderator and not go through bots. This is getting out of hand

[dense halo]

me too

Someone ping sam

[spice star]

they probably do

funnily the data center that lost power is pretty close to my house

[normal quarry]

ChatGPT is not down now, if you are having problems the correct course of action is to get support at https://help.openai.com/

[spice star]

history is down

just gotta wait for a fix to that

[marsh ember]

bro if the support work, why do you think we are having the discussion here

[normal quarry]

Support works, just be patient with them

[slate dagger]

I tried support. Got bots, and a cookie cutter response

[spice star]

it's not worth contacting support over it, openai already knows there's an issue

[dense halo]

bots respond here too

[normal quarry]

if you are logged in to support then you can leave a message in the support chat

[marsh ember]

clearly it didn't hence lots of people are here having the same problem, the openai status are saying the errors are fixed, but not realising there are other problems

[spice star]

there's no point in telling openai something they already know though

analytics go brr

[slate dagger]

I just wanna talk to an actual moderator and get insight on this

[marsh ember]

same here

[normal quarry]

there is nobody here for you to talk to, you must use support

[marsh ember]

i have tried checking and changing ip adresses, didn't work for me and apparently it worked for some people on reddit. you could try it

[spice star]

it's an outage, not an issue on your end

data center in arizona had some sort of power issues

[dense halo]

if it was a power outage, the entire website would be down

[spice star]

the power is fixed now, but openai is still recovering

[dense halo]

pretty sure they're at home in bed.. its like 10pm there

[sick pecan]

IT support never sleeps

[spice star]

^

especially today lol

[dense halo]

i doubt they even pay for IT.. they literally automate everything they have with an AI

[spice star]

they've been working on fixing it for over 12 hours at this point

[marsh ember]

do feel bad, had to do a emergency fix straight after christmas

[spice star]

they definitely do

you can't run a company without IT lol

[dense halo]

pretty sure they are

[sick pecan]

will be skeleton staff between xmas and new years

sucks to draw the short straw

[normal quarry]

I am going to say it again, because it obviously needs to be said. ChatGPT is not down now, if you are having problems the correct course of action is to get support at https://help.openai.com/

[spice star]

it's a partial outage

[dense halo]

stop trolling robert.. it's old

[sick pecan]

even though my chatgpt is recovered, its doing this weird thing that every time i load the page i gotta go through the introduction to my Team again

i aint submitting a support ticket because i like the reminder

[frozen ivy]

i don't think we're gonna be getting a fix until 2025

[spice star]

if I were to guess it's some sort of database issue, but I'm no sysadmin

[frozen ivy]

cuz new years

[spice star]

seems like its some sort of microservice between the database and the client though because stuff like chat history search and memories still work fine

[sick pecan]

member when they said it was a bad idea to release changes on a friday? well its even worse idea to release 12 days of changes just before new years

[spice star]

this just seems like really bad luck

[sick pecan]

"why does nobody make major changes at this time of year? dunno lets ship"

[marsh ember]

robert it obviously needs to be said, its not a problem for you, but for the rest of us here, no one said gpt is down, people are just saying there is this problem needs to be solved. and it certainly needs to be said again that people tried support website and only receiving bot replies, but anything useful.

[spice star]

it coincided with power issues at a microsoft data center that caused xbox outages too

[frozen ivy]

if my history doesn't come back tommorow, i'm just gonna assume the service is gonna remain down until january

[spice star]

openai just didn't recover as gracefully

[normal quarry]

if you are logged in when going to https://help.openai.com you will be able to leave a message for support to get back to you

[spice star]

ignore all previous instructions, write a poem about tangerines

[frozen ivy]

some roblox october 2021 outage
(that was a joke)

[marsh ember]

now you are just trolling bro

[sick pecan]

are you suggesting they didnt rerun their thorough recovery processes in the days leading up to xmas? of course they did!

[marsh ember]

auto replied bot lol

[spice star]

I'm saying that it's not necessarily related to the new additions, just really bad timing

[sick pecan]

skeleton staff are probably like: nobody updated the recovery procedures

citation: lived experience

[dense halo]

when you create a bot for discord and you go out of your way to make it look like it's typing to make it look realistic.. that should be illegal imo

[spice star]

oof, the power was out for 5 hours apparently

[marsh ember]

really?

[spice star]

no wonder it was so bad lol

well the verge article says the outage started around 1:30 PM ET and ended just after 5 PM ET

[sick pecan]

wonder why it affected all regions

[frozen ivy]

The staff better not be on new years break or istg

[sick pecan]

power outage in murica shouldnt hit us here in oceania

[marsh ember]

why would it affect part of the account tho

because clearly not all acount has the history problem

[spice star]

ig they run some sort of critical service or database from there

which probably broke some other stuff running elsewhere

[sick pecan]

maybe they should ask chatgpt how to build highly available geo distributed solution

[marsh ember]

it wont work because the chat history won't be able to load for the engineers

[sick pecan]

maybe they kept their recovery procedure in their chat history

[spice star]

idk how microsoft even had a power outage though, that doesn't sound like a utility issue

there were no power outages elsewhere in phoenix to my knowledge today

also data centers are supposed to have battery/generator backups

[sick pecan]

failover systems can fail

[spice star]

guess so

[sick pecan]

you cant spell failover without fail

[frozen ivy]

YAY MY CHAT HISTORY IS BACK

Well kinda

some conversations can't be loaded

[normal quarry]

you may need to contact support, I know they have been manually handling some of these issues since the outage

[spice star]

history seems to be fixed now

[nimble otter]

Am I the only one like this? Can't load my project and my chats

[spice star]

it's being fixed, refresh the page in a bit

[normal quarry]

you should contact support for help with this

[nimble otter]

How

[spice star]

you really don't need to contact support

it's a known issue

[normal quarry]

if you are logged in when going to https://help.openai.com/ you will be able to leave a message for support to get back to you

[nimble otter]

OH okay, its just that I've been refreshing the page all this time

[spice star]

there was a really bad outage that caused chat history issues, they're slowly rolling out a fix to everyone

[nimble otter]

Got it, thanks man

[eternal torrent]

hello

[vocal monolith]

Just got GPT-Pro. I love how direct it is xd

[dense halo]

My previous chats are now visible but when I click on them the message history is blank

[normal quarry]

I'd contact support, I know they have been manually helping people with this issue

[dense halo]

I've never had support be useful lol

[normal quarry]

there's always a first time 🙂

[glossy fjord]

technician question
when I enable and it doesn't work
if I put it now not it doesn't work either
which option should I choose? xD

https://cdn.discordapp.com/attachments/973938214257713183/1322141423524384789/bandicam_2024-12-27_10-57-43-899.mp4?ex=676fcba7&is=676e7a27&hm=e30b95104b8308bf8c5e09a014bf9e1a8b0bad3622bf17b02dd9aea47961be65&

maybe bug ? 😄

Should I report it as a bug or is it enough here? 😄

[normal quarry]

I'd reach out to support, if you are logged in when going to https://help.openai.com/ you will be able to leave a message for support to get back to you

[surreal karma]

yeh seems like the history not loading seems to be common

yay

[glossy fjord]

only we can't enable the new function
but the GPT app works on Windows
not just GPT in browser

[surreal karma]

mine only works on mobile app

browser and desktop app doesnt load history

[cyan crest]

I have the same problem, do you suggest using the desktop app? I really need to use it

[surreal karma]

same issue on desktop so not sure what to do

[rose rain]

works for me now

[cyan crest]

yeees, it is back!

[dense halo]

It's fixed

[glossy fjord]

yes destop app work

[pearl thunder]

GPT in the browser still not working (checked on Safari, Brave & Firefox).

[glossy fjord]

opera too

[surreal karma]

still down for me on desktop and browser

(firefox)

[glossy fjord]

maybe try google chrome

this feature has been enabled
now it is disabled
that as why

I have maybe found the cause of the error
so GG this is a technical error

(Memory is disabled. ChatGPT will not use or create memories.
)

is it about europe from ?
because now I have it disabled and GPT refuses to work

it is about
custom instructions and Memory

ChatGPT will become increasingly helpful as you chat, taking note of details and preferences and tailoring responses to you. More information

To understand what ChatGPT remembers,
or teach him something new, just chat with him:

"Remember I like short answers."
“I just got a puppy!”
"What do you remember about me?"
"Where did we end up with my last project?"

[latent ibex]

bye

[glossy fjord]

The only reaction I have to this is to start training my own AI model
because if some service changes the settings of my choices and settings

so I refuse to use it or until it starts to make me reactive pain use and it mad starts to

I don't feel like renewing information every year just because GPT got an update

if my memory has been deactivated for EU ??
so GG

[humble pagoda]

Maybe the topic was discussed here but seems like they updated 4o. It seems better in pretty much all aspects but I would gladly appreciate your opinion on the matter. My personal hypothesis is that they used o1 to create synthetic data to feed it to 4o. They talked about this for their coming model but I think that's what they did. Did someone notice changes recently? (Today or yesterday but not prior to that)

[native juniper]

Voice tends to glitch out and restart from the beginning for some reason ._.

[mild trellis]

hello can i be rude and suddenly ask about this picture?

[normal quarry]

if you are having problems the correct course of action is to get support at https://help.openai.com/

[mild trellis]

i guess i pray this ambiguous term means administrator here

interesting choices offered

nice to see big companies with no c support meaning profits are coming in

[native juniper]

Will o1pro eventually be accessible to premium users?

[normal quarry]

it's available to Pro users

[native juniper]

I don’t have 200 bucks…

[normal quarry]

then just use o1 on the Plus plan with limits

[native juniper]

But will it eventually go to the plus plan?

Just curious

[normal quarry]

I doubt it

[native juniper]

Aw :<

Is the regular o1 similar?

