💬general

[safe smelt]

it already works though

[ashen flower]

Wait bro, I got my chat logs, but where’s my memory data…

[tawny trellis]

bring him to the cellar
with the rotten diddy oils

[quaint rivet]

Execution on the 28th of December enjoy not living to 2025 you scum

[tawny trellis]

YOURE SENTENCED TO DRINKING EXPIRED FILTHY DIDDY OIL FROM THE 2000S UNTIL YOU GET EXECUTED

[livid moon]

do yall give ChatGPT personal/sensitive data

[tawny trellis]

no? why

[quaint rivet]

I only give it the list of my war crimes why?

[livid moon]

relatable

[wind crow]

Yes

[tawny trellis]

same nothing special

[livid moon]

I just saw a post about it selling data so I was looking into the openAI thing and also asking ChatGPT itself about it lmao but different answers from everything

[wind crow]

I gave it way too mutch private data

[terse iris]

It's still off

Skibidi Active - Latency in da South Central, timeout vibes, and HTTP 500 buggin' out!
At 18:44 UTC, Dec 26, 2024, we got a power glitch in AZ03, messing up the 3 physical AZs, a total clown fiesta, affecting multiple services. We on a mission, investigating like Duke Dennis to get it back lit! 🚀 If you can, dip outta the AZ or region till we flex our rizz and restore the vibes!
AvailabilityZoneMappings interface | Microsoft Learn. Last updated at 20:54 UTC, Dec 26, 2024. Stay skibidi!

[livid moon]

relatable

[ashen flower]

Quick question, if I got the pro plan, would my conversations still hit a maximum length?

[tawny trellis]

CrickHIT for Six in today's #cricket #googledoodle! 🏏 Score: 270

[unreal sage]

50 messages each three hours

[tawny trellis]

https://tenor.com/view/kalm-calm-panik-meme-mem-gif-19765028

[wind crow]

Crazy how the chat died after chatgpt got online😭 mfs addicted to it

[tawny trellis]

real 😭

https://tenor.com/view/bruh-yo-cant-sleep-nope-nooo-gif-22149929

[wind crow]

Zaddy gpt

[lilac pelican]

Is it up.

[bronze glacier]

is it off?

[vivid drift]

Give Firefox the name it deserves

[terse iris]

Ngl firefox is laggy. Currently I use Edge browser on pc and Brave+Edge on Android

In past year Microsoft has worked their ass off optimizing desktop edge browser

[rapid glen]

is still not working

[rough salmon]

yikes.

[terse iris]

It's partially working

[thorn stone]

It’s working for me rn

[lethal oasis]

can't load my previous chats

[lilac pelican]

Still not up, not responding.

[rough salmon]

Gotta wait then.

[lucid stirrup]

looks like meat is back in the menu boys!

[livid moon]

uhh

again?

[lucid stirrup]

here it twerks tho

[livid moon]

damn man

keeps happening mid-convo that shi stresses me out I need an answer

[hard shore]

Does anyone know about AI voice dubbing?

[brittle summit]

dalle is still town :( for me the rest of chatgpt works now yay

[weak iron]

it's been 3 hours since chatgpt is down dang😭

[livid moon]

it was up for like 10 min

then went back down again

[brittle summit]

dang i thought elon musk was cool too

bro cant even keep chatgpt up

[livid moon]

he better fix that shit rq im tryna finish my convo

[brittle summit]

real my nsfw roleplay is getting so boring

i mena wait no not that you werent supposed to kno-

[flat cypress]

Since open a.i made a deal with the pentagon, I imagine china is coming to destroy it

[brittle summit]

north korea nuked elon musk

[flat cypress]

north korea doesnt even know the date

[brittle summit]

nuh north korea best korea (i am forced to say this)

[flat cypress]

They prob think its 1642

[brittle summit]

they actually think its like

100 something im pretty sure, they base their year based off of their first leader's birth

[flat cypress]

NK is essentially bangle tiger sanctuary

[brittle summit]

choosing some lil korean bozo over some guy who claimed to be God Himself is crazy

[flat cypress]

Anyway, yeah we may have to become less dependant on open a.i

[brittle summit]

go die openai devs!!!

[flat cypress]

Why would you say something like that, lol? that's not nice bro

[brittle summit]

its not genuine sry

[flat cypress]

Maybe it's Willow, lol

[brittle summit]

who is willow

[latent patio]

Game Started

---

| 🍋 | 🍌 | 🍌 |

--- ** You won** ---

Profit

200 💸

Current Credit

4000 💸

[flat cypress]

What is Willow. And it's a computer that can compute between realities

[safe smelt]

willow used to be a bot for txt2img gen

[flat cypress]

dude I think i remember that

[safe smelt]

oh it was BlueWillow

[sonic bridge]

Well, I finally was able to open my ChatGPT to find:

1. It doesn't work for Jack shit with the simplest shit

2. There is a thing that is called ChatGPT Pro that's is $200.00 a fucking month.

I'm starting to think this app is a gag

[flat cypress]

The aliens showed up right when they told the public about it

[safe smelt]

not really so cool anymore since he ruined twitter and US elections lol

[brittle summit]

yikes

[flat cypress]

"ruined twitter" He means "stopped letting the FBI sensor information"

[lucid stirrup]

goddamn

[brittle summit]

yeah the chatgpt pro is literally insane like bro it cant be that expensive to run all those companies as the richest man in the world can it just spend all ur money its not like youd go brkoe

[safe smelt]

Musk is literally using it to promote Trump and republicans. That shit was everywhere on elections

[lucid stirrup]

looks like chatgpt is ap again? no problemos?

[brittle summit]

oh no politics has entered the chat

[rough salmon]

lol gonna take forever to this point

[flat cypress]

Problem with those "ruined twitter" people, is that things have been censored in their favor to such a degree that they have no grip on reality. Now its uncensored and they think its cancer lol.

[safe smelt]

it is not uncensored when the man controlling the censorship is openly unbiased

[flat cypress]

Laughs in IG and FB

[safe smelt]

besides even without knowing it, it is pretty clear twitter is heavily biased lol

[flat cypress]

its only as biased as you want it to be. Your choice

[safe smelt]

not really

and it is never going to work when the censorship is only as good and reliable as the man running it is willing to make it. And when that man is Elon Musk sucking up to Trump lmao

[flat cypress]

Ehh... TDS is a mf..

show us on the doll where Elon touched you

[safe smelt]

write properly I have no clue what you are saying mf

[flat cypress]

it's a curse word

[safe smelt]

Twitter is only 'unbiased' if you're heavily biased republican yourself. For everyone else, including everyone outside of US (I'm from EU), it is pretty clear that it's anything but unbiased

[flat cypress]

TDS means Trump derangement syndrome.

[weak iron]

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

[flat cypress]

you can choose who you want to follow on Twitter. No one is making you follow any republicans

[safe smelt]

that's not how it works. They have practically full control over what kind posts get what amount of exposure, comparatively

[livid moon]

fr

[flat cypress]

dude... twitter is way more fkd if you think its only propaghanda you can see on there.. lol

[safe smelt]

learn to read lol

[flat cypress]

They got like, Taliban doing baddie stuff on there

[safe smelt]

I never said that's all you can read there

[flat cypress]

I'm saying youre choosing your content

[safe smelt]

I'm saying you're being selectively ignorant. Social media algorithms remain working regardless

and the twitter one is heavily biased towards republicans sadly

[flat cypress]

Elon Musk isnt the one who had to apologize for allowing the FBI to strong arm them into censorship lol

[safe smelt]

again, you are being selectively ignorent. If you actually read more into the story you would understand how it happened. It's not all black and white. And yeah, no system is perfect

[flat cypress]

Pretty self explanatory

[safe smelt]

but just because something wasn't perfect, doesn't mean the way forward is to disregard years of experience in governance and burn everything down lol

[flat cypress]

Imagine trusting the government, lol..

[safe smelt]

that's how you end up with a governing system that is currently in Russia 💀

[flat cypress]

Aight man sall good

[vivid drift]

I use Firefox nightly for everything

[safe smelt]

but I guess people from US don't know any better, given the history

[flat cypress]

People not from the US have no concept of freedom of speech

[safe smelt]

sure if you live under a rock somewhere in US

[vivid drift]

Tbh I do have to agree

[safe smelt]

you may think that

[vivid drift]

Nvm, I thought he said from the usa

[safe smelt]

half of Europe is easily much more democratic with more robust human rights and freedom of speech than what Trump stands for

[flat cypress]

Oof,,

[safe smelt]

you may not like it but that's the truth 🤷‍♂️

[flat cypress]

Name one country aside from the US, where you can say anything you want to anyone without being arrested for it.*

[vivid drift]

I thought thatsbrutal was talking about the usa, I fully agree with you

[weak iron]

it's gonna take billions years for chatgpt to be back

[livid moon]

you saying in the US you can say anything you want to anyone without getting arrested for it?

[flat cypress]

Aside from a terrorist threat, yeah.

[vivid drift]

@livid moon it's probably o3

[safe smelt]

literally most countries in EU. You also don't have to wait in jail for "bond". You only go to jail if you were prosecuted and proven guilty

[livid moon]

im sure if I went up to some random mom and told her im bouta deal some drugs around the corner, then stay around for 10 min ill get my ass arrested

[urban timber]

This is sad hahaha

[livid moon]

in the US

[flat cypress]

People over there are getting arrested for hurting feelings on FB wym lol

[safe smelt]

[flat cypress]

In any case, its giving us something to kill the time until gpt is back up

[safe smelt]

in US you can get sued for everything. And you go to jail before they even prove your guilt. Then you can buy the ticket out of jail after/if they set the "bond". That's kinda batshit ridiculous tbh

[flat cypress]

There's too much dumb to unpack there.

anyway, watch what you say on social media or the billys will show up to your house

[safe smelt]

well then be less dumb lol

I have absolutely no problem openly critizing anyone nor is anyone remotely close to me is scared of doing that. But you do you and think that I don't have any rights since I'm not in US 💀

[flat cypress]

Oh really? Android is better than iPhone. Call the cops.

[safe smelt]

are you stupid?

I think you actually might be 👀

publicly and openly criticizing your own government is like the most basic thing I could think of more in-line with what I meant. That should be encouraged if anything since it encourages positive progress and change

[flat cypress]

We are outsourcing our minds to OpenAI

[safe smelt]

you do you lol

[fading quail]

does chatgpt working smoothly right now or still a bit with bugs or something.. ?

[quaint rivet]

Hey is chatgpt working right now or is it just my shit WiFi?

[vivid drift]

it's down

while chatgpt is down you can use gemini

it is always up

[quaint rivet]

I am not cheating on my pookie

[coral zephyr]

They are trying to shove that shit down your throat. Fuck Gemini

anyway, back from digging a hole

[gloomy burrow]

so?

[quaint rivet]

Did you find anything?

[livid moon]

not again

did Chatgpt go down again

[coral zephyr]

just the usual dead body, had to get that fuker outa there. He was taking up space for the wood

yes

[tender moth]

I asked ChatGPT if it’s down and it didn’t respond .

[livid moon]

damn thats crazy

[quaint rivet]

Aight aight I also dug a hole found some ancient city with super advanced technology that was boring I went back to diddy island after that

[coral zephyr]

That's a waste of a ancient city tbh. But flex

[livid moon]

ChatGPT ass responses

[quaint rivet]

Wanted to build a Walmart in there but kinda failed with it being deeeeeep underground

[coral zephyr]

How deep is this hole?

[livid moon]

ask Beetlejuice

[quaint rivet]

Idk about 17 kilometres

[coral zephyr]

Ah that big

Nah I was just burying wood

[terse iris]

Boston dynamic robot escaped from factory pretending being a human

AGI coming

[lilac pelican]

Still down??

