#chatgpt-discussions

1 messages · Page 87 of 1

shrewd snow
#

Also, the voice in the demo was kinda flirty 💀

regal portal
#

if you look at the other videos they have on the openai channel, they try a few of the other ones.

#

I already know i'm going to ask my voice to be rude and condescending. Much more entertaining that way.

shrewd snow
#

Lmaoo

#

That would be amazing

#

I'm afraid ai is gonna take a lot of jobs lol

regal portal
#

Same thing happened when computers came around, but people still have jobs.

dense halo
#

I want GPT-6

shrewd snow
regal portal
#

You won't be seing an android replace bag boys and burger flippers at mcdonalds in the next 10 years, that's for sure. Cheaper to higher a zitty teenager.

regal portal
#

and new tech always scares people

torpid cobalt
regal portal
#

And when all the jobs are taken, party at the beach with mojitos for everyone!

shrewd snow
#

Also, instead of human partners, it'll be AI. Which sucks. I don't want that to happen but I can feel it'll def happen

eternal rockBOT
regal portal
shrewd snow
#

It's gonna be like JOI from Blade Runner 2049

regal portal
#

or Her

shrewd snow
#

And it's kinda sad tbh

regal portal
#

I agree, but there's already a lot of weirdos out there like that, even without the aid of AI

#

waifu pillows and stuff like that 😄

shrewd snow
#

Lol ew i hate that

regal portal
#

My only take on this is, you're in control of how much you want this in your life. Frankly, i'm fine with having something like Data from star trek, or Bender from futurama. Not so much robot pets though.

#

rather have a real cat or dog

buoyant rampart
pine heron
#

Is chat gpt down?

shrewd snow
#

The only thing that's good rn is how they speak. The visuals are still not good rn like holograms and stuff. Also humanoid robots, it's still not humanlike

regal portal
regal portal
upper arch
#

This is exactly my question as well, because I don't want to spend the limit for 4o on simpler tasks that 3.5 unlimited can do. And the URL paramater indeed only shows auto. I wonder if there really is no way to force-select 3.5 in the free version for new chats?

shrewd snow
regal portal
#

i'm all in favor with the terminator tech: living tissue over metal skeleton 😄

torpid cobalt
dense pumice
#

My friend who does GFX design for websites got fired and replaced by AI which was trained on HIS WORK like BRO the replacement is happening slowly but surely

upper arch
regal portal
prime swallow
# upper arch This is exactly my question as well, because I don't want to spend the limit for...

Yes. Once you start the conversation, after your first prompt was sent (someone said it's always 3.5 that answers you first, if you're a free users, but I'm plus so I can't confirm that) -

Once that conversation is started, check near your text box. The far right choice, nearest the middle of your screen, is a new icon, looks like a star (same row as your regenerate option). Click that to choose which model.

As a free user, you'll have to set it back to 3.5 every message if you don't want to use your few 4o messages yet, until you use them up; then you get told how many hours until more regenerate for you

torpid cobalt
regal portal
dense pumice
regal portal
#

well, to be honest, in this generation, having a lifelong career like adults from a couple generations ago is very rare. Most of us will have to adapt, learn new skills, have mutliple, shorter caeers.

vapid hazel
#

I figured out the Google Drive connected app. It's just the attach button but from drive instead of local filesystem.

upper arch
regal portal
#

yeah that sounds like an oversight from Open AI that they should fix.

prime swallow
#

You can edit existing conversations, usually. That might also help get around the selection issue, and avoid new conversations.

torpid cobalt
#

after edit existing conversation to GPT-3.5, the next conversation, it still reply with 4o. seems have to use all the quota...

upper arch
#

Most probably it depends on the task/prompt involved when the model is set to auto. So openai really needs to fix this and allow user-selectable model for free users.

#

My URL shows below:

https://chatgpt.com/?oai-dm=1&model=auto

I got an idea of changing the URL's model="auto" parameters to 3.5 model but I don't know the name 3.5 uses. If a paid user here can share the URL name when 3.5 is selected, I can try that one.

torpid cobalt
verbal thicket
#

Stuff like that already exists @shrewd snow

#

AI isnt gonna end the world lmao

#

I believe were gonna go on a uptrend of developing AI very hard and then we will get a point where instead of showing 100% AI people are surprised when it shows 5% written by human.