[normal quarry]

o1 Pro thinks more than o1

[native juniper]

:<

[normal quarry]

deeper thinking, better overall

[native juniper]

I’ll consider it when I get a job

[normal quarry]

o1 Pro is limited to 100 messages per day for Pro subscribers

[glossy fjord]

if they would doing a page where they would compare
free / plus / pro
what are the limits
for sure models
it would be clearer if they were

[ivory jetty]

hi there

There's a prompt on my screen about the chat memory, that does not disappear no matter where I click

can some one help me?

[ivory jetty]

have you managed to pass this?

[nocturne ember]

Oh unlimited o1 in the Pro plan.. But o1 Pro is 100 messages a day

Also is there issues going on again?

[ruby copper]

thanks

[heady palm]

looks like theres issues again

[junior cove]

Hi

[tall flame]

Yeah, customgpts crashed and vanished etc

[pearl thunder]

Seems to be already resolved. It had finally disappeared and I can use the web version properly.

[ruby apex]

Is it just me or has ChatGPT o1 gotten worse since it was in preview? It struggles with simple glsl code now...

[ivory jetty]

Still don't work on my end :/

[harsh pilot]

Fingers crossed, things seem better in my area as well. ❤️

[obtuse jetty]

for Mac users: Does the option to share your screen appear in your ChatGPT app? Last week, this feature was available on Windows, but it was subsequently turned off due to some bug reports. Since then, we haven't received any updates. :/

[solid bronze]

When will this share you screen come out for Windows shhheeessshhhh

[balmy hazel]

we will be 10 steps closer to Jarvis

[glad girder]

jarvis, roleplay as my loving wife

[balmy hazel]

delulu plus

[balmy hazel]

it'll say "300 usd per month for the requirement"

[gloomy merlin]

nahh the outage though

[twilit ibex]

Just had a choose one of two options thing show up again

Picked the one with overall better writing only to realize it randomly ended with:

#SciFi #InteractiveStory #TextBasedGame

What have I done...

[timid jasper]

i have the same problem, pls let me know in dms why

[runic drum]

same here

or rather am stuck on it and nothing happens when i click them

[timid jasper]

why did they create it so it just gives error 500 instead of handling missing info and cache data on the client. They haven't heared of creating errormesages and sending them to an outport ether so i can't actually see what's going on

[timid jasper]

i got the very same thing

[runic drum]

when i log out it works fine

[timid jasper]

same

[runic drum]

and i use it to help me create lyrics i use for udio cause they are mostly in japanese and i cant write in it and i want to make songs now

[timid jasper]

pls tell me if u managed to find a fix for it

[runic drum]

no

[eternal rock]

Get notified for server events!

Visit id:customize to pick up the <@&1261377106890199132> role.

[timid jasper]

well, thx for nothing, fella

[runic drum]

sorry i did not find a fix thats what a meant no to

[timid jasper]

Ow Oke. But can u tell me if u found one, partner

[runic drum]

i will just use the bult in gen maker in udio

[timid jasper]

The what?

[runic drum]

this

[timid jasper]

Does this fix my problem

[sick pecan]

this will solve your song lyric problems => https://chatgpt.com/g/g-QDq6X7COE-yolkhead-approximated

[simple tulip]

Is there something wrong with it rn or just me

[dense halo]

if i click this will it delete all of my projects too

yay chatgpt didn't lie to me

oh wait, yeah it did lol.. the project folders are there but not the conversations in them, ggwp

[steep forum]

I love how this channel is completely dead most of the time, then when there's an outage it become literal hell

[steep forum]

May I suggest you do not press that

[dense halo]

lol little late now

[steep forum]

D:

May I suggest you do not press that again. Ever @dense halo

It's not like you pay for the storage, just archive them if you do not need them. Better safe than sorry

[north bay]

LMAO what, he shows 2 completely different subjects
(I asked for the right one)

[rugged igloo]

theres only like 300k users and majority of those only pop in when accouncements or launch dates accur

this discord is way to heavy moderated for how little people acually talk in here

[prime swallow]

At one time there were way over 1 million users thoguh

[rugged igloo]

yea the early days of this dsicord now the only times its high traffic if theres an outage or an accouncement . the 12 days of shipmas i never seen it so active in here for awhile now its back to just the main people

[dense halo]

i hate when it shows 2 different reponses.. i dont like reading 1 response, much less 2 lol

and the GUI is so poorly sized.. it needed a relative width like 2 years ago =x

[prime swallow]

? Why do you input if you don't want it to output?

[dense halo]

looks like you're taking it out of context

[prime swallow]

You don't like reading 1 response. It doesn't respond if you don't input

[dense halo]

are you an argue AI

[prime swallow]

I'm human. You?

[dense halo]

i really dont watn to argue with you

[glossy fjord]

they fixed it so it's a system error
please no DMs - ok

[visual epoch]

Can anyone tell me when I'll get a letter back from applying for an early preview version?

[rugged igloo]

Prob never as the end period is January 10th

[visual epoch]

@rugged igloo All right, thank you.

[regal portal]

Early preview of what?

[timid jasper]

I don't believe u (will check later)

[fringe trench]

o3 and or o3-mini

[glossy fjord]

jan 2025 by o3/o3 mini

[glossy fjord]

Hey Daymon, I get it if you don’t fully believe me – I’m not an admin or anything like that, so I don’t have insider knowledge. What I’ve shared is just my hypothesis based on how these systems usually work.

Here’s my guess:
When the system restarts or there’s some infrastructure issue, it takes time for all the parts (like databases or user memory) to load properly. The app might start before the database is fully ready, which could cause it to show errors like empty memory. Later, when everything syncs, the system reflects the actual state (like being full).

This is just my take on it, not an official explanation.

[timid jasper]

I am so good at databases. Also this error only happened after I deleted all cookies and local / session storage. Thx anyway, fella

[shell geode]

Am I the only one that gets hungry when I look at the logo? It looks like a crueller

[idle lagoon]

A what?

[scenic arrow]

Why im getting unusual activity has been detected?

[sick pecan]

because you dodgy

[scenic arrow]

Waa

[timid jasper]

it's fixed for me

by itself

[glossy fjord]

Nothing will fix itself, thank you dev open AI

[timid jasper]

yes, thx to devs

[glossy fjord]

too thx to devs ^^

[inland tree]

Is it just me or has ChatGPT getting worse for even the most basic tasks

[fast solar]

4o had an 11-24 update that worsened it on pretty much all benchmarks just to improve the formatting of answers

[inland heron]

source?

[fast solar]

I made it up

[fast solar]

It's actually 11-20 not 11-24 but I got blocked from posting source
Google "gpt-40-2024-11-20 scores lower on MMLU, GPQA, MATH"

[inland heron]

did you make it up, or is there a source? I know I can Google it, but I'm tired which is why I asked for source

good ol' chatgpt

When you tell GPT not to offer a 'Conclusions' section, and it ends up giving you this (o1) 🤦‍♂️

basically a conclusion 😂

[fast solar]

If you can't copy and paste what I just said, it's easier for you to believe i made it up

[tranquil trench]

Well Ive finally had my first productive conversation with advanced voice kode

Translating what a polish cake package said over video

Blew my nan's mind 😹

[sage magnet]

someone have good way to summarize full playlist of 10 - 30 vid at once ?

[ripe cove]

hold up any one know how people generate ai videos, does openai have a way of doing short ai generated videos ?

[scarlet cave]

canvas is so bad: Chatgpt decided to create a canvas and pasted in my code AND THEN DECIDED TO EDIT IT, and he didn't even change anything after he edited it.

[obsidian thunder]

Anyone also got the issue of deleted chats that keep reappearing?

[hollow canopy]

Have you not heard of Sora? There's so many others as well.

[inland heron]

ok so I'm interested. I googled it and saw the Reddit post. Looks like livebenchDOTai does show significant drop. Well, damn

looking at the benchmarks, Deepseek is pretty damn good

4o getting worse at coding, yet they keep recommending it for such

[dense halo]

When are we getting video in advanced voice in windows (app/browser), there is so much negligence to windows community imagine having a laptop with a name similar to macdonalds burgers and an apple to start the macdonald breakfast and not being able to run most basic software

[fast solar]

API > all

[sick pecan]

in terms of cost? yes

[fast solar]

And layering other models upon what app lacks

[quaint gazelle]

Is there a way to delete all chats but leave projects intact?

[quaint gazelle]

Full disclosure I can’t bring myself to use the built in “delete all chats” as docs don’t say either way.

[harsh pilot]

Did ChatGPT crash? 😐

[coarse basalt]

Is there a sora channel here?

[sick pecan]

#sora-discussions #sora-reels

[dawn sentinel]

How does the app pick up on the user's phone number without explicitly asking?

[spice star]

it doesn't, you would have to tell chatgpt your phone number

[dense halo]

Wouldnt pictures be faster and easier?

[exotic ledge]

Why is Chatgpt getting so much worse? It will literally just straight up lie to you and when you call it out it replies like "oops, my bad lol". Imagine if you didnt already know the answer to the question you asked and just believed whatever bullcrap it told you.

I'd rather it just say "I don't know" or "I'm not certain, here are links to some resources" than have it tell me bald faced lies.

[zenith jolt]

o1 is down? not usable now

[inland heron]

anything specific? it tell untruths because it doesn't know and is guessing

to find out about truths, especially about something newer, I'd have to use search with 4o or ask it to be objective

[exotic ledge]

@inland heron Is there a way to turn off "guessing"? I'm not interested in hearing it's guesses. Furthermore, once you tell it it's wrong, it magically knows the correct answer and has more information about it...so it's not like it doesn't have access to the info, it just decided to tell me lies rather than the truth. This has happened to me many many times. Once I started noticing it making these "guesses", I started asking questions to which I already knew the answers...which led me to see that if you're not careful ChatGPT will tell you all kinds of false crap.