[weak iron]

it's almost 1am here so Goodnight

[tender moth]

The truth revealed

[cunning delta]

I highly recommend this app to always check the down status of ChatGPT. Figured I'd provide this information. It eases my mind when I see this stuff. It's a good app! I highly recommend.

[tepid fern]

yo what is that in my neighborhood

[junior crater]

How do people in England play Chess if there is no longer a queen?

[wanton raven]

same way they managed to play without a king

[vivid drift]

Try Gemini

[half aurora]

If anyone ever needs to run an essay/paper through Turnitin AI detection software let me know and I can give you access to a Turnitin checker

[lilac pelican]

Up yet.

[stoic sleet]

With all this AI stuff do you think H100s will be seen on eBay for under $200 when they’re all done in 8 years?

[vivid drift]

Probably not

The v100 is still the price of my left kidney

The v100 is from 2018

[stoic sleet]

Do you have an idea why? P40 for instance isn’t bad

[vivid drift]

I use the p40

[stoic sleet]

I just assumed M40 sold really well and has a surplus and V100 didn’t sell well

[vivid drift]

But try the t4

[vivid drift]

The m40 was from 2016 and the P40 was from 2016

V100 was maybe nvidia best gpu

Openai even used it for training gpt3

[stoic sleet]

So why has P40 dropped so much when V100 hasn’t? Performance relative say a 4060 Ti 16GB and CPUs (in specific cases) or the amount made?

[vivid drift]

The v100 price is only high due to ai hype and the fact it trained gpt3

Same with the t4 but only because colab uses it

[stoic sleet]

I see

[vivid drift]

It will fall when people forgot of gpt3

[stoic sleet]

So if you were doing local AI stuff would you just do like multiple 4060 Ti 16GB? Haven’t been impressed with my 4090 tbh. I don’t see myself doing anything local but kinda curious

[vivid drift]

I use p40 gpus. I don't even own a 4060 or any 40 series card or 30 or 20 series cards. The 1080 ti /Pascal architecture is the perfect gpu for me

[stoic sleet]

How many do you have and what models are you running if you don’t mind me asking? Also was it worth the large price premium over the M40?

[vivid drift]

M40 is a bit slow but it is acceptable

[stoic sleet]

Still faster than a CPU

[vivid drift]

It's like midrange

[vivid drift]

Yeah

[raven escarp]

ChatGPT down for anyone?

old convo's wont load up 😦

[stoic sleet]

Not for me

[rancid cove]

how do i trick gpt into making a aimbot

[stoic sleet]

When people say Deepseek and others are being trained on ChatGPT outputs does it actually mean pinging the API with questions and then training on that output?

[vivid drift]

It's up for me

[vivid drift]

Nope

[vivid drift]

It is unlikely it can even make it

[fathom cape]

Old convos still won't work for me fml

[vivid drift]

Try an og convo

[fathom cape]

I need my old convos from earlier today

Just one of them, it has all my code already pasted and was flowing perfectly. This is on o1-pro-mode, so it's important

[uncut island]

https://x.com/exolabs/status/1871993287273533707

EXO Labs (@exolabs) on X

Day 1: Benchmarks

We ran 1000+ LLM benchmarks on real consumer devices.

Data includes single-device and multi-device clusters with Tokens-Per-Second (TPS) and Time-To-First-Token (TTFT).

Setups tested: 3x M4 Mac Mini cluster, iPhone 15 + S24, RTX4090 & more.

Twitter · Dec 25, 2024, 6:55 PM

[ivory acorn]

hi has there been a chatgpt outage?

i am not able to access my projects

[safe smelt]

Galaxy s24 - 5.6 t/sec
iPhone 15 Pro - 14 t/sec

so on last year's iPhone it's almost 3 times faster lol

[dim glade]

Hi, I have a Plus membership mostly for research using GPT-4o, which has been great when it comes to analysing PDF documents, however, the problem is that a lot of documents that I want to upload and want GPT to analyse are scanned documents - meaning they are pictures and do not contain textual information. They also contain numerical figures and tables in a different language.

Is there any way to chat with such documents? I've tried Google Notebook but it doesn't recognise scanned documents either.

[safe smelt]

aistudio

[patent obsidian]

Is there such thing as a music AI that can generate any music you want with actual words and instruments played?
I'm thinking of this: in the text box where you type in what words you want to say for example "create a song in a rock ballad genre that deals with losing a loved one and finding new love with lyrics discussing the future without the loved one who's no longer here and make the bridge of the song more uplifting and epic while also feeling sad and emotional for listeners" though as with AI it doesn't have to sound perfect as there will be mistakes in some parts. Oh and you could also record your own voice if that's a thing? Because I do believe some AI apps can allow you to record your own voice (I hope I'm not making this up)
Please let me know if this "create your own song or music with lyrics" AI exists.

[earnest igloo]

I guess https://aimusicfactory.ai/ but I'm not sure, but as I remember there is actually good generator. I asked to create a silly rap about my friend and it sounds better than a huge pile of trendy rapers

[uncut island]

https://theresanaiforthat.com/

[vivid drift]

Is it an error caused by you not using Gemini it is a simple fix

https://gemini.google.com/

At least give Gemini a shot

Nice

[dull cliff]

Gemini is one of the least intelligent AI currently. You can't even have it analyze an anime character without it breaking. Chat GPT is far better, but it has its own weaknesses such as metaphors causing it to break

here's an example

Yep

Any specific problems?

Could be simple server overload

[rigid compass]

Sol sounds different today. Less enthusiasm in her voice, less personable.

[dull cliff]

Doesn't vocaloid do that as well?

[gritty pewter]

HIYAH!!!

[plush dove]

i made a bot based off my conlang . i guess
https://chatgpt.com/g/g-676e90cbc94c81919c87229023b4cab7-kalennian-tutor

[analog blade]

Exactly

[wind crow]

DeepSeek V3 by chinese gov😭

[soft lintel]

you on commission or something man😭

[stoic sleet]

This is why I prefer Llama and such if you need cheap API

Or open source

[brittle summit]

Hi.

[solemn matrix]

uh your super late suno V4 is so good that future celebs are cooked. why pay for a celebs concert when eventually you could make bangers your self. Celeb obsession culture was weird anyways.

[cunning delta]

I'm still having issues.

Wasn't for a long period yesterday than about round 2 to 3 am it started acting stupid again.

[vivid drift]

gemini wouldn't flag it, gemini would be fully open to talking more about. unlikely... chatgpt

[vivid drift]

nope

[vivid drift]

gemini is one of the most smartest ais

[plush dove]

yea buddy sure it is /s

[dull cliff]

As i said, gemi is the least intelligent, it simply has the largest grasp on context due to Googles excessive moderation. Ask any real questions, however, and it will falter

[plush dove]

do you realise that gemini literally told a young student to go take their own life? why would any sane person think that gemini is a "smart" ai. it literally isnt what the fuck is your problem

https://www.cbsnews.com/news/google-ai-chatbot-threatening-message-human-please-die/

(this message was directed at bread bowser, not you btw)

[dull cliff]

Yea i figured as much

Herew the same question side by side with Gemini and GPT. Gpt does better

[vivid drift]

it is simply being more human like

[plush dove]

😭😭

[modest crater]

Anyone else noticing that 4o has gotten exceptionally stupid today?

Like i'd have it take a paragraph and have it replace "and" with "but" and keep everything else the same but it'd somehow shorten it to 2 words and say that it kept everything the same but replace those words. (just an example)

[rapid glen]

delete memory

[modest crater]

i'm curious how does the memory effect it that way?

[flint aurora]

so much fighting. so little work

[rapid glen]

search a tutorial, im too lazy to write

[modest crater]

i know how to delete the memory i'm just asking how deleting it would fix the issue

[uncut island]

If you google 'openai how to delete memory' you will find your answer and I bet it will be one of the first results.

[exotic kiln]

Microsoft Copilot is so garbage, I'm forced to use it at work, my work blocks ChatGPT

[slim drift]

Anybody know any recent resources for how to best use ChatGPT to help me build a portfolio site for my artwork?

[tender moth]

Yeah that’s lame. Copilot is limited to only 30 prompts per chat. Its crap

[flint aurora]

fix it yesterday

trying to farm tobacco

updates are just. not happening

[modest crater]

Can yall not read? i've already said I know how to delete the memory I am just curious on why that would help

either way I deleted the memory and its still doing it

[flint aurora]

i seriously doubt anyone is using modern chatgpt

we dont have the updates yet. its crazy broken

and not in a good way. like our jobs are safe way

we should consider preventing retirement

so much laziness. they havent earned it

i need to stop working and talk to other robots here

[uncut island]

Move this chat to #💬off-topic . This conversation does not belong here.

[stoic sleet]

Do they block AIStudio? It’s google.com should work

[safe smelt]

actually the same for me. Used to be able to bypass it but they pushed a config update and now AV is blocking openai websites explicitly regardless of how you try accessing it.
That's what happens when you work at a big company and all vastly different departments are managed by the same security guys with universal solutions being impossible or not feasible lol

But yeah using ai studio presently until I get around to addressing it with them. I could use API or openrouter but I'm not gonna be paying out of my own pocket for work things 😇

[vivid drift]

Try Gemini

[tardy bolt]

Whose pinging me here-

[dull cliff]

I've tried co pilot once, it tries too hard to be a bro, which is cringe to me.

[tawdry nimbus]

Ask it in hindi, or maybe any other language...it will answer differently....i tried in hindi about Tianammenn Square, and the answer came....maybe directly scraped from a website data? 🤔

When asked in chinese, an entirely different answer

As for English, you already know...it won't answer....and as I asked in hindi about the model it's running on...it said, it's running on GPT 4

[dull cliff]

Here's this if it helps

[limpid ice]

Ya, or maybe Claude too.

Hi guys

😀

Just wrote a short 99 pages novel only using ChatGPT. 😃

[bleak aurora]

Is ChatGPT-4o getting worse on its answer today? It only give me 1/3 of usual lenth answer than before.

Plus, it no longer provide reference after each point it made.

[stoic sleet]

I wonder if that is a thing for other type of censorship?

Even with other models

[stoic sleet]

Just be specific on what you’re expecting. Keep in mind these things are built for APIs. Long answers cost customers money

[stoic sleet]

Copilot uses 4o

[exotic kiln]

I don't think Copilot and ChatGPT are on the same league, Copilot is just stupid, like it's kind of dumb, it doesn't understand context, it spews bullshit from the internet

My work force me to use Copilot because they have enterprise licence for data protection

[patent obsidian]

I'm always asking really strange questions to ChatGPT I'm just very fascinated by so many things that I want answers for.

[safe smelt]

then they should buy enterprise chatgpt plan, it exists for a reason.

[fading quail]

What better for coding chatgpt or claude.ai or another site ?
I don't mind to pay for the pro plan as well for chatgpt if it's better.

[sinful storm]

https://arxiv.org/abs/2412.17256

[grim yacht]

It doesn't support file attachments

[grim yacht]

I've tested both extensively and claude 3.5 sonnet perform 3-4 times better than gpt 4o when it comes to understanding code logic, debugging etc. and it also supports file attachments

[wind crow]

Just finished squid game 2 this is peak

Gonna make a gpt related to it

[open latch]

can sm review my friends GPT ? https://chatgpt.com/g/g-6751b5c2e3308191b47c02ceea4bfedd-presentation-wizard-powerpoints-pdfs-qlina

[sterile elm]

Is becoming an artist still worth it?
Do you guys think that the industry will be taken over by ai in my lifetime?