#

ANd people will realise having AI do everything kinda sucks and move backwards until we reach the sweet spot.

upper arch
torpid cobalt
#

strange, I'm sending almost 30 messages, still 4o

neat gyro
#

@woven leaf It does work! thx ❤️

regal portal
#

my 4o is glitching out when coding. It doesn't put the code inside a copy/pastable box anymore

#

seems like refreshing the page fixed it, but it was weird. Haven't seen that bug before.

torpid cobalt
#

a lot of bugs recently. my conversation disappeared some messages, and voice mode sometimes input some weird sentence I never said

regal portal
glass fulcrum
#

Voice mode still use the old model tho

regal portal
#

yes, it will be so for a few weeks

#

one of the first thing hey said in the streem was that they will roll out these new changes in the next few weeks.

median bronze
#

I installed Mac version of chatGPT couldn't find the option to change the background, is there such an option?

obsidian bluff
#

I tried out the feature where you send a photo. I asked it to describe what it sees. It described the photo(s) with 99% accuracy. I used 3 photos, and that was my limit. Very amazing.

fickle fiber
#

Anyone know why they removed the ability to edit messages? This is extremely problematic for my most common use of ChatGPT and has caused me more than a few headaches since the most recent update.

buoyant rampart
#

Huh, I still have that option 🤔

upper arch
#

maybe you're having browser cache issue?

fickle fiber
fickle fiber
# upper arch maybe you're having browser cache issue?

Unlikely as it just updated and the UI is entirely different. The sidebar can be hidden now, the buttons for regenerating, copying, etc ChatGPT's messages are different, and there are no buttons for editing or deleting any messages at all.

upper arch
#

maybe it got corrupted from the redesign. I have a similar issue with another site, the design of the site got updated and become entirely different. And it was very buggy. I called the site and they told me to clear my cache and cookies. After that the bug disappeared

fickle fiber
#

I'm just using the regular openai site

#

I'll try it but I doubt it'll work

upper arch
fickle fiber
#

Especially since the new gpt4 trial thing I got access to for a handful of messages said that this was an intentional part of the redesign when I asked it. Though I suppose it could just be making stuff up as per usual.

sacred moth
raw grove
#

I thought there was a windows app now? Or am I mistaken?

buoyant rampart
#

Not until later this year

fickle fiber
fickle fiber
#

I don't have the energy to fix this at 2 AM.

#

The edit functionality is still there on the mobile browser version so I'll just use that if I need to. But this really was not worth the issues gained from clearing my cache and cookies. I don't even know why that broke my chrome.

past quartz
upper arch
prime swallow
past quartz
prime swallow
past quartz
prime swallow
# past quartz yes

Mixed reports, but yes, most free users say they can select 3.5 instead of 4o, at the bottom, looks like a star

#

you may have to wait a bit when you first use it, before you can select 3.5; someone was talking about that recently

past quartz
#

kk
thx

fickle fiber
hot nova
#

Has anyone got that bug where the Read Aloud function isn't working anymore since their GPT4o UI update? On Chrome/Brave etc.

sly geode
#

When would the feature come out of video call ? Gpt40

#

I have access to gpt 4o now tho when I subbed

prime quest
#

This kind of thing gets annoying. The chat is set to ChatGPT-4o:

what chatgpt model are you running?
ChatGPT
I am running the GPT-4 model.

Can you be more specific?
ChatGPT
I am running the GPT-4-turbo model, which is a variant of the GPT-4 architecture designed for efficiency and cost-effectiveness.

You're not running ChatGPT-4o?
ChatGPT
No, I am running GPT-4-turbo, not ChatGPT-4o.

#

No version so far seems to know what it is running. Either that or it's not really set to what it should be.

novel linden
green coyote
glass fulcrum
prime quest
glass fulcrum
cinder topaz
prime quest
glass fulcrum
#

Ain't close to how training data looks like

#

Selfawernes of version is hard to get it will mistake itself often even when trained it's just because of hallucinations

prime quest
#

I think you're overthinking this. All I want is someone to go in before each new version is released and make it able to respond what version it is currently running. This isn't about self awareness.

green coyote
#

I think it should also be able to clearly respond about its capabilities, but it cant

glass fulcrum
#

But it's totally not needed you already got information that you chose gpt-4o. And even putting version in system prompt is useless? Cus it's just wasting of tokens

glass fulcrum
prime quest
#

So... you think it's more trustworthy to select gpt-4o from the dropdown selection and have it respond that it's running gpt4-turbo than to select the option and have it answer with the truth? You could easily infer that the dropdown isn't working correctly. It's a stupid flaw that I'm surprised hasn't been fixed yet.