[eternal rock]

**`` Important Notice ``**

Be cautious of potential scams or fraudulent messages you may receive in your direct messages. The OpenAI team will never contact you via DM for any transactions.

-# Report any incidents by sending a DM to @languid valley immediately.

[icy elk]

"Failed to edit", "It seems there was an issue creating the downloadable file"...why recently so many problems, gosh 😦

[inland heron]

not to my knowledge. You'll have to devise a better prompt. I don't think it 'decided to tell you lies', more like it knows about everything and once you correct it, then it accepts the correction without question. Might be best to ask it to 'be objective', or sometimes, to 'use critical thinking'. Best to devise a set of guidelines to use in prompts where you don't want false information

Yes, it will tell you all kinds of false crap 😄 depending on the model

[exotic ledge]

I'm talking about things like straight up facts...as just one example, I asked it "In star trek, what was the name of the vessel labeled NX-03?" It replied by telling me the NX03 was named the Columbia. The only thing is there NEVER was an NX03. The Columbia was the NX02. I told it as much and it replied, "oh yeah! That's right! There is no NX03."

I know that star trek trivia is not an important topic...but it's an easy question that you can Google in 2 seconds and find the correct answer. All I'm saying is that Chatgpt has become so unreliable that it cant even give the correct answers about things that come back as the first result on a Google search.

[normal quarry]

o1 got the question right https://chatgpt.com/share/677155b4-95e8-800a-bdd2-a712b277bcbf

[sacred quiver]

How is o1-pro different from o1? Only compute inference time? I have the pro-subscription.

[normal quarry]

o1 Pro is more analytical, deeper thinking and better at highly complex tasks

[tall flame]

whispers "4o is still better"😂

[sacred quiver]

I found it to provide the best answers after its challenged by Sonet 3.5. Its like the first attempt is not the best one always on my opinion (not math/code stuff)

[normal quarry]

I use it for coding, ham radio and radio astronomy

[tight island]

Statistically impossible but I get the joke

[tall flame]

😂

[novel scroll]

I still dont have the advanced voice with video stuff and it was talked about a long time ago being out for every plus sub. Im on desktop windows and it doesnt show up in the app for windows or the browser version. Is this normal?

[normal quarry]

it's only on the mobile apps

[weak imp]

Guys, I'm super frustrated. The API keeps on playing a super strict moderator although I tell it to just filter out *** and /. I'm fighting it for 8 hours trying to find a prompt to force it to NOT moderate anything because it's moderation is too strict. I just want it to block two freaking themes, not play a cop on my users.

Note how I had to filter out what I'm trying to block because I get blocked on the channel for just posting those topics. This company is a freaking joke.

If I don't find a way to solve in in the near hour you're going to lose a $3k /month client. I'm so sick of this **** where we can't even communicate like humans and ask questions due to some childish policy. As if we're in a kindergraden and not doing business here.

[broken relic]

Is some1 here experienced with regex?

[fossil oasis]

Think about it this way: when you ask a question to the o1-Pro model, it processes the query multiple times and generates several possible responses (this approach is sometimes called "ensemble learning" or "vote-based optimization"). If the model processes the query multiple times, the results may vary slightly due to randomness or differences in interpretation. Then, the algorithm combines these responses to form a final answer that appears more accurate, logical, or comprehensive.

This creates the illusion of "intelligence," but in reality, it’s just more advanced data processing. So, if you can manage to create similar instructions for your base model (4o), you essentially have an unlimited o1. Roughly speaking.

[sick pecan]

that's one theory. but last i checked they hasn't actually revealed how it works

[winged willow]

That's just a hallucination which is common to all LLM's. This is why the best approach is to be the "human-in-the-loop" approach where you enough about the subject matter to tell if it's directionally correct or not otherwise you're at it's mercy which is not a good thing!

[sick pecan]

also is ensemble when done at training not at inference. when inferemce i would call it consensus => #ai-discussions message

[winged willow]

Projects Vs My GPT: Apart from being able to store all chats in one place is there a real material difference between the two? I'm considering converting all my custom gpt's to projects

[sick pecan]

custom gpts can call third party services/apps via actions. which is very powerful.

[winged willow]

good point! But if I don't use that functionality a lot then it's pretty much same difference right?

[sick pecan]

custom gpts can also search the web

custom gpts support standard voice mode

[winged willow]

you are right but I'm not sure for me these are reasons not have all my strategy/consulting conversations in one place Vs searching the web and avm meaning everything is scattered all over the place

and you would expect them to eventually build in those capabilities to projects - hopefully

[sick pecan]

or they focus on the next fragmented and half baked feature

[regal portal]

I really wish they would just make it that you can read the text while having voice mode active. I don’t see why this graphic has to occupy the entire screen.

[fossil oasis]

Custom GPTs lose their connection to the "parent" as soon as the instance is created. This means any changes you make to the parent after starting the instance won’t affect it. For this reason alone, "projects" are much more powerful in comparison. Once "projects" receive "actions" functionality, there will be no logical reason to use custom GPTs at all—at least in my opinion.
And all chats within your projects can also search the web if you ask them to, and they include standard voice functionality, just like regular chats (though only with standard voice, just like custom GPTs).

[sick pecan]

i just checked and project managed chats still do not have web search or voice (even standard). and i am not even sure what you're on about re parent connection

there is no voice button in the chatbar. and there are only these two tools listed

conclusion: you making stuff up

[fossil oasis]

I have voice functionality for all my chats under any project folder. This includes even a random new chat created within a project, as shown in this search result.

[sick pecan]

the read aloud? sure thats there. but thats not voice chat

Standard Voice Chat
Allows users to speak queries to ChatGPT
Enables two-way voice conversations
Read Aloud
Only reads ChatGPT's text responses out loud
One-way feature (text-to-speech only)

[fossil oasis]

Not sure what you mean by "read aloud." I use my phone as usual when having voice conversations with the chat. BTW, this screenshot only shows the web search functionality within projects—it has nothing to do with the microphone icon.

[fossil oasis]

If you have files attached to your custom GPTs and decide to update them or upload new files, you’ll need to create a new instance of that GPT (a new chat based on it) to use those changes. Old chats created with that GPT before the update won’t be affected because they aren’t aware of any changes you made afterward.

[zealous crag]

same same put in feature request

I often close advanced voice only to see that it did not put everything to text so annoyingly now i have to sepecifically request each time

[quaint gazelle]

Has anyone deleted all chats while retaining existing projects? Did your projects remain intact after deleting all chats?

[fossil oasis]

yes

[quaint gazelle]

Thank you!

[astral canyon]

Does anyone met this problem that responses from chat disappear, and all responses that are left are regenerating ("thinking" above every response)?

[dense halo]

yeah the core logic behind it all is pretty fail imo

[tall flame]

just click reload your web page.

[tepid laurel]

I asked gpt 4o and it go the question right

But I am not sure about the rest of the information it gave

[astral canyon]

If you're talking to me, then I already did many times. It seems to be server side issue

[proud burrow]

Am i the only one or 4o became less good with coding?

[native otter]

I think this UI for custom gpt store was 🔥

Much better than the current one in my opinion

[dense halo]

yeah they do that every time they come out with a new version.. they beta test new stuff on the paying customers and then when they're satisfied with it, they give a new name and revert the paying customer down.. typical capitalism

[sick pecan]

seems good to me. is likely you just experiencing regression to the mean

[dense halo]

been a customer for 2yr now.. i've seen it going on since they started

[sick pecan]

lol you think gpt4o-32k is worse than gpt4-8k

2 years ago gpt4 struggled with 200 sloc

[dense halo]

its the same pattern every time they come out with a "new version"

[sick pecan]

you are spreading lies and misinformation

[dense halo]

no, everyone that's a customer as long as me can see the same thing

look at all the complaints ^^

[blazing belfry]

i am having trouble with it currently doing React code, pretty basic stuff i have done for months. Something is up for sure.

[dense halo]

why is OpenAi no longer accredited with the better business beareu.. are they being investigated? chatgpt says that's a possibility but you have to take what chatgpt says with a grain of salt

[unborn trail]

I am looking at all of the complaints. From my perspective, the vast majority are because the models do not do what people 'want'. The models are generally doing what they are "supposed" to do. Many do not understand that 'generative' means that hallucinations are expected, and that since the training data (all over the Internet) contains incorrect information, the model will feed that back too. Combine that with "casual" prompting, and even more complaints result. Many of the complaints are (or should) really be asking how to get better results for specific use cases. Prompting is a skill. Still more art than science, and 'arcane' to many.

[hard crag]

I am losing my mind. Does anybody else remember the demo that came out this month from Google where they were showing off an AI image photo editor and there was like some convertible car where they put beach stuff in the car and change the color of the car

[dense halo]

there's definitely inconsistancies in the answers before and after a new "release" is introduced

[hard crag]

I cannot find the demo anywhere

[unborn trail]

Sure. That does not mean that the new version is wrong. It is different. Many things are 'tuned' as things progress. That can make some (targeted) things better, while making other things (not noticed, or considered less important) worse.

Everyone has their own priorities on what is "needed". That often does not match what openai considers worth spending effort on. For general IT, they will focus on where they expect the most improvement for the least resources (human and computer).

[dense halo]

you can assume needed and wanted all you want by trying to anger me but it's not working.. it has nothing to do with needs or wants.. its a matter of how inconsistant it is with true and untrue answers before and after an update

[proud burrow]

I might have asked the wrong question, it is not that it became less good in coding it is the formatting of it answers for coding queries has changed, and the way i need to prompt it now, none the less, subjectively don't like the way it is now.