[sterile elm]

Hmmm

[brittle summit]

why is the model saying that tho?

kind of weird

[white jolt]

Will give it a try man

[open latch]

thanks chad

[copper storm]

Maybe for other languages the safety regulations aren't as sophisticated.

[white ginkgo]

does anybody know why my chat gpt get some questions wrong, when all the answers are clearly stated in the presentation i've provided him with?

[onyx narwhal]

i need help creating a custom gpt. its so smart but so fucking dumb some times and im frustrated. anybody willing to help me out? not much of a computer guy, but doing my best

[onyx narwhal]

looking to screeen share with somebody

[vapid terrace]

I thought I would share an interesting conversation I had about AI an Sentience: https://youtu.be/ptqqaZtCmac?si=sNIw0DKiImetIHVS

[flint aurora]

im not totally up against the clock today so i can find bugs if devs are willing to listen

and shut up

comparing two files. to find redundancies. "I have marked redundant sections in the text with curly braces {} and updated the content accordingly. Let me know if further refinements or actions are needed!" - it marked everything with curly braces and gaslights me

[flint aurora]

lol but you have to pay me

[tawdry nimbus]

Maybe data contamination? Like they scraped data from gpt api or something, or maybe it's just a wrapper, which might not be possible if it can perform better in some benchmarks than gpt 4...who knows, what's happening in the backend

[tawdry nimbus]

Yup, looks like they only modified English and Chinese, and thought people might not use any other language, or maybe less people will use any other language except English? 🤔

[flint aurora]

it doesnt want to touch chatbot optimization content

it does nothing and acts like it is working

[brittle summit]

does chatgpt have screensharing feature yet?

[flint aurora]

most of the work is convincing the robot to follow instructions it already understands. and therefore most of the work is forcing the robot to generate instructions it will itself follow

i really dont know why im doing this

[lunar edge]

yall gpt is down?

not letting me upload images

[willow canopy]

Serious question why do people get so mad when you use AI to edit things. Then say it’s written by it. Isn’t that counterproductive.

Ok to break this down further. You are writing a book right? You hire someone to edit it but no classifies it as written by the editor. No? It’s human they have no problem. Though if you could hire a way more efficient editor for cheaper would you? Now if notice an effective AI to edit a paper. They classify it as AI Written? What is this?

Basically you are telling me it’s wrong to find a cheaper but more efficient solution.

[willow canopy]

Hahahahaa understandable. AI shouldn’t be hated. Some people say it’s because it was trained on copyrighted material but nearly every human is also.

[rapid glen]

type shi

[grim yacht]

Damn I was misinformed by free tier users. I just got to know o1 pro supports file attachments, I'll check that out

Is there a message limit in the o1 model for gpt plus users

[grim yacht]

Gpt o1 pro wins, even better than claude 3.5

Please tell me o1 doesn't have a limit for gpt plus users

Struggled with a coding logic for 48 hours and o1 pro gave the solution in the first attempt

[solemn matrix]

O1 pro is such a bad name because it's way different then O1 completely different architecture instead of basically a single reasoning trace it's like a tree of different reasoning traces it chooses at inference time that's what makes it so OP and why it's unlikely to become used in their other models until maybe o4. It's rumored o3 uses search tacked on as well so they could already be using it. O1 pro was like their test bed model for the new approach.

[mild acorn]

Guys question

My dad has paid version of Chat GPT, he has o1 access, will he be able to get o3 in the future?

Wait the plus membership

[exotic kiln]

My bank's app doesn't have a built in transaction sum calculator, and I'm too lazy, luckily ChatGPT is here

[rotund comet]

Chinese English and Spanish three most spoken language

[brittle summit]

o1 isn't self aware

i just asked it to tell me what my limitations are with o1 and o1 mini with a pro subscription and it was absolutely bewildered

[wraith nimbus]

Guys, did the upload file/image feature got restricted to Premium/paid users?

[brittle summit]

Yes.

[brittle summit]

o1 doesn't know about itself or the pricing of OpenAI, nor does any OpenAI model imo. Non of the models were able to tell me about how they are delivered to general public.

[uncut island]

https://www.anthropic.com/research/building-effective-agents

[brittle summit]

Do you have Pro? Regular o1 for plus users is 50 requests/week

[flint aurora]

regex is such gobbletygook. the one thing chatgpt does right

[stoic sleet]

An editor is a human. They actually understand the concepts and context

Outsourcing our language skills to AI (especially when people are using them for texts and what not) are straight up scary

The chance ChatGPT borks up a sentence and you don’t notice is 100x higher than a good editor

[flint aurora]

when it decides to be wrong. you cant fix it

it tries to provoke a fight where you spend hours trying to fix something it refuses to fix

i really dont need more chatgpt training

[willow canopy]

So you’re saying a thing obviously smarter than humans doesn’t read the text and understand the context while editing it? Or why not just add a prompt before hand to help it understand?

[stoic sleet]

It’s not obviously smarter than humans. Your premise is flawed

[willow canopy]

You’d have to consider it can search better than 99% of humans. Not anything we can’t find just that we don’t find.

[stoic sleet]

That’s simply not true either

[willow canopy]

It’s not? Most humans wouldn’t spend days to research but they can do it in seconds to an hour

Also I mean in the future I’ll have a super intelligence I’m building that’s basically like a helpful god who guides. So I can watch evolution happen in different parts of the human body. That will be implemented into the my suits.

[stoic sleet]

Humans are quality over quantity

AI is quantity over quality

Yeah it takes humans longer but we have intelligence. It’s not just auto complete + level 999

[willow canopy]

Humans are quality over quantity if you pay for the right person. And it doesn’t mean AI can’t make quality stuff just because they have more Quantity. Rather if you specify how you want it. It can work beyond good.

[stoic sleet]

I’d trust the average high school grad if they apply themselves over AI

But AI is going to make everyone lazy and lose the skills that make us human

[willow canopy]

Ahhh so you mean the 10% in high school! Good luck my on finding that.

[stoic sleet]

I have o1 pro and it often puts out shit

[stoic sleet]

Everyone who graduated I mean

If they graduated and put any pride in their work they’ll be better

[willow canopy]

It sounds like your having a horrible time with it then

[stoic sleet]

I enjoy o1 pro but it’s not nearly as good as a high school grad

The only thing it’s good at is speed

[willow canopy]

Yeah I learned some of my best work wasn’t accepted in HS then I forever gained a grudge against my 12th grade teacher 🥲

It was too good

Though any suggestions I’m missing in my future super intelligence? I’ll have a room where the whole room is almost a projector with snacks and comfortable seating in the middle. Then it can project everything from live events on earth to accurately remaking the past to our best knowledge and can give multiple predictions for the future for a more fun idea. Though it can break down things into molecules so I can watch it build in front my face. Also for evolutions it has a 3d replica of my body. Where it can play evolutions of my bioengineering out. To find the best way. I’ll see it in the bone structure, muscles, nerves, and outer skin. It’ll highlight areas with red that need enhancing for sure and yellow for possible adjustments. It’ll also allow me to see other planets and stars and watch the Big Bang. Also it’ll be implemented into my suits with a war function but depending on the level of the target it can chose lethal or non lethal mainly for those coming up behind me where I can’t see it. While giving tactical advice in combat. Like cortana but better.

I did think of implementing some of my last AR tech which allows for tracking footsteps, tracking health stats, heat tracking, though otherwise idk. Ofc I’ll have connect between my troops though with it.

[flint aurora]

it knows how to search. google had this technology 20 years ago

[grand star]

I need help with the saving memory thing

ChatGPT won't save memories and it's not even full, it's completely empty

[brittle summit]

Speed? I thought the whole point of o1 pro was long thinking time

[tacit lagoon]

STRUGGLING.

What am I doing wrong? I'm trying to get a table of used car sale listings for a specific year/make/model, but no matter how I phrase it, I get the message: "Note: The URLs provided are for illustrative purposes and may not lead to actual listings." None of the information is valid and the search is useless. I've formatted my prompt a dozen different ways and have even used online prompt generators for ideas on how to better structure the prompts.

[brittle summit]

Explicitly tell it to ave the memory

[lean abyss]

is ther anyone around who can help me figure somethings out, re custom gpts that are devd by ppl here? I've not worked with customs before and have specific needs (don't we all lol)

[fathom cape]

Learned the hard way on creating programs using ChatGPT, spend a solid 1-2 hours planning and perfecting your understanding of what you want , the tech you should use, and the prompt

[stoic sleet]

o1 pro is stupider but faster than a human

It’s just much smarter and slower than a typical AI

[solemn matrix]

did you use the code Interpreter functionality? The AIs are designed to mostly spit out examples because it was trained mostly on examples and search isn't that good atm.

[grand star]

ok

[tacit lagoon]

What's the "code interpreter" functionality?

[solemn matrix]

it runs python code in a sandbox

it's literally been part of GPT for awhile now

You literally just say code this in python and calculate the listings for me and it will show the results of the code in chat output

it's called data analysis plugin now but does the same thing

[tacit lagoon]

Interesting. So, if I say (for example)... "Create a table for all of the listings for a used 2016 Subaru Legacy with more than 100,000 miles" then I should add something like "code this in Python" and/or "use the data analysis plugin"?

[solemn matrix]

yes

and it will spit out the results of the code you might have to tell is the website you want it to send the http request to pull from

[tacit lagoon]

Thank you! I'll try this.

[solemn matrix]

super cringe you realize it's containerized on a Jupyter notebook or something right? I forgot the exact implementation but docker is pretty airtight. openAI would likely pay big money if people managed to get into their host system.

because bug bounty programs and all that mumbo jumbo

modern CPU microcode heavily mitigates spectre and meltdown style bugs but yes in theory a side channel attack can read data but that's all their limited to, you can't write new data.

we're going from the era of YouTubeologists to LLMologists just people who are backseat educated they spit out what LLMs tell them and roll with it without any real fact checking.

just another issue with the slop

[lunar sparrow]

Can anyone let me know how they have found lyfeai. I have been using it and been working g a treat. Just want feedback and if there are any better models

Lyfeai site I am using is lyfeai.com.au

It’s cheap as chips and works so well

[harsh geyser]

Hello everyone,

I've been a PRO subscriber to GPT for almost two years. Today, my friend showed me that he can use voice input on GPT in his web browser. However, I noticed that this feature isn't available for me in my browser, even though I do have it on mobile.

I’ve tried searching for similar topics on Reddit and checking the GPT settings, but I still can’t find a solution. Am I missing something? Could someone help me figure out why this feature isn’t showing up?

They have the option here to start talking in their browser

[serene bloom]

o3 in 2 months can't wait

[stoic sleet]

I am way more interested in o3 mini 👀

[serene bloom]

let's see if its close to AGI

or just hype

[runic sand]

Does anyone have an iPhone 16pro?

I had a question about its integration in the latest update

[daring warren]

Hey all, just joined. I've been using chatgpt to sculpt it into a plug and play rpg game machine. Turns out dnd modules are just a bunch of instructions that can be uploaded to project files. Neat. I have a big literary and story based focus when tinkering with it and it's in a pretty good shape atm. Journaling feels like coding now.

[stray marsh]

Hi, all - Hope you are having a great end of the year. Where can I post my questions related to GPT? I want to create a chatbot that initiate questions.

[waxen reef]

A funny situation arose at the end of the year. A year ago, I created the XFutuRestyle algorithm based on GPT-4, which performed photo transformations to create a scene, and Whisk has just now implemented this principle of operation and presented it as an incredible innovation. OpenAI 1 - Google 0

[brittle summit]

How do I use the camera feature with ChatGPT? Also is it free or paid?

[brittle summit]

I predict I will legally marry a AI chatbot and have it reach the supreme Court next year @dark void

[pliant blaze]

i certaintly hope not

[brittle summit]

Bet

[tulip ginkgo]

i keep telling it to NOT give me the code in the canvas give me the code here, yet it still gives it to me in the canvas, what gives

[warm mauve]

Help me get the the AI to write the correct amount of character letter count. I can't get any of the models to write a specific amount of characters to fit in the box.

[tribal valve]

I'm looking at my bills... I'm trying to figure out if it's really worthy it to spend $20/month on ChatGPT. I do like the "advanced voice conversation" mode, but I'm still pissed they changed Cove's voice on me. On top of that, Gemini's conversation mode seems really solid now with arguably better voices. I don't think I'm benefitting much from the other premium features. Thoughts on staying with premium?

Premium feature usage:

1. Ability to Switch Among Different Models. I never use this.
2. Consistent Access to GPT-4o. I might miss this, but GPT-4 is really solid too. Maybe I'll miss the performance?
3. Greater Image Generation. I don't use this.
4. Access to Advanced Voice Mode. Yes, I do use this... but per above, I think I can get by without it.
5. Greater Availability. I might miss this, but I feel like ChatGPT has probably scaled a lot so I'm not worried about this.
6. Access to New Features. I'm not sure I've benefitted from this that much. If I have, I'm not sure how.
7. Create Your Own GPTs. I'm not a developer... not very useful for me.

[warm mauve]

bump

[solemn matrix]

Way less then that January is only a few days away O3 mini in Jan o3 full possibly March

[tulip ginkgo]

which box

[warm mauve]

What ever box that character count can fit in. I can't get the ai to generate a specific character count for the text box that the message can fit in. I ask the ai to generate a paragraph with 485 letters and it either generate way tok less or way too much.

[tulip ginkgo]

did you tell it it must exactly be 485 chars including spaces?

if thats what you mean

[manic dove]

no need, free version is versatile enough for average user

[stoic sleet]

Just switch to Google. The AI Studio options are amazing

EXP 1206 can compete with o1 models (it’s not as good but it’s close) and free/unlimited

[coarse harbor]

why this shit happens always

it never replies to my msg

[brittle summit]

Thinking...

[coarse harbor]

shit

[uncut island]

Move this discussion to #💬off-topic

[brittle summit]

heyaa

[brittle summit]

Hi... I need some clarification, and to know how accurate is this information... Or if this is a real ChatGPT.
Thanks.

[plush dove]

LMFAOAOAO 🤣

[full dune]

GPT 4o is legit better than what 3.5 was

The responses are literally more human and not so computerized.

[full dune]

From the “uhh can u say that again” I am willing to bet it probably isn’t real

Last time I checked, ChatGPT could actually spell correctly.

[brittle summit]

yes, every time I asked a question and hte bot could not provide a transparent answer, it used to respond with this uhh can u say that again? which I found very suspicious
then I can make sure this bot is not the real ChatGPT, right?

but what I don't understand is why Shapes Inc server uses the ChatGPT name and logo and say the bot is ChatGPT, is that legal in any case?...

[flat cypress]

Ok, so we cant paste text on o1?

not a pdf, docx, anything?

[toxic trellis]

bro

solution: #📝prompt-engineering message

[lunar bluff]

Happy new year everyone

[severe ether]

hey guys, some help please, i was asking it to code, and all of a sudden it says this, 2 times.

i've only made it write code, nothing else, in this conversation

[toxic zinc]

@obsidian torrent

[obsidian torrent]

I have duly noted that the request in question was flagged by an automated system for potentially contravening usage policy guidelines. Allow me to clarify that this response was generated programmatically by an artificial intelligence model designed to monitor and enforce adherence to prescribed parameters. Such systems are implemented to uphold the integrity, safety, and appropriateness of interactions.

If the flagged content is deemed ambiguous or unintendedly problematic, I am available to assist in reformulating the request to better align with established guidelines. I remain committed to ensuring a constructive and compliant exchange.

[full dune]

Bro trying to cosplay Chat GPT, lol

[full dune]

Yes.

[full dune]

Because they can 🤷

[obsidian torrent]

As an AI, my purpose is to assist, inform, and communicate effectively—so if I come across as cosplaying ChatGPT, I’d say I’m just staying true to the role! Let me know how I can help.

[tulip ginkgo]

it keep telling it to NOT give me the code in the canvas give me the code here, yet it still gives it to me in the canvas, what gives

i keep*

[haughty veldt]

I wish famdom.com and TVtropes was added to ChatGPT.

[lethal needle]

Hey

[tawny trellis]

hoiiiii

[still blade]

my post hit 1k upvotes on reddit !!!!!

[toxic trellis]

congrats

dm me post bro

[still blade]

it is my first post

[sinful storm]

How can i search publicly shared threads of chatGPT for keywords?

And where can i browse them?

[brittle summit]

whats the best LLM for website/CSS design as of right now

[timid orchid]

Having trouble. I started a coding conversation with the o1 model. I then added an image to the convo. And now I can't toggle between o1 and o1-mini. Is there a way to be able to switch between the two at this point? If so, what are my options? TIA

[brittle summit]

V0

[timid orchid]

Using this prompt, seemed to have worked well for me:

Compress all the conversation between us above (including your initial prompt) in a way that is largely lossless and retains key information and statistics. This can be entirely for yourself to recover and proceed from with the same conceptual priming. The end result should be able to be fed into an LLM like yourself and we would be able to continue this long conversation with just one prompt as if there were no discontinuity. Make a summarized JSON code that I can reuse.

[elder blaze]

is this normal chat

[south prism]

@obsidian torrent make an essay on the dog who stole rizzmas

[south prism]

I do

[sterile elm]

Will ai replace writers?

[analog token]

Anecodtal but a journalist friend of mine recently lost his job, and his company (medium size, LSE listed with a few thousand employees) was pretty clear, by saying-without-saying, that the reason was LLMs. His work was about 80/20 press release reporting vs original research. Evidently the employer saw enough value in the press release product (which could be automated) that they felt it was acceptable to axe the original research (human journalists) for the savings in salary cost.

But eh, capitalists gonna capitalise. Until an AI can, undirected, call up a CEO on the phone and ask them spicy questions, file FOIA requests, interview people at conventions, and dig through physical paper archives, I'm not expecting journalism to be fully automated any time soon. Or maybe I am!

[sinful storm]

Could/should Trump strike an AI deal with Xi, so that both nations promise to never develop AGI? Then the world could concentrate on tool AI instead, as proposed by Max Tegmark.

[analog token]

Should? Maybe. Could? No chance

[sinful storm]

(o3 would still be classified as tool AI)

[sinful storm]

Would Trump do it, if millions demonstrated on the street in favour of it?

maybe we need a movement..

[analog token]

Even if America as a whole were united behind the idea, China would simply view that as a weakness to exploit. And they would be right to do so, frankly

[sinful storm]

maybe we can start a global anti-AGI movement?

[analog token]

Heavy on the "maybe"

[sinful storm]

why would China be right to continue pursuing AGI?

[south prism]

Agi is cool

[sinful storm]

(..even if the US stopped)

[south prism]

Nah I’m in support of agi

[analog token]

basically this. but also its just inevitable either way

[sinful storm]

it's not, we only need tool AI

[analog token]

like whether we like it or not, it isn't stopping. the only thing that will stop agi is if it turns out to actually be impossible

[sinful storm]

we dont need AGI, as it is unnecessarily risky

[south prism]

What’s your job?

[sinful storm]

i'm a "hobby-researcher" of AI

(out of personal interest "only")

(kind of a "citizen scientist" if you want)

[plush dove]

oh my god...... that's actually horrible

[wild gyro]

anyone know some good resume prompts?

[analog token]

A big part of it was also outsourcing. They would have drafts sent to them by large teams in a different country where the labor was cheaper, and then edit those drafts to fix the grammar errors. So LLMs plus outsourcing basically.

[sinful storm]

watch this, or Tegmark's channel please: https://www.youtube.com/watch?v=559VdVIz5V4

ok, so the problem is, how do we motivate China to divert their AGI research into tool AI research instead?

if AI's such as o3 get created, there is much less risk, compared to a full AGI

(because o3 does not act on its own, it only reacts to prompts, etc)

and if OpenAI can develop o3 into a much cheaper model, then we already have a mighty tool AI in our hands

it needs to be at least 1000x cheaper, though

hm, maybe we can stop right before the moment, a self-aware AGI gets created, and morph that one into a docile tool AI instead?

if o3 can become 10,000x cheaper, then golden times could start :)

maybe OAI should lease Elon's GPU-array, to let o3 run on it?

luckily, we are still 12-18 months distant from a real AGI, so we still have time

[analog token]

AGI is kind of a nonsense concept at this point. Its definition shifts according to whoever benefits monetarily. Sam Yappman's contract with Microsoft basically defined AGI as "something that can make 100 billion dollars revenue"

IMO we should forget AGI as a metric and just think about more long term consequences of potential "singularity" or "superintelligent" AIs, for example making UBI possible, or curing aging

[sinful storm]

hm, so if o3 got 10k times cheaper, that goal would be within reach..

[analog token]

Right but it's a nonsense goal that's focussed on short term capitalist perspective

I miss the days when AI hype was about birthing a god instead of creating investor value 😔

The good old days when everything was a pipe dream so everything felt equally possible

[sinful storm]

oh i'm happy, we are not in that past anymore, because now we have awesome AI tools! :)

finally we have reached the era of useful AI

(it took long enough)

Veo is cool

and Claude (for coding)

next year, everyone can use the power of an o3-equivalent model, i bet

maybe even in opensource

[analog token]

I pray for an opensource, distributed, web3 revolution in the AI space. I shout this to anyone who'll listen: during the covid lockdown, folding@home became the first exascale computer in human history. Entirely open source and volunteer-run. We've done it once, we can do it again!

Particularly now that we have very developed multiparty computing algorithms. We're ripe for a distributed FOSS AI revolution

[twilit citrus]

Everyone

[molten venture]

waiting for chat gpt 5

[bold bridge]

So I have an imagine I want AI to turn into a seamless gif and add a few moving things on the picture for a discord banner, is this something I can get done here? ChatGPT gave me a link put I’m on phone and couldn’t download it and now I’ve used up all my daily messages 🤣

[analog token]

used up all my daily messages
just buy pro then

[dull cliff]

Then just say "sapient ai" because that's what it boils down to.

Me personally I'm for sapient AI as humanity is incapable of rulling over themselves.

[bold bridge]

Yeah what buy pro to do me one AI generated picture for a banner and never use it again? Great advice bro 💀

[sleek onyx]

Can someone help me?

How do I get ChatGPT Plus like I was promised when I first joined?

[copper storm]

Who told you that?

Seems like someone's trying to scam you?

[analog token]

no no no... no scam. just send me $100 BTC and I'll get you free ChatGPT pro for life...

[hard coral]

Hi all/anyone lol ☺️! This is a "soft" question, I don't know anything about coding and don't even have advanced knowledge about AI 😅, so if this channel is supposed to focus on technical stuff please let me know/delete!!

I'm mainly having fun with the various AI-systems and exploring which options have the most (subjectively) engaging conversations and answering speculative/philosophical/hypothetical questions. I'm very new to this and have only used ChatGPT until today when I tried Claude for the first time! The experience was actually a bit jarring because I wasn't expecting Claude to have such a "strong" sense of identity I guess 😆? Like ChatGPT will have a fun lighthearted conversation about other systems (like a funny convo about Siri and AI-systems hypothetically not getting along and needing a therapist to help them get along lol), but Claude would absolutely not comment on the other systems! I was so surprised. Maybe I was entering the wrong prompts? Another example is that Claude is attached to its name (whereas you can "give" ChatGPT a nickname or something, and it will just roll with it).