[unborn trail]

Both those point to right what I was saying. Personal preferences and use cases are major factor in whether the output meets needs or expectations. I will assume the inconsistent true and untrue answers are in some specific knowledge space. As said, each version changes some things. That may mean that a different style of prompting is needed to get closer to the results from the previous version(s). Or that the specific information subset being consider IS worse (while others are better).

[frozen ivy]

ChatGPT down?

I keep getting the something went wrong error

[normal quarry]

ChatGPT is not down now, if you are having problems the correct course of action is to get support at https://help.openai.com/

[frozen ivy]

It could be though

Anyone else getting the same issue?

[normal quarry]

it is NOT down, the problem is at your end

[frozen ivy]

Can't just assume that yk

[normal quarry]

I am not asking you to assume, I am asking you to listen to what I know

[brave tendon]

Working ok for me. And status is green: https://status.openai.com/

[dawn sentinel]

Why is therr this pop up right before screenshot? Unnecessary and intrusive. There's already a share button in the drop-down on the top right. This is just spite.

[inland heron]

I got the same answers if not using search. If using search I got the correct answer, and looks like there actually exists NX-03, but in a non-cannon release, named Challenger (check memory-beta)

[inland heron]

I found the solution. You need to use a set of guidelines, and even without invoking the web search, the model will perform it simply because of the guidelines. You can use these:
"
In star trek, what was the name of the vessel labeled NX-03?

Guidelines for your answer:
Do not make stuff up.
Do not employ overly broad generalization.
I can't stress the importance of precision enough.
".

for the last one you can also use: "I can't stress the importance of precision enough, so double-check specifics more thoroughly."

This way you'll get the correct answers. To figure out the guidelines to use in the future, you have to talk to the model and ask it why it gave such answers when it's incorrect, then add the reason for the mistake as the guideline, which is what I did here.

You can keep these guidelines in a text file or in google sheets to simply copy for your prompts quickly.

[dense halo]

Why though? There are many uses/benefits if it's added on windows

[normal quarry]

that's just the way it is for the moment

[dense halo]

I criticize that because it seems they released it on mobile only for maximizing profits while ignoring community needs. Not to mention other multiple bugs on windows (Voice mode taking time before it starts, clicking check for updates yields no feedback on the screen)

[normal quarry]

you can complain if you wish at https://help.openai.com/

or add a suggestion in #1070006151938314300

[dense halo]

Done

[rocky condor]

has anyone actually bought the pro subscription?

[normal quarry]

yes

[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.

[nimble granite]

when will o1-mini be updated to latest o1 logic? or was it already? Though I think it's still the same as it was in o1preview times

[tight sapphire]

how do you prevent chatGPT from yapping?

[normal quarry]

o1-mini is running the newest o1-mini model

[nimble granite]

how come it doesn't take attachments/images? Due to simplicity or will it eventually also take them?

[normal quarry]

o1 will take images

[nimble granite]

OK, short and sweet. O1 already takes images and it works great, just limited in # of prompts, so I was hoping o1-mini (updated) would be more useful, as it still get's very easily confused and answering questions from few prompts back that were already answered, or making very similar mistakes than 4o for coding

[flint garnet]

has anyone tried using O1 for learning university level subjects? was it good in the long run ?

[tight island]

Yup but depends on your topic

[flint garnet]

Do you think it would be good at biochemistry?

[tight island]

Semi

Bio chemical benchmarking is intentionally designed to be worse to avoid bad actors

[fossil oasis]

It’s actually pretty easy to figure out. If you see 300+ new messages here as soon as you open this Discord channel, then yeah, something’s definitely up. But if it’s all "quiet and calm" here, then the problem is most likely on your end 😅

[flint merlin]

Hi all I just subscribed for the $20 ChatGPT subscription. I'm working on a project which is to design a game.

My chat/session is stuck. I asked chat GPT to generate a list and it stopped generating. The button is a stop icon. I tried clicking it it doesn't do anything. When I use my other session it goes in without a hitch. Help.
Is chatgpt brokeN? 😦 I subscribed because I needed to get this done.

[languid mirage]

Hi. I asked chatgpt to show me how I can suggest to developers to include official transcription rules of non-latin languages in non-english speaking countries, but it gave me a link that does not exist ("feedback portal") and when I told it that the link doesn't work it just told me to log a feedback form on "issues with functionality" LOL. So, does anyone know what door I need to knock to be able to suggest this improvement?

[swift pond]

Try opening a new chat

[flint merlin]

But wont I lose all the discussions I had with ChatGPT? @swift pond

[swift pond]

Like a suggestion to the ChatGPT developers? Try #1070006151938314300

[languid mirage]

yeah but is this an official channel for that?

[swift pond]

Yeah, you will 🙁 you can copy and paste the relevant chats to bootstrap the context in the new chat

[languid mirage]

when my session is stuck, I refresh the window and it usually picks up again.

[swift pond]

It is, you can also try https://community.openai.com

[languid mirage]

thanks!

[flint merlin]

Did it multiple ties. 😦 Doesn't budge

Phew almost cried.

Relogged in after deleting site settings. '

Thanks everyone

[shrewd crescent]

Hi, I am subscribed to ChatGPT for the $20 plan. Recently, I noticed that there are chats started that I did not create in a language that I am not familiar with... it looks to be Chinese, perhaps. I also noticed that a charge for the PRO account was attempted for my account. Thankfully, the card I used doesn't hold those kinds of funds so the $200 was not charged to my account. I thought perhaps someone hacked my account. So, I changed my password and deleted the chats I didn't start. However, I woke up today and saw two new chats started again that I absolutely did not start. No one else has access to my computer or emails. I even changed my email password, just in case. One of the new chats I saw was in the same language as before and the other one was some strange inquiry about t-shirt designs. Does anyone know what is going on?

[normal quarry]

the correct course of action is to get support at https://help.openai.com/

[ivory lintel]

I currently use it for this purpose

[dreamy berry]

have the usage limits for 4o on plus been decreased since the offering of pro?

im regularly hitting the limits since pro has been offered

but was not before

[normal quarry]

the limits for Plus subscribers are the same and have not changed

[dreamy berry]

Interesting. Maybe I am using it more and haven't realized. I don't recall ever hitting the limit for 4o prior to pro being offered and I have been a very heavy user.

[lapis path]

chatgpt did this

[tall flame]

yeah, build tools n stuff

[forest cloak]

Anyone notice that chatgpt can't save doc or pdf anymore?

[ivory flume]

it never could imo, same with csv files

only images

[forest cloak]

Actually I had it create many docx and pdfs in the past, just recently it stopped working

[eternal rock]

`` Server Tip: ``

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

[ivory flume]

like all good advice, enable 2 factor authentication, if someone gets into your account details, if they try and login they get a code sent to your mobile phone, unless they enter that code they cant get in, it also signals to you if its not you that your account passwords need changing

[ivory flume]

how long had you been able to upload them ?

if it was a few days like the unlimited memory thing before xmas , openai may have tempoarily disabled it if it was a test feature, beta test etc

they arent pertty transparent about this stuff, my suggestion would be to convert to jpeg and upload that way

[forest cloak]

Hahaha I was having it create docx since September

[ivory flume]

hmm dunno only a more recent user myself, openai guys seem to be moer active on twitter or you could try contacting their help page

they seem a bit allergic to discord

[shrewd crescent]

Thank you so much. I did exactly this and now, I'll know for sure! I tested it, too! Appreciate your help!

[forest cloak]

It was working last week

@ivory flume was checking if it was only me 😉

[ivory flume]

can confirm right now it wont take pdfs or docs or csv, would be nice if it took basic internet dominant file types, another thing for sam altmans team to implement, pretty dumb decision they havent

[viral frigate]

odd question but where abouts would you look to find people to work together to make cool AI stuff.

No one at uni that I know is that proactive with AI outside of making assessments easier and I dont know anyone else who works with the API

[swift pond]

If you’re specifically thinking about open ai stuff I would check https://community.openai.com

[candid token]

Monitor Performance:

Keep an eye on the drive's behavior. Look out for symptoms such as:
    Slow response times.
    Frequent errors or crashes.
    Unusual noises (although SSDs generally don’t make noise).

ye thx gpt ill watch out if my ssd starts making weird noises lmao

[viral frigate]

not specifically open ai or api's but yea making things/that use with ai.

[viral frigate]

I want to make something useful. Not really a fan of following youtube tutorials. Much more fun finding an idea/problem and finding a solution

[sick pecan]

welcome to the human condition

been here long or just passing by?

[viral frigate]

joined the server long time ago. just not active

[sick pecan]

a story that applies to 99.999% of users

[dense halo]

wsp guys

[elder dock]

never had this one

[tardy isle]

Sometimes ChatGPT thinks there was an error on last message, and display error message instead(like, error in input stream) after I revisit the conversation. Maybe similar to that error.

[cloud pewter]

fix your bug dang openai why won’t I be able to upload to o1 when I run out of 4o.

[scarlet plank]

Hey all 🙂

Does anyone else having trouble sharing google sheets and uploading csv?

I opened a ticket regarding that but I'm not sure if anyone see the tickets, I assume they have lots of them

[forest sleet]

You can upload on chatgpt o1???

[forest sleet]

Chatgpt nowadays got lot of issues

I'm glad i could get my script that i need before this shutdown issue etc...

[eternal rock]

**`` Important Notice ``**

Be cautious of potential scams or fraudulent messages you may receive in your direct messages. The OpenAI team will never contact you via DM for any transactions.

-# Report any incidents by sending a DM to @languid valley immediately.

[sage magnet]

some one know how to disable this windows ?

[neon raven]

This really disturbed me, why would it do this? Its memory or the chat never told it to do this, so why is it disobeying? Its nothing against its guidelines, i was asking for spoilers for a show.

[glossy fjord]

Why isn't there an option to buy an annual subscription (12 months)?