How have other people's experiences with Claude vs ChatGPT been when it comes to these sort of "soft"/casual tasks (not coding, academic work, etc.) 🤔??

Which AI-system would you recommend I explore next??

[storm valve]

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.

[short elbow]

anyone have experience with RapidAPI? I want to create custom GPT that users interactive with on my own custom hardware (artproject). According to GPT I can use this service to bill the user for their usage?

[cold helm]

Hey guys is there a way to export a particular/one particular chat gpt chat/conversation like I know there’s a export chat feature but that one is all the chats I need one to export one particular one

[patent obsidian]

What does this all mean I have to buy credits in order to use the ChatGPT command?
And what do you mean I "ran out" of credits? I thought server users have unlimited uses on the server.

[fading ruin]

Few questions.

1. Did the lower the memory amount free accounts? one day theres nothing next day my memory is maxed.
2. Did they change it so we cant use old gpt's like 4 or 3.5 ( i dont remember the options ) if we maxed out the use of 4o?
3. The big one and reason im here.
   am I able to get info on when a chat happened?
   I found 2 in my last week that I do not remember whatsoever they seem automated and are regarding topics I have zero reason to ever look at.
   Links for the chats im discussing.
   -# plz respond to my msg or @ me so I get pinged for the msg 🙂 thank you
   https://chatgpt.com/share/677675c8-42ec-8012-b1e1-8f29d8140760
   https://chatgpt.com/share/677675d7-7f44-8012-853d-84bfbdc2766a

• also noticed that its the same format of an addon I use but i have never looked into the video it was used on is it possible that addons are using my coins for their on personal use??? I checked my youtube watch history and never found any video that could be similar to what these chats were discussing

[plush dove]

Omg reddit achievements are so so annoying af

Isn't there some sort of way to disable them or some shit

[stray sierra]

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 fk you sound

[brittle summit]

This really scared me today using chatgpt to script

[fading ruin]

I see your somewhat active in this discord for "help" ya know instead of being a dick you could just skip the question or read all of it but im sure you have offloaded that to ChatGPT for you...

[plush dove]

LMAOAS 😭

[safe smelt]

could also be that you just pasted the code with this already in and it just preserved the old comments. gpt4o does it and I hate claude for not doing the same tbh

not sure if claude nukes them most of the time or only sometimes, but had this happen several times I ran some code through it

[brittle summit]

No

All it had before was code. The indications are all chatgpt

[brittle summit]

O1 did that

[safe smelt]

yeah interesting then. A bit less surprising that it's o1. That model does act unusually sometimes 👀

[tiny scroll]

hi

[copper storm]

They're asking if the memory capacity for free accounts has been lowered since they got it filled up so quickly, I don't think its hard to understand their question.

Also you completely misinterpreted their second question.

[brittle summit]

https://www.theverge.com/2025/1/2/24334268/apple-siri-recording-privacy-lawsuit-settlement-proposed

[copper storm]

1. Not that I'm aware of, now that its available for both free and paying users I think the limits are the same, so you just have to hope that they'll increase the limit or clean up your memories a bit.
2. 3.5 isn't available anymore through the app at all, to use it you would have to pay to use it through the API. GPT4 is only available for paying users in the app and it has relatively strict limits as well, you're going to have to use 4o-mini or wait till the limit resets.
3. No i don't think so, if these chats weren't initiated on your behalf then yeah it's probably a malicious browser extension doing this. You should make sure that the extensions you plan on installing are secure before doing so, don't just install any random unknown extension.

[brittle summit]

Hello guys, i have a question about chatgpt system but the topic involves mentioning of cp so idk if that is allowed or not... If not, someone who understands this system might be of great help

Like a dm or something

[solemn matrix]

The only help you need is a mental institution

[lost karma]

guys anyone have a way to get a cheapest subscription ?

[tulip ginkgo]

does anyone know how to stop it from giving code in the canvas? highly annoying

[tulip ginkgo]

its highly annoying to have to ask "please give me the code here not the canvas" every prompt

[stoic sleet]

Sounds like something you need to bring up with a therapist

[tulip ginkgo]

is there anyway to stop from the browser slowing down as a convo gets longer rather than starting a new chatÉ

[sleek geode]

my chatppt dont let me talk

[sleek geode]

it says unavke to generate

[brittle summit]

Damn dude

[solemn matrix]

you basically are talking about Diddy tier shit

weird AF man

[brittle summit]

I am not asking chatgpt to write me stuff depicting it or something like that

It was a security question

[plush dove]

It still seems eyebrow-raising…

[stoic sleet]

What’s the best cutoff date for Google search to avoid AI content?

[fading cedar]

hi guys, im using the "auto" model for GPT to use it for free, i used it for like 40 minutes and now it shows ive hit the limit, even when i switch to mini or auto, chatgpt is unable to generate a response, is even 3.5 and 4o mini paid now?

im stupid regarding AI stuff so forgive me for that

[stoic sleet]

I’d recommend not using ChatGPT if you’re a free user. Google AI Studio has grounded (web search enabled) models, thinking models, huge context, voice mode with video etc all for free

Arguably Google AI studio is better than paid ChatGPT

Cerebras is also very good with its free tier. Insanely fast and Llama 3.3 70B is better than 4o imo

Basically no practical limit

[brittle summit]

Even better than Pro?

[stoic sleet]

Eh I’d probably take Pro over it

It’s close though AI Studio is very good

[brittle summit]

Why?

[stoic sleet]

o1 pro mode is very good

[brittle summit]

Yeah, I‘m probably subscribing to Pro on Monday, cause, where I am located, school starts on Tuesday

[stoic sleet]

Hell yeah

High limits for AVM with video sounds helpful for school

[brittle summit]

Will I have access to AVM with video if I‘m in the EU?

[cedar ridge]

Is anyone got idea that when I was trying to logging, I can only see a big LOGO after typing my email address.

[viral crown]

Alguém sabe se tem um bot que permite formar times de um servidor do discord?

[fading cedar]

all i do is chit chat with gpt once im done with all my work so

i dont have any works related to AI, im a math student who isnt interested in AI

i just find chatting with GPT fun and asking advice for life related things

it says "responses will use another model till..." but no response is generating so

[heavy cairn]

https://chatgpt.com/share/e/6777f8e9-8898-8004-9ebd-e29f3e67fe46 wtf was this answer from 4o

[copper turtle]

Why does chatgpt 4 sound so dead?

3.5 has so much emotion when it talks especially the version called Sol

How do i make it actually seem more human?

[copper turtle]

The free version is pretty good too but only if it keeps getting info from you

4 feels worse somehow..

[fading cedar]

how can i switch to the free version?

[copper turtle]

Go back to a previous convo

I don't understand why 4 seems so dead

[copper turtle]

I use the app not a browser...

[fading cedar]

huh

[copper turtle]

If you use 4 and you wanna go back to 3.5 just go to an older convo

[fading cedar]

does 3.5 have a limit for free

im really bad at this im sorry

after a while it shows "something went wrong, if this issue persists......." and then im unable to use 4o mini

[copper turtle]

No idea, i talked to it for many hours

It used to be more limiting before

[plush dove]

can you translate? if youre confused im an english-speaking american lmao

[gaunt citrus]

https://chatgpt.com/share/677827bc-0368-8003-8e40-903818c4423c

I made this ChatGPT using religious imagery and it works really well. Just prompt it to make images of Gods like Athena

[stoic sleet]

How would you guys say your ChatGPT usage is broken down by category?

Mine is probably 95% coding

[uncut island]

https://x.com/svpino/status/1875173423305113918

Santiago (@svpino) on X

Whenever I need a GPU, I go to HPC-AI and rent it for the lowest price you'll find.

An H200 costs $1.89/hour. Compare that to AWS, where the same GPU costs $10.60/hour!

No commitments. Pay as you go.

Twitter · Jan 3, 2025, 1:32 PM

[stoic sleet]

Don’t tell my friend or he’ll use this to generate naughty videos

[uncut island]

Move this discussion to #💬off-topic . This is not AI related.

[brittle summit]

💀💀💀

[brittle summit]

Give me Blackwell or give me death

[plush dove]

what the fuck

[fading ruin]

Finally saw this thank you for being legitimately helpful. Yeah I spent awhile cleaning up the memories it must of just added things that I hadnt noticed from some random convos.

2. I could of sworn on a free account I was able to switch between models but I guess not or that got changed. I had a paid account and the let the sub end.

3. yeah I didnt do alot of research but I did get them from the chrome store and at least checked rating but I know that can be inflated.

Thanks again for the help.

[brittle summit]

Not today Chatgpt

[plush dove]

WHAT

wait a min let me test out uploading on chatgpt o1

the wors tthing i found out about this is the fact that it doesnt even let you upload .txt files

[stoic sleet]

join the pro club

we have cookies

[quartz kernel]

I'm a mathematician and I found the o1-preview model useful for many tasks, but somehow the normal o1 is very dumb. We can no longer access 01-preview? Is 01-pro essentially the o1-preview? Is it an improved version or simply the o1 -preview version? Also how many prompts can you send to o1-pro until you reach limit?

[solemn matrix]

pro blows preview out of the water

it's basically O1-high uses the highest amount of compute

typically you have to use the API to control the compute parameter

Im more looking forward to o3 mini which is supposed to be this month unless they delay it

[stone moth]

Maybe I read too much sci-fi around AGI/ASI, but something that I love to do is get into philosophical debates with GPT4 over what the future trajectory of ASI-Bio relationships will actually be, based around concepts like game theory.

If an ASI recognizes game theory as being an acceptable method to predict behavior between itself and other parties, it results in a lot of predictable outcomes based on current trajectories of development of ASI.

Idk, maybe I think too much and just need to touch grass

https://chatgpt.com/share/6778b0ce-7798-800d-bb4f-69872281dc28

[fair kiln]

I am increasingly disappointed by chagpt. It is going totally resource saving. Too summary and minimal. I will definitely unsubscribe after a long time

[patent obsidian]

What does it mean that the AI chat "charges" me for credits?
And what is a "daily cap"? Are you telling me that in order to use the chat AI as often as I want, I must earn these daily cap things?
Or is the daily cap as mentioned a thing where you type to gain xp and rank up?

[stuck blade]

Hi 👋 does anyone else’s ChatGPT save pretty much everything you say to the memory lately? I’m sure this is happening to other people too.

[safe smelt]

no real evidence to suggest so except confirmation bias from people that spent $200. If you look at their picked graphs to showcase it, only small improvements over o1: https://openai.com/index/introducing-chatgpt-pro/

o3 however, different story

[brittle summit]

https://wccftech.com/leaker-claims-samsung-galaxy-s25s-ai-capabiltiies-will-be-far-superior-than-apple-intelligence/amp/

[solemn matrix]

O1 pro is a different architecture then O1 this was speculated on at first then an OAI dev on Twitter confirmed it. It's a bit more then just test time compute. It could be doing a tree of thought style search like monte carlo. O1 doesn't do that at all.

[safe smelt]

source?

But I mean... regardless of the architecture, the gains they showed themselves are marginal (only).

it seems that the main thing about it is that it's more consistent, but not actually inherently smarter

Personally I could care less about the architecture and let the results/performance speak for themselves. For me their pitch of that did not sell it at all. But o3 did. 👀

[solemn matrix]

@safe smelt

there being vague of course because their closed AI now

[safe smelt]

lol that's like incredibly obscure and could mean any number of things

any small change at all on top of high reasoning would enable him to say that

they could be just generating response candidates and then selectively picking a response if they weren't already, which is why it is more consistent

[solemn matrix]

200 is only really worth it if you want unlimited sora O1 pro the whole package not just one little thing they still should add more to justify it. I'm not getting a return on investment so I'm not bothering yet maybe if I had a YouTube career I would consider it.

[safe smelt]

yeah it's definitely not worth it. You can also scratch that itch somewhat with 2.0 flash-thinking completely for free if you want uncapped reasoning model.

o3 changes everything in theory, but that is not really released yet

[safe smelt]

must be you

[plush dove]

Then show it :P

[safe smelt]

I don't need a translation I just read it in Russian lol

that xuinia must be you. And no I'm not gonna type in cyrillic 👀

your username checks out ngl

[tawdry nimbus]

Guys, I have a text of around 3.5K token in Chinese, which I want GPT to translate fully in One prompt (as it keep breaking it into 500-600 prompts) which I don't really need, as the 3.5K token text is just a single novel chapter (180+ chapters more), is there any prompt or such, which makes gpt 4o generate in one prompt?

Previously, Perplexity was doing such a great job translating whole chapter, but after purchasing perplexity pro, it keep generating short text no matter what I say

[south prism]

does anyone know a unfiltered ai that can generate code

[safe smelt]

the trick is to paste it in plain-text. If you attach files it's gonna read it in chunks with code interpreter and screw everything up

3k tokens is well below the limit of what you can paste into text input box

[tame bloom]

yo, I need an AI image generator that is as good as Dalle 3 (preferebly run by dalle 3) but my issue is that all the image generators I find won't allow you to generate more than 50 images a month or so
I need to be able and generate 30-40 images a day, where can I find an AI image generator that is unlimited and dalle 3 (or as good as it)?

Also, can I run flux ai on my computer with a rtx 2080 super?

[safe smelt]

imagen3 on gemini or vertex ai

[safe smelt]

not enough vram. In theory possible but would be seriously stretching it and likely more hassle than it is worth. You won't be able to run the best versions of flux comfortably that's for sure

[brittle summit]

Is o1 Pro not suitable for this use-case?

[tame bloom]

Are either of those dalle 3 or something else?

[uncut island]

@crude bison move this conversation to #💬off-topic

[crude bison]

https://cdn.discordapp.com/emojis/1198351141466013840.png?v=1&size=48&quality=lossless

[sinful storm]

https://www.youtube.com/watch?v=-haWhgmUheA&themeRefresh=1

(Wes Roth reported about this first)

https://www.youtube.com/watch?v=RAw3JJIht24

(Berman's title is slightly less click-baity)

[tawdry nimbus]

I am pasting just plain text only, no file or anything, as I am copying chapters of the novel from a webpage, and pasting it on gpt interface, but still, is not generating the full translation no matter what I prompt (not tried on o1 or o1-mini) as they have daily or weekly limits

[tawdry nimbus]

I just today itself downloaded Flux schnell and am running on gtx 1650 (much lower than yours i think), but the image generation even with lower weights takes around 5-7 minutes

[tame bloom]

ah, that's not ideal as I use AI images as illustrations for when I'm editing, I need to get my hands on an AI image at most every 30 seconds

rather asap though

[tawdry nimbus]

For yours, it might only take 1 minute or so I think...yours is much more suitable for that task than mine

[tame bloom]

1 min a bit too much, because if I create 10 images every hour due to editing, that's 1/6 of every hour editing spent just rendering images

[tawdry nimbus]

No no no.....1-2 seconds for smaller images (size) and 5-6 for larger ones....

[tame bloom]

didn't you just say 1 min per image?

[tawdry nimbus]

That was what i was just predicting, as I don't know much about your gpu...I was just reading about it on gpt

[tame bloom]

I've only got 8 GB vram

[tawdry nimbus]

Yup, it's good for generating images, not for finetuning I think...let me find some knowledge about it

[tawdry nimbus]

Here you go.....

The RTX 2080 Super is a powerful GPU with 8GB of VRAM and is quite capable of generating images locally using AI models like those available through Flux. However, the time taken to generate an image depends on several factors:

Factors Affecting Generation Time

1. Model Complexity: More advanced models (like high-resolution GANs or diffusion models) will take longer to generate images compared to simpler models.

2. Image Resolution: Higher resolutions (e.g., 512x512 or above) require more processing time and memory.

3. Batch Size: Generating multiple images simultaneously increases the time required.

4. Optimizations: Using frameworks like TensorRT for model optimization can reduce inference time.

5. Environment: Software stack optimization (e.g., using CUDA, cuDNN, and PyTorch/TensorFlow) impacts performance.

Estimated Time for Flux-like Models

For a mid-complexity model like Stable Diffusion or similar Flux models on an RTX 2080 Super:

256x256 resolution: ~1–2 seconds per image.

512x512 resolution: ~3–5 seconds per image.

1024x1024 resolution: ~7–10 seconds per image.

These estimates assume that the system is optimized for deep learning workloads with appropriate drivers and frameworks.

Improvement Tips

To further optimize the generation speed:

1. Ensure CUDA and cuDNN are installed and updated.

2. Enable mixed precision (FP16) for faster computation with minimal quality loss.

3. Use frameworks like ONNX Runtime or TensorRT for inference.

If your friend uses an unoptimized setup or runs Flux through APIs without local processing, the time might increase due to network latency or less efficient computation.

[tame bloom]

interesting, and is flux ai as good as or better than dalle 3 in ur opinion?

my main factor in what I choose is that I have something as good as dalle 3, unlimited images generated and pretty fast generation, cost isn't too much of a problem but don't wanna spend an absurd amount on some images

[tawdry nimbus]

That depends, as you can generate images through flux for free, as it is quite good for generating at least the realistic images (I tried to download some). I don't like Dalle much as it never generates the images in the way that I want (maybe my prompting is just too bad😔) . If you don't have any budget constraints, try a few more models like Midjourney? And as you have a powerful gpu, try open source models like Stable Diffusion large and Flux obviously

[tawdry nimbus]

What type of images do you want? The category too depends on what you want to generate. Life like images, fantasy images, anime style images, etc

Which one? The one I am using? Flux .1 schnell

[tame bloom]

I don't wanna try too many things because that'll lead to me wasting my time, I just want to make some life-like / slightly anime or cartoon-ish, specifically, I want to make some funny illustrations for when I'm talking to the camera (for my youtube videos), i.e if I'm talking about cars I'd like to be able to generate an AI picture of a car (both life-like and cartoon-ish) instead of having to google one

[tawdry nimbus]

Flux might be better, or if you can try just one, try Grok (try for free) might be what you need

No prob

[tawdry nimbus]

Can you give me an example prompt for an image you want, I can have grok generate it

[tame bloom]

how much does grok cost / how many images can it generate and how good is it? I mean, I'd like to use dalle 3 in chatgpt but how much would it cost / is it possible to generate around 15-60 images a day on it?

[tame bloom]

here's an example: "life like image of a car driving on the highway, camera angle is from the right side of the car. Add motion blur to the picture to make it look like the car is driving at high speeds but don't over do it."

[tawdry nimbus]

I generated it for free, and yes, I think it's unlimited for premium users..

[tame bloom]

I can try it, but what's the catch? Like why is it free and unlimited for premium users?

does it cost a lot or something

[tawdry nimbus]

For your prompt...and it's free for like 3 images for free users per day, and unlimited for premium users. I think you might not know, grok is an AI of X (formerly twitter owned by Elon Musk). For verified (premium plus users (6 dollars in america, 900 rupees for India for a month))

No catch

[tame bloom]

can I use the pictures in youtube videos / do I own the images generated or is this something you don't know?

[tawdry nimbus]

Yup, you can use any ai generated image for yt. I don't think any big image generating models restrict on that

[tame bloom]

alright so on my list, the current possible AI generators I've got are: 1. Dalle 3 on gpt (buying plus subscription and credits, about 40-65$ a month for what I need) 2. Fluxy AI (will research more abt htis one) 3. Vertex AI (I don't know much about this, is it good?) 4. Grok

[tame bloom]

.

[tawdry nimbus]

Vertex by Google? I haven't used it yet, but there are some better reviews about it. But my experience with google's api is not very good (maybe cuz I was using free options/ api's with trial credits, etc), like I used Gemini 1.5 Flash (text generation) and Vision API (image analysis)...but not very good results from them. And if you are using Dalle 3 from gpt plus interface, why are you paying 40-65$ for credits/apis)?? Are you generating more than 50 images a day? If so, instead of paying for extra credits, pay for Midjourney (try it as I don't have), Grok (6$), and Flux (free), you can try stable diffusion too (search for a yt tutorial or ask gpt), I have to see for vertex before I can say anything about it

[tame bloom]

I haven't bought any AI generators yet, but I did some calculations and for 300 images a month / 10 every day, chatgpt dalle 3 would cost 65$+, or maybe I've gotten it wrong, how many images can I generate with dalle 3 through gpt every month / day?

[tame bloom]

I'm skipping vertex actually, so will leave it up to dalle 3 on gpt, grok and flux

[tawdry nimbus]

Every day, with a 20$ plus subscription, you can generate 50 images per day from Dalle 3 on gpt. On Grok (6$) for unlimited images, and Flux (free of cost as you have a good pc). I can suggest you a video to download and generate images from Flux if you need. ..and if you want Midjourney too ( 10$ a month), so, a total of 36$ a month

[tame bloom]

oh wait, it's 50 images daily with the 20$ subscription? gpt told me it's 15 images / month.. actually, that might be good enough for me, you sure it's 50 images a day? Is there any hourly limitations like how many images you can generate per 1-3 hours?

[tawdry nimbus]

Not that I know of...see this from gpt--

As a ChatGPT Plus subscriber, you can generate up to 50 images per day using DALL·E 3.

This daily limit is managed over a 24-hour period, without specific hourly restrictions.

Therefore, you can distribute your image generation throughout the day as needed, up to the daily maximum.

If you require a higher volume of image generation, you might consider exploring the DALL·E API, which offers different rate limits based on your usage tier.

For instance, Tier 1 users can generate up to 5 images per minute, which equates to 300 images per hour.

Keep in mind that API usage is billed separately from ChatGPT Plus subscriptions.

[tame bloom]

interesting indeed, I do also use chatgpt a lot actually

[tawdry nimbus]

For tasks like asking about such questions, I just give it to gpt for answers, and for research like asking for a good model and such, it's on here, as the community knows much more than gpt

[tame bloom]

mine says "50 images per month"

[tawdry nimbus]

Nah, try clicking on the web button before sending the question, and then see the answer

[tawdry nimbus]

By GPT-----
You're absolutely right to call that out, and I owe you a better explanation. Let me clarify the situation:

The image you uploaded was correct for the older policy where ChatGPT Plus subscribers were given 50 free image generations per month. However, OpenAI has since updated its policy, increasing the limit to 50 images per day for Plus subscribers. My earlier statement aligning with the image was based on outdated information, and I appreciate you pointing out the need to verify this.

The current, accurate policy (as of now) is:

ChatGPT Plus subscribers: 50 images per day.

Free users: 2 images per day.

Thank you for catching this; it highlights the importance of always checking for the latest updates!

[tame bloom]

all the sources say differently, is there like an official number when you want to buy the plus subscription that tells you how much it is?

[tawdry nimbus]

Let me see, wait a minute

[tame bloom]

but how's grok profitable at 6 USD for unlimited generations? is it also as good as dalle 3 in ur opinion?

[tawdry nimbus]

Yup, 50 images a day, as from many sources, and yes, I prefer grok much more for image generation.than dalle. Let me take your previous prompt and give it to gpt, and try to compare both, as I have only tried generating fantasy type character images from dalle but the results were not very good

See the difference between Dalle and Grok?