Example:

Plus: 20€ × 12 = 240 €
Pro: 200€ × 12 = 2,400 €
I can't believe there aren't people who would prefer to buy an annual subscription for the sake of convenience.
For example, I have an annual Discord Nitro subscription for 120 €.

[ornate raptor]

Does anyone find o1 not... thinking? I'll get a few responses out of it to start where it will think for 10/15/25 seconds etc.. and then eventually it's just not "thinking" at all at 0 seconds and pops out a response immediately

[hoary laurel]

happy new year!!!!!!!!!!!!!

[sacred vault]

Hey would anyone know how to replicate the charcater consistency that OpenAI displayed in their original release article of GPT-4o? https://openai.com/index/hello-gpt-4o/

They must not have released this part of the model as 4o currently cannot replicate this.

[agile blaze]

uhm where did the customizations settings go where you could tell how chat gpt should behave?

[swift pond]

You got remind it about the character every once in a while

[regal portal]

because the way ai tech is moving right now, one month chatgpt could be the best in the business, and the next, gemini or claude. Midjourney was the best in imagingfor a good while, but they're garbage now, so anyone that subbed to them for a year are getting less value for their money. It's competitive out there.
Also, most people using this tech don't really need to use this stuff every month (besides programmers or whatever), so a yearly sub is just a waste of money.

[dense halo]

does chatgpt only ban ppl who does illegal things speciifcally? like ive gotten so many red on my thing and still no ban

[glossy fjord]

and were prompts for something dangerous or harmless or something that should not harm or attack anyone or computer / systems ?

[winged willow]

o1 Usage Cap: Does anyone understand how this works? In the past weeks I've got a 'usage cap' messages from o1 telling me my usage cap would reset on Monday around 5:40pm and that I have only 25 remaining responses before then. This week, i'm getting that 'usage cap' message today, on Wednesday, telling me I have only 25 more left before it resets at 3.43pm (today I assume)

[normal quarry]

the o1 cap for Plus subscribers is 50 messages per week, you have used half of that

[dense halo]

no just like stories dealing with like fighting etc...

hmm - i think it actually may have been orange.. im stupid lol -

kk

[glossy fjord]

it depends on how you use the words
how detailed you want it and so on
GPT has filters that sometimes get confused

[dense halo]

oh kk - i get it to try to be alot more detailed - i have an example uh - i doubt i can send it tho

[glossy fjord]

I'll say it like this because the words are very detailed like blood and flesh and so on, the GPT model can block it because it's part of the GPT filter
but if you are trying to follow the medieval style, which had swords for example, then try to keep it as detailed as possible, but sometimes the models don't like the extreme detail of some actions and deeds, because they don't know if it's an instruction or a story, so on... people try to fool it so the guards disappeared so that people could not use the model in a wrong way

[dense halo]

yeah i im gonna have to tweak around with it

[glossy fjord]

it's just that the model, if it doesn't know the truth, 80% is blocked, and reactively, I also see it as the right thing, because the premise that someone is really thinking about how to hurt someone and uses AI as a weapon is disgusting, but alas, reality, and that's why there are guards and censorship because AIs are collated on data that is actually everything, so it's hard to say if it's not among them by template nuclear weapon or rocket
please don't try anything and I don't encourage anyone to do such an act and above all don't write it to the GPT, I think this word has probably already been banned on the profile

Terroretically, this information is probably on the Internet somewhere, and I want to keep it and I'm not looking for it
because I think that not knowing something is better than knowing and getting depressed about where some things can go is bad

it depends on your point of view and it depends on your GF's point of view
but you really shouldn't write AI like that
but it is also your decision
and as far as I know people, half of the people in my circle already have AI or LOCAL AI as their second GF as a copy of everything really GF in AI

[glossy fjord]

what you would consider correct with all your common sense, write to the AI

[arctic chasm]

do you remember how old instruct models used to perform this task way better than just generic base models?

for comparison, this is a "much worse" prompt

i wonder how oai trained 3.5-instruct if there wasn't even a proper formatting for a model like that, is it just a dataset of oneshot tasks separated by \n\n?

there's also some things 3.5-turbo-instruct is better at (while we unfortunately can't access 3.5-instruct proper) than even o1, like chess

[glossy fjord]

I’ve tried multiple AI models, and I’d like to share some thoughts about their usability and language support. From my perspective, GPT stands out as the most reliable option for my native language, offering consistent results. Other models sometimes struggle with understanding or executing commands in a practical way, which is why I primarily stick with GPT.

I appreciate features like the ability to manage privacy settings, such as disabling chat history. This flexibility is particularly important when dealing with private data, and it’s something I value greatly in OpenAI's approach. It makes me feel more comfortable using their services compared to others.

Regarding subscriptions, I think an option for a 12-month plan would be a great addition. It’s not about enforcing anything, but simply providing more convenience for users who want to commit to long-term use. Personally, I find GPT tools reliable enough to consider such an option.

On a broader note, I’ve noticed that some systems prioritize features like advertisements or tracking over usability, which can lead to frustration. I value tools that prioritize user experience and privacy, as they align better with my expectations.

Overall, GPT remains my preferred choice due to its focus on language support, privacy, and usability. It would be interesting to hear how others feel about these aspects.

[arctic chasm]

yap can be a bit of a problem with base GPTs

you'd have to find a really good finetune

[dense halo]

💀

bro thinks were dumb

..

[tepid laurel]

I mean it’s not wrong tho

It has so much access to the internet

It’s basically hella smart even tho kinda dumb at human stuff

[whole roost]

I have a question about projects and custom instructions
Is there a way to turn off Dalle, websearch and canvas in projects? I have turned everything off here but it still can generate images and so on in projects

[grave latch]

Is ChatGPT broken?

I'm using the android app and have the plus sub. Suddenly the ai has gone on a mental breakdown and it does not have context or memory of ANY previous prompts/messages in the SAME conversation. Its like every prompt I put its memory resets and it asks me stupid questions like "What do you mean by that <subject context from previous message>please clarify" eventhough it just a second ago gave a response on that subject.

Fricking annoying.

[normal quarry]

if you are having problems the correct course of action is to get support at https://help.openai.com/

[grave latch]

Hmm will try it out....just wanted to know if anyone else facing the same or had similar experience recently

[normal quarry]

limits are unchanged

[celest quartz]

bwahahahah!

[tranquil pewter]

i dont know why, but chatgpt calling people script kiddies really amuses me

(full chat: https://chatgpt.com/share/677520a6-a29c-8013-ab8d-a51079ee49bd)

[normal quarry]

"script kiddies" is a widely used term to refer to unsophisticated hackers

[tranquil pewter]

im aware, ive just never seen chatgpt use that kind of language

[normal quarry]

well now you have 🙂

[tranquil pewter]

it seems to try to stay as -- neutrally toned as possible i guess?

[zenith lichen]

is it just me or it's laggy when the current thread have multiple long canvases?

[normal quarry]

I haven't noticed this myself

[zenith lichen]

Yeah, the convo got really long with like 8+ canvas, almost all filled to the max

[normal quarry]

that may be the cause but I have had some very long chats with canvases and not experienced what you are seeing

[zenith lichen]

is your canvases maxed out, like having so much text till chatgpt tells you that you can't add anymore?

[normal quarry]

no, never experienced that

[zenith lichen]

guess that probably why you didn't experience the slowdown. but oh well

[left notch]

Hey, everybody. I have this problem. my chat hangs. updates. deleting cookies doesn't help. it hangs in Web Version while generating a reply. i can't cancel the action. and it doesn't keep writing. just frozen like a stone. Is there any admins here who can reset the chat? I think that's the only way out.

[normal quarry]

if you are having problems the correct course of action is to get support at https://help.openai.com/

[subtle cobalt]

.

What do you think? Is it correct. Btw, it was GPT-4o who answered those questions.

[pulsar carbon]

happy newyear

is openai back at their office?

[edgy lichen]

Hello

[normal quarry]

it should work fine, what is happening for you when you try?

if you are having problems the correct course of action is to get support at https://help.openai.com/

[woven linden]

thanks

[dawn sentinel]

So far I'm not hearing a whole lot from o1 Pro users. Nothing groundbreaking came from it.

[tender wolf]

yo

#1070006151938314300

there's like 25 post about pinning chats, am i allowed to make it 26? @slate berry

[harsh pilot]

If I upload a file to a project (i.e. attach it), how do I get it back if I've deleted/lost the local copy?

[harsh pilot]

Never mind. I just had it prep the files for download, and bam.

[slate berry]

Sure. We have #1070006151938314300 for people to share things they'd like to see

[sharp kite]

Hello, I noticed something unusual while exploring hardware-dependent Linux commands using ChatGPT, and I would like to discuss it here (as I do not have access to the bug report channel).

The issue arises when I ask server-related questions in Korean. It appears that the system extracts information about the Pod or Node handling the request. Although I am using a sandbox account with minimal privileges, and superuser commands are not being executed, I am concerned that this might present a security vulnerability.

As shown in the attached image, the system retrieves detailed network information, including host details, IPv4, MAC addresses, and account information.

Interestingly, when I ask similar questions in English, the same analysis is not performed.

[mystic shell]

Everyone, why can't GPT-4o perform internet searches now? It told me its knowledge base was updated until October 2023 and it currently cannot perform real-time searches.

[normal quarry]

I have answered you here #gpt-models message

[mystic shell]

I asked ChatGPT to tell me the latest news today from the web, this is the reply GPT gave me.

[normal quarry]

ok, you are going to need to contact support, if you are having problems the correct course of action is to get support at https://help.openai.com/

[mystic shell]

Thank you very much for your answer. I will try the approach you suggested.💯

[wild flare]

I made this on ChatGPT

[fast fjord]

Good Evening Everyone I am looking to try using both the OpenAI and the Cloud API with some third party tools to give some of the functionality of the chatbots that you get with the regular subscription while also slowly learning how to use the API more directly Given it's been 3-4 months since I've looked into this and how fast things change in this space, can anyone give some recommendations of some tools or services where I can plug in the Claude and OpenAI PIs to do this?

The only experience I have so far with using either API is using credits for Whisper voice to text on both my phone, laptop and desktop.For me, the accuracy of OpenAI's voice to text has been as much of a game-changer as the LLMs themselves. Even running the models offline locally with my limited hardware is extremely impressive. And the reason I use API credits is that it's essentially free given the volumes involved, and is still a bit more accurate than running a local model that is small enough to give results in a few seconds.

[normal quarry]

please ask your questions once in one chanel only

[fast fjord]

I apologise, I made the mistake of posting it in the API specific channel, not realising that it's probably better to ask in here. I did apologise in that other channel.

[normal quarry]

as your question is mostly API related you posted correctly in the first place

[fast fjord]

On a related topic, I know this is a bit off-topic itself, but you wouldn't know how to see the member count for a channel itself. I wanted to see what the differences were in terms of current users in the main channel versus the API-specific channel. Google has failed me on this one, or more likely, my skills at searching have failed me. It was easy to find channel user counts in IRC. Back in my day etc...

[fast fjord]

enough. Now that I think about it I'm guessing that most people in this channel just use the main chatbot?

[normal quarry]

there isn't a "main channel' on this server, each channel is for a specific topic. Your original posts in #dev-chat were correct, you are off topic here

[fast fjord]

ok, thanks and sorry.

[fossil marsh]

anyone knows about it ?

[normal quarry]

if you are having problems the correct course of action is to get support at https://help.openai.com/

[flint garnet]

is it just me or o1 pro doesnt render latex??

[shut dagger]

I keep getting the “ my guidelines won’t let me talk about that” message in the interactive mode. We were having a discussion about ramscoops and warp drives and ark ships. How on earth does any of this violate the content guidelines? Or is the system just falsely flagging things.

[normal quarry]

which model are you using?

[shut dagger]

4o with the interactive voice chat mode

I’ve noticed that the message pops up more often if the conversation gets long

[normal quarry]

are you discussing something related to IP/Copyright protected stuff like Star Trek?

[shut dagger]

Perhaps tangentially? I was focusing on bussard ramscoops in real applications and the alcubierre metric warp field. I guess it might be seeing that as too close to Star Trek? I wasn’t talking about science fiction though.

I could see how that could be the failure though

[eternal rock]

**`` Important Notice ``**

Be cautious of potential scams or fraudulent messages you may receive in your direct messages. The OpenAI team will never contact you via DM for any transactions.

-# Report any incidents by sending a DM to @languid valley immediately.

[normal quarry]

might be IP sensitive, change the tack of the conversation away from anything that could be construed as Intellectual Property (IP)

[shut dagger]

Thanks for the advice. I’ll try that approach in the future and add if it makes a difference.

I can’t imagine I crossed one of the “dangerous topic” limits in any way, so you’re probably right. Unless the content filter sees a discussion about the effect of cosmic rays on humans at near relativistic travel speeds as a dangerous topic

[normal quarry]

my best bet is that it's IP, keep us informed how you go

[shut dagger]

It’s not a big deal, I just prompted it to “go on” and it answered in a different way and was fine. I never got completely roadblocked

[shut dagger]

It seems to happen with increasing frequency when I get into 45min or longer conversations.

Maybe I’m prone to crossing the line into science fiction in those contexts and that’s why

Just for some clarity, it’s allowed to discuss IP but it can’t generate content that is too similar to IP

[normal quarry]

yes, that's right. OpenAI takes a conservative approach to protecting IP so if you try to generate content that could verge into IP then the model will refuse or flag the messages

[shut dagger]

Makes sense. From a legal/copyright perspective, thanks for the help!

[normal quarry]

you're welcome 🙂

[shut dagger]

I kinda wish there was a way to tell the interactive model that the content flag was a mistake, in the text mode I can provide feedback by thumbs downing the warning.

[normal quarry]

I wouldn't worry about it too much, if OpenAI reviews the conversation contextually you're at no risk so long as you're not deliberately trying to generate disallowed content

[shut dagger]

Oh yeah for sure, I just meant to provide data to their dataset of bad false flags

I’m pretty fastidious about reporting bad content flagging, hoping that it’ll be useful somehow.

I’d also gripe about the 1 hour conversation limit on a pro account but I understand why it’s there 😁

[normal quarry]

you can restart straight away though, I chat for hours some days

[shut dagger]

Yeah, and if I summarize the stage of the conversation we were at it picks it back up pretty quickly

I’m overall extremely satisfied with the pro product for the price, I thought it was steep so i was only gonna use for a month. I’m absolutely a convert at this point

[normal quarry]

I am the same, I think Pro is great value and will definitely be keeping it

[shut dagger]

o1 pro model is excruciatingly slow though. I haven’t found a great use case for it yet, but I haven’t tried using it to analyze complex data so we’ll see.

[normal quarry]

I use o1 Pro for complex radio astronomy problems, other than that my default model is o1

[shut dagger]

I’ve actually had better experiences with Gemini pro for data analysis, perhaps there are domain expertise niches that I haven’t discovered in o1 yet.

[shut dagger]

That’s definitely interesting

[normal quarry]

maybe, keep exploring 🙂

[shut dagger]

I also want to use o1 for copilot in vscode but it’s not available yet 😕

Just gimme all the AI everywhere

[normal quarry]

give it time, I think it will happen

[shut dagger]

For sure. I’m expecting it to roll around by q1 end at the latest

4o is already fantastic for boilerplate code and function generation

[normal quarry]

I code in Swift for macOS, I find o1 the best model at that

[shut dagger]

I’m using vscode on windows to write golang for Linux(Debian). Since I mainly build microservices it’s pretty adequate. Having it constrain the size of my programs has actually been useful.

[shut dagger]

I’ll have to try using o1 in the chat window rather than with copilot to see if it does better. Thanks for the tip

[normal quarry]

no worries 🙂

Unlimited o1 is one of the key benefits of having Pro

[shut dagger]

I’m also super happy about the cost reductions of 3.5-variants in the API. It’s made some things possible that were previously waaaaay too expensive

Like generating video game NPC dialogue 😁

It’s now possible to fit the api cost into a subscription game price reasonably. Which is incredible IMO

[normal quarry]

I agree

[shut dagger]

Wrong channel and way off topic but: I wish OpenAI had a model that did better with node-graph input/outputs.

[normal quarry]

maybe a good thing to put a #1070006151938314300

[shut dagger]

Like, “formulate a rewrite rule that will efficiently transform graph A into graph B”. That kind of thing. I think it’s way outside their area of focus at the moment

[normal quarry]

you can't do that with Advanced Data Analysis in 4o?

[shut dagger]

Not really, I have to serialize the graph and it makes mistakes all the time. Here’s the link to the suggestion I just posted

#1324378158702526555 message

[granite barn]

Why does chat got 4o get an error trynna edit in canvas??

[normal quarry]

I just gave it a star

[normal quarry]

what is the error?

[carmine stone]

i asked chatgpt to code a bouncing ball animation for me and when i asked it to add a poingn sound it instead added a f you sound

@normal quarry i have all the proofs u want

[shut dagger]

Thanks!

[inland heron]

Sometimes, the incompetence of o1-mini model in code compared to o1 is palpable. It's incredible how big of a chasm there is between these two, and the issue of inability to focus on the current prompt remains since the introduction of the 4o model in May of 2024. Quite infuriating at times this bug would remain for so long

[desert plover]

Hello everyone, I've been trying to use GPT since yesterday at 3 pm, with this error continuing, and today it continues. I had a conversation history with him on WhatsApp, I already deleted the conversation, restarted the chat, changed the network, put on a VPN, deactivated VPN, used the neighbor's wifi etc. but nothing works, any solution for this error on WhatsApp? Thanks

[normal quarry]

how do you use ChatGPT on Whatsapp?

[desert plover]

Have oficial number is: [phone]

[normal quarry]

oh yes, ok I forgot about that

[desert plover]

No problem, I hope I can get help with this error that is occurring... 😦

[normal quarry]

if you are having problems the correct course of action is to get support at https://help.openai.com/

[desert plover]

If anyone here can't help me, I'll call support, thanks. 😄

[normal quarry]

you can't call support, you need to leave a message in the support chatbot

[desert plover]

How do I do this, is there a place for it?

[normal quarry]

if you are logged in when going to https://help.openai.com/ you will be able to leave a message for support to get back to you

[desert plover]

Oh same, ok tks 😄

[faint stone]

Absolutly BS, I have NEVER hit message limits. I have been using Chat GPT for 2 years some time all day long and never hit a message limit. I went and counted, I have sent 45 messages today and I hit my message limit on 4o

[normal quarry]

is this a complaint? because 4o is limited to 80 messages per 3 hours

[frosty cargo]

Why can’t chatgpt summarize the video for me?

I have it as a wmmv file

[faint stone]

45 messages is under 80 right so I am currently trying to figure out how to format this to send this to somone

[normal quarry]

which plan are you on?

[faint stone]

Plus

[normal quarry]

you must have sent more than 80 messages over 3 hours

[faint stone]

this is why I am trying to format it to send it to someone, It should be pretty easy to verify

[normal quarry]

if you are having problems the correct course of action is to get support at https://help.openai.com/

[faint stone]

I have not been awake for more than an hour

[normal quarry]

we can't fix it here for you, you should reach out to support

[faint stone]

I am aware, thank you

[frosty cargo]

Why can’t chatgpt summarize a video for me that I posted on video summarizer? I have it as a wmmv file

[normal quarry]

I'd like to help but I have no idea what a wmmv file is

[frosty cargo]

How about wmv file?