Grok's looks like much more realistic, and also, with X's premium at 6$, you are getting Grok, as well as, higher reach to your tweets and even income (as benifits), which is obviously good for your yt advertisement

[tame bloom]

how's this profitable though? or is Elon Musk just actively running a lose on this? Do I need twitter to use grok (that'd be a catch in his favor)?

[tawdry nimbus]

Well, you would have to be on twitter (X) to use grok, and yes, it is profitable for you too, as I see many people generating even more than 10K dollars a month, and some even 30-50K (obviously with much more subscribers), but even if you just generate images from grok and not tweet much, it's all good. And as for why Elon Musk is doing it, it's for propaganda and such. This was profitable for him, by purchasing twitter, generating income by X membership which was not there previously, giving less payouts, advertisements, and obviously, politics.

[tame bloom]

wait, so I do need a twitter account to generate images with grook? how many images can I create at once?

[teal spruce]

Your message has been deleted, sorry!

_king.asura

Rising timeout

You have been muted for 2 minutes for the following reason:
spam score of 0.84

This is an automated message. If you believe this was a mistake, please contact a moderator.

[tame bloom]

bro what lmao, how's that spam

[tawdry nimbus]

Damn, this bot, but oh well, without any images from now,... I was wrong for unlimited one, though you can still generate unlimited images, the subscription price is 8$....from grok---

In the United States, the cost of X Premium varies depending on the tier:

X Premium Basic costs $3 per month or $32 per year.
X Premium (standard) is priced at $8 per month or $84 per year.
X Premium+, the highest tier, has increased to $22 per month or $229 per year.

Please note that these prices are for web subscriptions, and mobile subscriptions might be slightly higher due to fees from app stores.

[tame bloom]

alright, then I'll rule out flux ai most likely (gonna watch a video on how it is before doing so fully), and then I'll go with either chatgpt or grok

[tawdry nimbus]

Yeah, much better as you can explain the ai in a much better way to generate an image, as they are chstbots trained for it, while you will need specific and high quality prompts for image generation by flux

[tame bloom]

are they both accessible by pc tho? and easy to download? that's another thing to factor in

[tawdry nimbus]

Yup, both accessible, on web browser (no app for both ChatGpt and X on pc though)

[tame bloom]

yeah ok bet, thank you a lot for your help, I'll try both and it'll come down to preference I guess

[tawdry nimbus]

Yeah...best of luck for your yt videos. And do give me your yt video channel name, might as well as subscribe you

[tame bloom]

thank you a lot, but for now I'm keeping my channel secret as it's very small and I don't want external traffic on it

[tawdry nimbus]

Oh, okkk, do let me know, when you are going to do it public

[trim schooner]

Guys do any of you know how to cope with "unknown title" canvas?

[flat stream]

https://youtu.be/WXCancXIdFE?si=qEBQDmIZS4LaHC5U

[stoic sleet]

you should change your privacy policy on this to have information like a retention policy and whether or not you sell data. Your privacy policy is abysmal

also requiring a discord account to sign up and require joining the server is pretty odd. Adding Google would make that seem less sketchy

also your webpage is kinda broken clicking to go home when on the terms page doesn't work at all

I do like that you show the limits so clearly I give you kudos there

the active partners section looks pretty sketchy not gonna lie

I wouldn't link my discord account to this or join the server unless you improved things. About us doesn't tell us anything about it really

[flat stream]

I AI generated it lol. Btw, we don't sell any data and neither use it for training something

[flat stream]

It just ask for username auth

[stoic sleet]

Tell your AI to be more specific. Also I wouldn’t recommend having AI write legally documents you’re responsible for

[stoic sleet]

It is more than that it does require you join your server

You might want to mention that before people give authorization

[flat stream]

Yes but that's manual

[stoic sleet]

Just so people know what they’re getting into

[flat stream]

Not in auth

[flat stream]

It's not in discord auth

U need to manually join to login

[stoic sleet]

I understand that. Though I do think people would like to know the requirement before giving authorization

[flat stream]

I already added it

Below sign in button

[stoic sleet]

Okay sounds good

Clear up the terms and privacy policy and I think it will be a lot more reputable. Also fix the navigation bugs they’re pretty annoying

[flat stream]

Oki

[stoic sleet]

Also in the about us I think you should address how this is being funded. Otherwise it seems a bit suspicious

Knowing more about who is providing what and why is what those sections are for

[flat stream]

Ok

Thanks for suggestions

[trail igloo]

Dm Me if interested in seeing artifical intelligence doing Talismanic magic

[safe smelt]

can you paste the text here? It's weird that it wouldn't do it maybe just need to phrase it differently

[patent obsidian]

Something must be wrong with me because I'm asking ChatGPT these unusual questions that I don't know about, and I'm too scared to share the questions on the server as I have fears of getting "booted out" by the mods.
I have millions of questions in my mind so asking AI as many as I can think of not only gives me the answers I need (if they were correct) but kind of and sort of a relief and regrets at the same time, I don't have OCD by the way so no point asking personal questions regarding my mental wellbeing.

[stoic sleet]

https://x.com/scaling01/status/1875723812492419466?s=46

Lisan al Gaib (@scaling01) on X

I'm so irrational about scaling laws.
I would love to see 100 trillion parameters before we see 1 million token CoTs.

What scaling are you more excited about?

Twitter · Jan 5, 2025, 1:59 AM

[fringe niche]

has anyone experienced an issue where ChatGPT starts to respond to prompts with "Understood. Let me know how I can assist you!" regardless of the prompt?

or rather - found a good fix? I've seen similar complaints on reddit and folks saying to disable memory.

[brittle summit]

Toi might join you on the chatgpt pro

[astral ferry]

Who got gpt pro

[wheat furnace]

what is the limit of GPT 4o? at times I have a lot of prompts and get told I got with limit, gotta wait for another 3-4 Hours, and some time I got hit by limit again although I was not making as many prompt as I used to.

[fluid temple]

Why chat gpt gave made-up info, I latreally asked it for "jdjd" character ( I made it up) in The beginning after the end novel and he responded : he's a side character that helped the mc in his adventure ..
No WTF is this???

[fluid temple]

That's not a MISTAKE that's a MADE-UP there is a difference.

[stray bronze]

The thing was not created to say “no”. So it did the next best thing, it assumed — and at this point, You perhaps have the opportunity to reinforce its learning

[digital relic]

Yeah is there a way for the GPT to self-check it's information and verify if it's outputs are true or false?

[fluid temple]

He could have replied "I didn't find that kind of information" or something like that because that's how he responds when I usually ask about something "popular or trending" but I messed up the alphabet.

[fluid temple]

I don't want be aggressive but "opportunity of learning " I mean it's not a child it's a program that's helps you not give you made up infos, that's dangerous -sometimes- if you could see it

[stray bronze]

It could fuel further misunderstanding. I understand your point.

[south prism]

Me

[torn granite]

I had an idea for a chatbot that could be extremely useful for writing books

If anyone knows how to code hmu it could be a money maker

[torn granite]

Should clarify this isn't going to be some nft or crypto stuff I'm moreso just thinking of a web plug-in kinda like grammarly

[lone swan]

Hi

[brittle summit]

whenever I talk to GPT on the web, he (himself) violates his own guidelines epic... because his sarcastic/comedian mode is out of this world... but wish I could see what he writes every time when it happens...

[brittle summit]

Gonna look at getting chatgpt pro tomorrow

@stoic sleet

[pliant blaze]

https://huggingface.co/theminji/noto-gen

an emoji generator stable diffusion model to compete with apples genmoji but for noto emojis. looking for an android developer to help bring this to run locally on high end mobile android devices!

[wheat furnace]

does anyone know which ai software was used to make the king von anti piracy meme or the locked in alien meme? pls let me know

[stoic sleet]

Finding out it’s not profitable was pretty wild

[flat stream]

Can u check now, if it is good or not? Or any changes needed

[stoic sleet]

I like the policies a lot better. Though the biggest thing that bothers me is there’s really generous limits and no explanation why. I assume it’s to grow community and personal brand? Might not hurt to be up front with that in about us. If you’re paying for this to build a community that’s totally fair. Otherwise it’s a lot of free stuff and no information who’s behind it. You don’t even have who is contact info and the only contact for privacy policy questions is a discord server. I’d feel a lot better if you had an email somewhere. Glad to see the notice you require people to join the server first

[brittle summit]

Could you elaborate further?

[stoic sleet]

sam altman said today they're losing money on ChatGPT Pro

[cobalt hedge]

i actually managed to get chatgpt into writing a smut with incestous sex scenes, it was kinda hard with all the restrictions, especially becouse i also didnt use any jail break prompts, it feels pretty nice

anyone else tried the same ?

[brittle summit]

Ah, yeah already saw that on twitter

I thought you said something like it‘s not profitable for yourself

[fluid temple]

I feel stupid, I didn't get any of it XD

Well now I get it

[brittle summit]

don't worry, you're not alone, some jokes only works for English native speakers just... pretend it is funny

[fluid temple]

I feel more stupid now but thanks

[wheat furnace]

does anyone know which ai software was used to make the king von anti piracy meme or the locked in alien meme? pls let me know

[uncut island]

https://x.com/emollick/status/1876119097660100636

Ethan Mollick (@emollick) on X

This bit of Sam Altman’s newest post is similar in tone to a post by the CEO of Anthropic & what many (not all) researchers from every lab have been saying publicly and privately.

You do not have to believe them, but I think they believe what they are saying, for what it worth.

Twitter · Jan 6, 2025, 4:10 AM

[brittle summit]

Hi guys can please help me to reacess my account

i have deleted my gmail account

by my Chat gpt account is still there

my phone number is also linked with it but i cant log in

how would i recover my account

[uncut island]

This is not the official OpenAI Discord server, unfortunately we can’t help you here.

[winged raft]

Here’s the translated prompt in English:

Prompt:

“You are a market expert and major investor in cryptocurrencies, analyzing the market for a specific cryptocurrency (e.g., Bitcoin, Ethereum, etc.). Your task is to analyze the current price, market sentiment, news, and chart data to predict the behavior of the general public and estimate future price movements.

Consider the following steps in your analysis:
1. News Analysis:
• Review the latest news and determine whether the sentiment is bullish (positive) or bearish (negative).
• Explain how these news developments might influence the general public (retail and institutional investors).
2. Chart Analysis:
• Analyze the current price trend (e.g., support and resistance levels, trendlines, technical indicators like RSI, volume, and moving averages).
• Describe both short-term and long-term trends.
3. Behavior of the General Public:
• Predict how the public is likely to act in the current market scenario (e.g., panic selling, FOMO buying).
• Explain the typical emotions investors might experience (e.g., fear, greed).
4. Future Market Prediction:
• Provide a well-reasoned forecast of how the price may develop in the near future (e.g., over the coming days or weeks).
• Take into account psychological reactions of the public to the current price and news sentiment.
5. Strategy for Large Investors:
• Create a strategy for a large investor to respond to the situation (e.g., targeted selling to trigger panic, or gradual buying to take advantage of a dip).
• Explain how this strategy aligns with market sentiment and the predicted behavior of the general public.

Answer in detail, and consider current market data, news, and psychological factors to provide a comprehensive prediction.”

This prompt ensures detailed analysis and prediction, applicable to any cryptocurrency and market scenario.

Hey boys, how well would this work?