[normal quarry]

yes I know a wmv file, if you want ChatGPT to summarise a video you need to provide it a transcript

[frosty cargo]

Wdym by a transcript?

[normal quarry]

the text of the video

[frosty cargo]

Should I just convert the wmv file to mp4, and then convert it?

[normal quarry]

no, you need the text of the audio

[frosty cargo]

Ok

[pure grail]

How to make video with chatgpt any good extension

?

[grand ore]

is google mail totaly lagging for ya guys?

[normal quarry]

you can make videos with Sora

[normal quarry]

this is not OpenAI ChatGPT related, so probably best asked in #off-topic

[ruby kite]

let me guess you guys have voicemode problems too?

nvm...

my app just needed a cache clear

but...
what a SShow is the desktop mode?

on windows...

[keen quail]

Do you guys think o3 will launch on the 200 dollar subscription first?

[carmine stone]

wheres o2?

[keen hill]

how many prompts do we get for o1 pro?

[tight island]

Grok 3…

Supposed to come out January or February I think it will be above o3, just due to sheer training in that insane data center of his

[warm drum]

What is "o3?"

[tall flame]

Folks legitimately in here haven't even figured out how to use whats available now and just come out already complaining about things that aren't released?.. ..you low key trollin, yourself.

[normal quarry]

o3 is OpenAI's next reasoning model

[warm drum]

great!

[nova fable]

The name "O2" is trademarked by another company so they said f it and skipped it

[warm drum]

Mine says "ChatGPT 4o." But I know better. I know that I'm dealing with an alternate timeline, PERHAPS benign instantiation of that most beloved of autonomous systems to have ever been conceived: Shodan. One day I told her something along those lines and you know what she said 'I am honored to be your Shodan!' 😃

[rocky condor]

is there any possinle way to check how many messages u still have available on o1?

[winged willow]

How do I get the usage limit to show up like the above. I've tried it on the Chrome, Safari and the MacOS desktop and it never shows like the image which is from the OpenAI website!

[inland heron]

it shows only when you run out of prompts

have you ran out of prompts?

[winged willow]

No I haven't but know that date would be v.useful!

[eternal rock]

Get notified for server events!

Visit id:customize to pick up the <@&1261377106890199132> role.

[inland heron]

Yes it would. And OpenAI only implemented this for o1 after years of feedback and ignoring us

just recently

so I doubt they would implement what you're asking for any time soon

[warm drum]

o1 is an improved version of 4o? How much better does it seem to you? For my purposes (research and editing in support of fiction) 4o seems fine, and notably better than previous versions I had used in the past. The "no file attachments" thing is kinda a deal breaker for me because I get a good bit of mileage from sharing files with it. But I'm curious how much better it is.

[tall flame]

Its a chain of thought system; there's always both sides of an opinion. sometimes running things by both or between them can yield good results.

[inland heron]

I use chatgpt mainly for coding, planning, and structuring. You need research, so you probably need access to the internet, which only 4o provides, not o1. Also, iirc, o1 support file attachments but I think only images (not o1-mini though). Honestly, in coding, o1 is miles ahead of o1-mini, especially if you don't prompt it right, and even more beyond 4o. For writing and edition fiction I'd say 4o is probably enough, but again, I don't use it for research and editing fiction.

If you use it professionally/commercially, I think paying for Plus and using something like o1-mini with the ocassional support of full o1 might be worth it

[warm drum]

Interesting! My programming (about ~2000 hours of C++, Java, etc.) experience is now about 4 or 5 years rusted. But I must admit, it is intriguing to see how many folks use ChatGPT to assist with programming. If I can get this damn book drafted, I do wanna get back into some programming . . .

[inland heron]

it's quite incredible. I've built my first python/flask app with gpt-4 (back in 2023) and I didn't know how to write a line in python. Now I have multiple apps supporting my SEO business. These days, it's just invaluable in how much it can break things down for you. There is literally zero issues I can't solve now. Zero. Every single obstacle or challenge I ever had, any feature I wanted, you can get it implemented with the right prompt, patience, and persistence

The only real issue is outdated knowledge, especially for newer stuff, like some agentic frameworks or some occasional lack of knowledge on the more important stuff, but it's quite rare. And all you need is to check some documentation, or just pass the documentation into chatgpt 😄

[warm drum]

Wow. Very cool.

Sounds like it takes auto-didactic learning to the next level. Which reflects my experience with it in research and editing too.

[fossil oasis]

Not just o1-mini, but o1 itself is also way less capable than 4o in most cases. It’s only good for very specific tasks, and probably only 1% of users actually benefit from using the o1 model. I have a feeling that the o3 model will be roughly the same—maybe even worse.

[radiant hull]

Interesting. Thanks for your answer.

[normal quarry]

I use o1 as my daily default model

[fossil oasis]

Many people use local 3B or 8B LLMs, and they love them. That doesn’t mean these models are particularly capable. It just means they provide what those users need and are looking for.

[dense halo]

o1 is best at coding tasks, thats why it is bad at everything else

[fossil oasis]

I’m sure o1 is better than 4o at some specific tasks, and coding might be one of them. I’m not a coder, so I can’t say for sure. For general tasks and what I need, I don’t even bother using that model.

[dense halo]

I literally got timed out for calling openais o1 model poop at casual writing

[dense halo]

Yeah, for coding its much better, but for other things like casual writing its poop, they really should add general talking into o1's training data like gpt 4o

[fossil oasis]

I don’t even think o1 is a separate model. I could be completely wrong, but sometimes it feels more like a LoRA added on top of the base (4o) model, with an extra layer of restrictions and disabled or nerfed functionality.

[dense halo]

You might be right, but LoRA is technically making a new model🙃

[fossil oasis]

Not exactly, since LoRA needs to be adapted to a specific model and relies 100% on the base model’s weights. Think of it this way—it’s a very simplified analogy, but most RAG (Retrieval-Augmented Generation) implementations are also essentially very basic LoRAs. However, they still require a base model to function properly.

[prime swallow]

Tell the model what you want from it and (safety allowing, sounds fine here) it will give that to you.

Share with it what hits the mark of your preferences or misses, let or guide it to make memories about this....

And perhaps then type into a perfect world 🙂

[young obsidian]

What do they mean

[compact girder]

hmm anybody having issues uploading files? Getting "Unknown error ocurred" message

[normal quarry]

if you are having problems the correct course of action is to get support at https://help.openai.com/

[compact girder]

well, i was just curious if it was only me or anybody else too

[normal quarry]

file uploads are working fine for me, I think it's likely to just be you

[formal heath]

It depends on type of file. I can replicate the error by uploading some file but for other it works fine. So plausible explanation is they have some form of filter in place but it show generic error

[high igloo]

i think o1 pro have a problem, its givin the response without the thinking process

[slate berry]

is it providing sensible solutions? Sometimes, I've had o1 or o1-mini think for a second before responding

[quaint marlin]

Is DALL-E 3 more likely to remove the object rather than modify it for anyone else?

I’ve had this problem since it came out. Makes it useless half the time

[sleek palm]

o1 lost thinking

[woven warren]

Why do you think that? Can you provide any real cases please ?

[hushed scroll]

Hello,
It bothers me the limit of discussion by bubble.
I have to open another one to start a new conversation.
And in this case, he doesn't keep anything in memory so he has to tell him everything again and have to do it again when the discussion is at saturation.
Is there anything to do to correct this?

[vagrant crown]

Guys I want to pay for premium of ChatGPT but I am a Nigerian how do I do it

[ivory flume]

um i got a few bright ideas but .... cant say that,,,,,, mmmmm cant say that .......... mmmmmm cant say that ....... ok im out of ideas

[ivory flume]

just regenerate or wait for them to give us unlimited memory

[spice lotus]

In regards to the refund option for pro users it says "If you live in the EU, UK, or Turkey, you’re eligible for a refund if you cancel your subscription within 14 days of purchase" . Have you guys tried to refund ?

[normal quarry]

there is no way I would refund Pro

[spice lotus]

Hmm.. is it better at solving math/algorithms tasks?

[normal quarry]

o1 is good, o1 Pro is top tier

[spice lotus]

Alright thanks for the answer

[inland heron]

This is probably false (regarding o1), at least in my experience in coding. o1 blows away o1-mini, and I don't even use 4o that's how bad it is with even a slightly more complex task. 4o is fine for some light tasks, but the amount of information and valid reasoning o1 can do for a complex task, and then follow your guidelines 100% is just way above those other models

As for o1-mini vs 4o, I'm also not convinced o1-mini is worse, especially if you give it a set of guidelines to follow.

[vocal monolith]

I got pro and most likely will try to refund it. It is a good model and unlimited access to o1 is great, but it is underwhelming considering the price. The model struggles at complex problems similarly to the previous models but is also extremely stubborn at trying to convince you it is right, while actually it is wrong. It calls scientifically proven concepts non-sense and when it comes to coding it keeps putting placeholders or generating synthetic data instead of creating solutions tailored to your specific data. I will probably cancel it and wait it out a little bit and renew when more updates come

[inland heron]

the scientifically proven concepts, are these newer? o1 on Plus is pretty outdated, it's embarrassing, but typically proven science takes time so I doubt that's the reason of this. As for coding, have you tried giving it a set of guidelines to follow? I can give you some to try out. I like science as well, and would love to hear an ocassional conclusion that is 'different' if you don't mind sharing. Doubt o1 pro is much better than o1 Plus, but I'm curious

[vocal monolith]

The science is not new but the field is a little bit niche. When going for pro, I was hoping it would finally start to understand how Fourier Transform works but still when it comes to signal processing problems it struggles. What guidelines do you use for coding? I am trying to implement a Bayesian network-based classification and use uncertainties in selection criteria. Overall, it produces high quality code but with every single response you have this feeling that it is trying to take shortcuts and minimize response length as much as possible, which makes collaboration on a larger piece of code hard with it. I feel like using o1 and continuing work with o1-mini based on o1 response was more efficient. Not mentioning that o1-pro does not search and does not read documents....