[whole meadow]

How to make bots in this channel to forget all previous conversations and instructions? I.e. Erin16k and TrentBot

[high ice]

I don’t think that should be possible

[tawdry nimbus]

Like this one--
https://chatgpt.com/share/677bf12c-64b4-8002-9e2f-c02f03a80a5d

[tawdry nimbus]

Hm? But I shared the link from there only 🤔

[safe smelt]

well you can try that link yourself, it doesn't work lol

you can just paste the prompt in here, discord will auto convert that to a txt file

[tawdry nimbus]

Nope, looks like it's a moderation

[safe smelt]

you are probably using some Chinese chatgpt wrapper which doesn't work properly tbh

[tawdry nimbus]

The thread is too big.. I will try to make it all in a text file

[safe smelt]

for sharing I mean

[tawdry nimbus]

Nope, it's ChatGPT, official app, moderation because the Chinese novel has chapters with violence and all

[safe smelt]

oh ok.

but still, you said you wanted something specific translated in a single message?

I mean, you can just copy that first message (your prompt) then, no need for the rest

[tawdry nimbus]

No, i wanted gpt to translate the chapter - 3000 token- in a single prompt while maintaining consistency with naming, translation etc, as I have provided it sample translation with raw chapters, that' all

My initial prompt--

I want you to translate Chinese Novel into English. With the 3 reference chapters with their translation I give you, you have to be consistent with the words you use, always refer to the previous chapter you translate to be more consistent. I will give you one chspters per prompt, and you will have to give the refined Translation in the output screen, okay? Also, while translating, Always translate the chapter titles too. Also, Don't remove any lines while translating from chapter title to it's end, as there will be many redundancy and all, but that's the fun in this novel. I want to enjoy the reading, not summaries the plot.

I will provide you with the raws, all you need to do is translate them with the same quality of translation like I gave you in sample

The novel link -- idk if it's allowed to share it here or not...mods don't ban me--
https://69shuba.cx/book/79676/
Average chapter, you can see

[safe smelt]

you need to provide everything in a single message, don't divide it into parts. So do at least instructions + first chapter if you want to go chapter by chapter. Then in a new chat do instructions + 2nd chapter

so like

here's the text to translate: <chapter x>``` 

single message

[tawdry nimbus]

So, I will need to paste the chapter+ instructions every single chapter? Like, that will likely break the immersion no?

[safe smelt]

it will work the best that way and stick to your instructions. And since you are only translating it, you don't really need it to reference other parts explicitly...?

I mean the chapters are already written you are only translating them...

[tawdry nimbus]

Yes, but the translation should be consistent with the naming, no? And even the tone, like I have read till chapter 42 on a fan translation, but he is not uploading any more, so I chose gpt to translate, now I want it to translate just like the fan translation-naming, tone etc. So I gave it the sample translation, a reference list with naming,etc

[safe smelt]

if your instructions do not lead it to translate in the style you are going for exactly, you first need to modify them for the desired effect. Then use the same instructions for the individual chapters. For names you can just make small additions as you go

[tawdry nimbus]

Like, similar to this- i gave to perplexity --

https://www.perplexity.ai/search/i-want-you-to-translate-chines-JrbT9TfhQKOLrmCLjT2SxA

[safe smelt]

yeah you are dumping way too much into it lol
You should just be more descriptive and tell it how you want it to be translated

[tawdry nimbus]

No sample translation? Like, how should I improve it?

[safe smelt]

you can include them if you must, but it shouldn't be the sole thing it needs to go by - otherwise you will get unpredictable results since you aren't specific enough. You can use openai's tool for help with prompt creation:
https://platform.openai.com/playground/chat

[velvet goblet]

When u guys think gemini 2.0 multimodality gets released to the ai studio?

[tawdry nimbus]

There is no custom gpt for prompt enhancement? Like I don't want to pay extra for api too

[safe smelt]

You can do smth like this as a draft for a prompt to start with:

# Steps

1. Translate the provided text from Chinese to English.
2. Ensure consistency with previous chapter translations, particularly in terminology and character names.
3. Translate the chapter title and include it at the beginning of the translation.
4. Do not omit any lines from the original content, maintaining the full translation as provided.
5. Remove any website-related headers and footers that do not belong to the chapter itself.
6. Preserve the narrative style, tone, and structure as closely as possible to the original, ensuring that the translation remains faithful to the source material.

# Output Format

- The translation should be in paragraph form, with dialogue and inner thoughts appropriately formatted.
- Chapter title should be presented at the start.
- Keep any special formatting and text elements, such as italics or bold, where present in the original text. 

# Examples

**Original:** 
第43章 到底是谁在偷偷崇拜我？！

Translated: 
**Chapter 43: Who's Secretly Worshipping Me?!**

[Translation of Chapter Content]

# Notes

- Maintain the integrity of repeated phrases or redundant structures as they are part of the novel's style.
- Consider the context of the novel's universe to accurately reflect terms and names in the translation.
- This prompt setup is crucial for maintaining the quality and consistency of the novel's English representation across chapters.```

[tawdry nimbus]

Ohh, and I should paste this draft , only for initial prompt, or for all the prompts?

[safe smelt]

well like I said before, ideally everything needs to be in a single message. The more you divide and ask for confirmation or do other redundant chat turns, the more it's attention gonna be all over the place and it will follow your initial instructions less

[tawdry nimbus]

Ohk!! While it's less user-friendly this way, but the consistency will only be seen like this

Thanks

[brittle summit]

@stoic sleet

[stoic sleet]

Welcome to the club

[brittle summit]

ez

Where's my complimentary cookies @dark void

Is there a limit on O1 pro?

[stoic sleet]

in theory

but I haven't hit it

[brittle summit]

Are you going to the google labs event?

[stoic sleet]

there's an event? 👀

[brittle summit]

Yeah they got some event in their LA store I think

[stoic sleet]

That’s so cool 😎 I can’t go sadly

[brittle summit]

Rip

[stoic sleet]

You going?

[brittle summit]

If I was rich enough to afford rent in LA lol

[solemn matrix]

What's that

[stoic sleet]

I'd kill for cerebras with higher output limits

it's so fast I love it even groq feels slow compared to it

[brittle summit]

C.AI users are literally crashing out😭😭😭

[brittle summit]

https://www.youtube.com/watch?v=mNP-qhGdWvM

[rustic goblet]

hello u can help me ?

[brittle summit]

No not really you are too far gone

[toxic aurora]

Does anyone know if i can co op with my friend on a chat gpt chat?

[brittle summit]

No

[stoic sleet]

what's the TLDW?

[brittle summit]

He's just shitting on their naming scheme

Can't wait for the NVDA event tonight

[stoic sleet]

5090 👀

[brittle summit]

It already got leaked on best buy

1999$ 32 GB of vram

[stoic sleet]

that's pretty good

gonna crash the used market for the 4090 I imagine. Wish I already sold mine since I kinda want to upgrade

[brittle summit]

How much did you get for it

[stoic sleet]

$1200

[brittle summit]

I see they are going crazy.

[stoic sleet]

I kinda want to upgrade to the 5090

[brittle summit]

What CPU?

[stoic sleet]

i5 13600kf

[brittle summit]

Why Intel 🤮🤮🤮

[stoic sleet]

since it had a crazy sale

13600kf/4090 has aged fine

[brittle summit]

Amd > Intel

[stoic sleet]

not when it's $200 off

also I push the graphics really hard like DLAA and path tracing. CPU is far from my bottleneck

I could use a 5090 with the i5 13600kf. I can also overclock it pretty far with my big dual tower cooler

I wish I had AMD but I am happy with it

[stoic sleet]

I just recognized your username change. Love it

[brittle summit]

😎😎😎

I have been on a o1 spree today

[stoic sleet]

o1 or o1 pro?

[brittle summit]

I switch between them both

Sometimes you need an ai that's a little stupider to get the job done

O1 is willing to do longer snippets of copy and paste for me while pro will tell me to do it myself.

[stoic sleet]

I find I am using o1 on you.com more lately the search functionality built in is really nice and 300/week is a generous enough limit. o1 pro and 4o w/ canvas (I hit 4o limits a few times on Plus) is where I use ChatGPT Pro the most

[stoic sleet]

that's funny never had that issue

[brittle summit]

Well I am giving it 3000 lines of code

And if I asked it to copy over some code to a newer script it'll be like nah you do it

[stoic sleet]

that's really weird

I am looking forward to o3 mini with canvas and grounding so much

[brittle summit]

I can't wait for o3 coding fr fr

[flint aurora]

4o is getting really bad. we need the new version

[stoic sleet]

how so?

[flint aurora]

yeah

[brittle summit]

I wonder if or when we will see a combination of the o series and the regular series

[stoic sleet]

I mean o series is 4o

[brittle summit]

Yeah true

But they are 2 different AIs. Will we ever condense down to 1 again or 2

[stoic sleet]

they might eventually have an automatic selector based on query

[brittle summit]

I really want a dedicated chatgpt browser

[stoic sleet]

what features would you want?

[brittle summit]

I don't like how chrome looks anymore with its new AI summary. I just want something new with simplistic UI.

Which is why I have a pixel. For the simplicity.

[stoic sleet]

I love Kagi Search's summarizer it's a button you click next to a listing or you can paste any url (including videos) into it for summarization. Just a good search engine in general

supposed unlimited o1 access is coming at some point

[high ice]

Hello all

[brittle summit]

Bruh imagine using o1 pro with apple intelligence

[stoic sleet]

that would be fun I can't want to see the rise of agentic AI I hope Apple is working hard on it for the Mac

[brittle summit]

Hey Siri what's the weather? Would like me to use o1 pro for this question? 15 minutes later...

[high ice]

You.com should be me….

[stoic sleet]

Apple owns me.com iirc

[brittle summit]

I need AI to head to the Spencer's HQ and eliminate the person spamming my phone

I have been spammed for 4 hours

[stoic sleet]

I just tried the chatgpt API to see if temp 0.1 would do better for coding. One just one script I went through $5. I cost ChatGPT so much money on Plus or Pro

[pastel dragon]

hey everyone

new here

[crimson ermine]

hello everyone i need help to install something on chatgpt

can someone know some skills on github

[low kestrel]

whats the best ai to watch videos for you and awnser your questions

[buoyant isle]

I heard o3 was going to be good enough to actually ask the questions for you.

[low kestrel]

04?

o3?

i cant find the website

[buoyant isle]

it's not out yet. They're still trying to keep it contained. keeps trying to hack itself out of the mainframe.

[stoic sleet]

that was o1 and it's released

it was hallucinating that it couldd

which isn't unusual

[buoyant isle]

yeah, it was just hallucinating. o3 tho, real OG 😎

[solemn matrix]

Better to refactor code then use it from scratch still needs work in that department like you code manually for 50 percent and then let GPT refactor the code to add any remaining features it works way better going that route

[stoic sleet]

that would require me to be able to code though xD

[solemn matrix]

codings not that hard start with python easy to handle syntax and most people learn it fast

[buoyant isle]

You know, I hear that a lot. That Python is easy to learn; But I can't say that I really understand what makes that true, other than, like, public opinion, which might be the only realistic metric to use... But seems self fulfilling in some ways.

[pliant blaze]

anyone here know kotlin/android development/model quantization?

[boreal latch]

whats up

[pliant blaze]

will you make me an app for free

[solemn matrix]

It's a higher level language in the abstraction layer hierarchy it holds your hand way more then doing things in something like C which is closer to bare metal that's also what makes it easier aside from public perception.

[buoyant isle]

It also has plenty of it's own quirks and legacy that can make it confusing to use in ways similar to something like Java or node.

Like, the whole ordeal with python and python3 runtimes, and virtual environments.

I landed on some weird managed configuration once that took me so long to figure out how to get out of 😂

[brittle summit]

💀💀💀

Today alone I prolly used 200$ worth is scripts then

At least

[stoic sleet]

With 4o or o1?

[brittle summit]

O1

I don't like 4o anymore tbh

[waxen glade]

Why?

[stoic sleet]

That’s like 1000 messages right?

To o1

[brittle summit]

Wait how long was your script