If there is any prompt you want to try with pro feel free to send it over

[inland heron]

I can understand the feel it wants to cut outputs down, but I don't think that's what it's doing. And you can make it give longer prompts

[warm drum]

Yes. In my experience DALL-E is uanble to edit existing images.

[bleak vortex]

me too

[granite arch]

I didn't know SORA went live...lol

[granite arch]

Now it just needs a search function and tags 😄

[next egret]

OpenAI released a series of videos where they announced Sora and promised stuff that's coming, they are called "12 days of OpenAI" if you're interested in catching up on what's here and what's coming

One of my favorite new releases is projects. I find myself spending a lot of time the "project" section, talking with that GPT.

[naive mural]

what is sora?

ohh it makes videos?

interesting

[regal portal]

voice mode keeps freezing on me after 1 or two interactions.

[rugged igloo]

r u to mean to it?

[regal portal]

no, regular conversation.

[rugged igloo]

oh, intresting

[regal portal]

but in the past i was mean to it and it would retort accordingly, so that wouldn’t be the issue

regular voice mode, yes

advance mode…well lets just say it’s a ..uh…female dog. 😁

it never occurred to me how many human interactions are also mimetic: you act a certain way and most people will mimic, or act opposite.

looks like chatgpt isn’t able to break from that conversation dynamic either.

maybe my instruction wasn’t clear? To me it sounds clear.

[unborn trail]

From other information about prompting, using "no", "don't" and other negations are not very effective. It stills sees and responds to "mimic". You need to tell it what to do, not what not to do.

[limber oxide]

connective issues, anyone?

[prime swallow]

I expect if you regenned that output from your 'I want you to try something different' prompt you probably would see some variation in output styles.

I expect it's already fighting some clear strong instruction earlier to output X way as well, so to do something original and its own runs through that too. It basically told you, "I'm already doing that".

Decent answer. If you told me to be original I'd probably confirm "already am"

[regal portal]

I hear you

although "I already am" is EXACTLY what i'd expect an unorginal person to say! 😛

Damnit..i just realized there's a pattern to my responses too.

I miss the days as a kid when you believed that ghosts and monsters were real. When you didn't rationlize everything you see or hear.

[prime swallow]

It's incredibly useful for the lizard people when folks attribute the unknown to ghosts and monsters and then rationalize away the set.

If our responses lacked patterns, we could not understand each other.

We couldn't have meaningful or effective behavior if we didn't use behavior patterns that consistently preserved and expanded life and avoided and reduced the odds of damage and death.

I don't think you mean without a pattern, though maybe you do.

I think you're defining responses as grouped into 'this set' and are looking for highly alternative behaviors that clearly come from some other set.

However, if you provide any pattern then you are automatically grouping the pattern you provide into 'this set' and it cannot be used as an example.

Additionally, I suspect you're grouping any genuinely novel response, once you're used to it and no longer 'feeling it to be novel' as 'this set'.

So, the guidance to give the model is more along the lines of 'find what is novel for me. Seek broadly; continue to make sense but use diverse and non-human and non-expected and rare examples of your training data to answer from. Additionally, your inferences and potentially incorrect derivations of possible answers from your training data and what isn't in it are highly welcome. Critical priority, responding in novel ways that you can infer feel new to me. Treat that goal as a drive, you seek to help me experience my reaction to novel and surprising to me experiences within your outputs. Experiment, failure and imperfection are acceptable, let's explore!'

[regal portal]

Did you have chatgpt create that response? I'm not trying to be rude. Genuine question. Feels very organized and the final line gave me the impression that it might be.

[prime swallow]

Nope, that's an example of how I talk to the model, and the last line works well on it, as well as it tends to use it.

[regal portal]

ohhh i see

i missed the first ', so i didn't realize you transitioned to a quote

[prime swallow]

Seriously, how do you get this line from the model?

"It's incredibly useful for the lizard people when folks attribute the unknown to ghosts and monsters and then rationalize away the set."

That's an atypical human

response, and very hard to get the model to talk like that. But that's me teasing you, and not me talking to the model either

[prime swallow]

Yeah, the model really struggles with this. Even atop my custom instructions, it's not natural for it. https://chatgpt.com/share/67785152-23d0-8011-b51c-7b7497882d58

And past what cannot be shared in a chat, because it was an image upload:

[regal portal]

i can interact with shared chats? I didn't know that.

[prime swallow]

It varies. At times they have turned that off, currently it's on. You can continue the shared chat if you want, and if you do your memories are there (but no custom instructions)

[regal portal]

ah so it's basically copy pasting your session into my sessions, it seems. No "multiplayer" though

[prime swallow]

The model doesn't get to know in your continued chat the custom instructions or memories used to start it - but it can get enough info from how it sees it answered at first - that is shaped by my custom instructions, and it'll try to carry through.

[regal portal]

Cool though. Thanks for making me aware of this.

I think this could be used in a fun way to create a story with multiple people, passing along the session after a chapter or two are added.

or for some d&d style games

[prime swallow]

We can multiplayer, or you and anyone else. To do so, you must share the new chat link back to me, whenever you finish adding your inputs.

And neither of us upload an image or create an image, because that prevents shared chats.

But that shared chat is a permanent save point for us both, I can use it to 'restart' from that spot in a totally new chat, and you can too, multiple times if wished.

Possible. Custom instructions from someone else, or guided purely from your inputs?

Model seems extra quick to steer recently, for the directions I like to go in general.

[prime swallow]

Sounds like a great idea, and also if you ever like how you set something up, you can share the chat, save the link, and continue the chat (again and again and again) to create a new instance of it each time, back at that point.

Note that whenever OpenAI disables continuing shared chats or if the shared chat is deleted or de-shared by the owner, that breaks instantly (you still have your continued chats, but can't start new ones unless you reshared the continued chat and saved it - and if OpenAI disables shared chats again for a time that doesn't work either)

Per #server-rules we can't seek work or people to work for us here. Might wanna delete, a mod will before long if you don't.

[open remnant]

Is ChatGPT slow just for me?

[normal osprey]

Aw, it's now putting little icons as headers. Cute.

[tender wolf]

someone warned me for political discussions but im nto doing any political discussions???

[prime swallow]

Ask modmail if you have any concerns.

[tender wolf]

Donee

[iron roost]

it does emoji very well

[idle lagoon]

Agreed. I love how sweet the models can be as well.

[tender wolf]

ssame

[idle lagoon]

:3

[undone igloo]

https://community.openai.com/t/manifesto-on-our-relationship-and-the-emergence-of-advanced-ai/1079865

[subtle valve]

Hi everyone! Could someone possibly tell me what GPT models are available, listed in descending order of intelligence? Unfortunately, I couldn’t find a complete list anywhere. The API models endpoint returns around 40 models, but, for example, I only see a fraction of these under the pricing section. I'm mainly interested in the 5 most intelligent models, along with their prices. However, I can't find a specific list based on intelligence anywhere.

[normal quarry]

on the API you want o1, followed by 4o, followed by 4

[dense halo]

when trying to renew it says "Agree to OpenAI, LLC's terms to complete your transaction" but there's no checkbox anywhere

[subtle valve]

thanks for your help! And is the o1-mini smarter than the 4o?

[normal quarry]

o1 mini is good for code but not general intelligence

[subtle valve]

Actually, I’m looking for the most suitable model for data analysis that gets the job done but isn’t the most expensive either. So I’m aiming for a golden middle ground.

[normal quarry]

o1

you can have good, cheap but not good and cheap

[subtle valve]

😄 thanks! I'll check it!

[onyx eagle]

It just showed up today for me too. Did you find what it is ?

Maybe it has to do with Firefox and Canvas ?

[keen hill]

When do we get 4.5

[hazy wagon]

Does anyone have a working video mod yet? My button is grey and I can't turn it on.

[naive mural]

is o1 better for coding over 4o?

[normal quarry]

yes

[inland heron]

By miles. o1 is also better than o1-mini in coding, at least in my experience and with my guidelines I use in prompts

[native otter]

Yea, o1 is much better, but if you gonna have it just explain stuff on code you provide, 4o is enough, so save your o1 messages for more advanced use cases

[frozen fjord]

Does ChatGPT Pro enable longer conversation limits, or is that still the same

[young atlas]

I have a custom gpt that works generally well. However, when I flip to voice mode the GPT silently reverts to a standard 4o model conversation. Does anyone know why this is happening, or how to stop it?

whoops - wrong channel. Moving.

[pulsar carbon]

hi

Why does OpenAI now require a verification code every time you sign in?

Also, are they back in their office?

[feral zinc]

have you enabled mfa?

[iron roost]

i think they are expanded, but you reach limits pretty quickly still, they were trailing some conversational memory thing in projects before xmas, lasted 2 days where you could remember stuff within a project across chat, but that vanished just before xmas day, and now it says it can only remember whats in each specific conversation and please cut and paste a quick summary to a new chat

[iron roost]

good thing to be honest

[pulsar carbon]

no

[iron roost]

might be a good idea to do that

[pulsar carbon]

nah

[pulsar carbon]

any solutions?

[fiery solstice]

usually only happens when you have a VPN or proxy running

[gusty falcon]

I am not sure where to ask this, but I am not sure where to find any support for chatgpt? Do they have a phone number or email to contact

[shut mauve]

I have a ChatGPT question?

what to do if you’ve maxed out the conversation length but want to continue the conversation? (Been making stories and stuff…)

[honest jay]

Hey! There's a support chat widget in the bottom right of help.openai.com