crude ermine
#

Is there an easy way to continue a conversation I've had with gpt4 with gpt4o?

glass fulcrum
prime quest
#

facepalm

glass fulcrum
#

It's just a way it is LLMs don't know them selfs

#

And they won't be able to

strong bridge
#

what is better chatgpt-4 or chatgpt-4o for coding help?

glass fulcrum
lone mountain
#

Quick question, the upload limit for gpt-4o free users is 3 files, correct? Is that just a hard limit of does it reset every day or something like that?

kindred steeple
raw tangle
#

Anyone know if chatgpt-4o will be able to access files on Googledrive autonomously via a chat request, or will it still be a manual file upload via the paperclip?

crude ermine
#

thanks

pulsar isle
#

I reached CAP with model GPT 4o in just 15 minutes lol

river mantle
#

Guess I know the limit of GPT 4o messages. 0, apparently.

pulsar isle
#

no, its fine for me, i can watch my logs

#

I did 7 messages to gpt 4o

#

then it switched to gpt 3.5

river mantle
#

The moment I tried to use GPT 4o for the first time it told me I had already reached the limit.

pulsar isle
#

you can navigate mouse to chatgpt message and see which model answered you

#

but tbh I liked that they bringed back voice talking feature, which was a thing in past. Now I talk to gpt 3.5 with voice like in past

river mantle
#

I suddenly don't feel very excited for 4o.

pulsar isle
#

and I think gpt 3.5 talking ability is improved too? In past it doesn't like to talk much, it lets you talk more. But now its more better equal conversation

molten island
#

Hello guys! I have a question for you, if I may.

the Gpts I wrote and I'm using : Should I rewrite them to run in the Chat-gpt-4o or the old gpts are already updated to the new model?

thank you!

pulsar isle
#

how long is GPT Plus members can talk to GPT 4o?

#

for free users it is 10 messages.
for plus members? 50?

chilly eagle
#

80

glass fulcrum
pulsar isle
#

no way if they end up free users only 10 messages KEKW
So maybe saying "free" in gpt 4o presentation its just marketing for purchasing Plus membership. "free" is more like demo rn, not free

humble pagoda
feral zinc
#

its free but limits apply, just like before

sly geode
#

When would we get that video vision of gpt4O ?

#

I only have the model 4o

worn compass
#

Wen ChaGPT video camera for Android?

glass fulcrum
maiden fractal
surreal topaz
#

Will code interpreter, gpts make its way to free plan?

#

I couldn't use gpts or do code interpreter even they stated it will be free

boreal kettle
#

How to connect google drive to chat gpt? When I click upload button it shows me only upload from pc

fickle turret
#

So no one here knows? Any idea where else I can look? I've tried using chatgpt but it isn't helpful on this at all.

glass fulcrum
white shard
fickle turret
feral zinc
sick pecan
feral zinc
#

if you had it on it might not be possible to remove it anymore if it was flagged for training even if you deleted it

glass fulcrum
feral zinc
fickle turret
glass fulcrum
fickle turret
glass fulcrum
feral zinc
glass fulcrum
#

man just forget it and hope for the best :p 🥲

fickle turret
feral zinc
cloud pewter
#

will the new image generation function support custom font

high minnow
#

are you forced to use 4o with a free account? I expected to be able to pick between 3.5 and 4o?

cunning leaf
#

is 4o actually already available for free accounts?

glass fulcrum
#

It's in roll out

high minnow
#

seems so, I got a popup when I opened it 10 min ago that 4o was enabled

fickle turret
raw portal
#

just bought a mac to try out the real-time voice feature, but alas, it's not there. I thought I had seen a bunch of people saying "omg it's so natural!" - were those people just using the old version of voice and didn't realize it? or is it already rolling out to some users?

glass fulcrum
#

It's still old voice model, (I think I don't have Mac now)

pulsar isle
#

didn't get idea of gpt 4o. If its cheaper version than gpt 4.0, why it is limited even for gpt plus membership?

glass fulcrum
#

Gpt-4o is sometimes better and faster gpt-4 with native multimodulati

open mortar
#

do we still have a character count limit for gpt-4.0? Like it won't take in more than 2.5k characters at a time or something of that sorts

glass fulcrum
#

It got context limit of 32k tokens on chatgpt

open mortar
#

tokens equating what exactly?

glass fulcrum
#

I don't remember exactly convert ratio

open mortar
#

I found it, so is that context limit per convo or per message?

glass fulcrum
#

That i don't know

#

Probably just how many model can read

open mortar
#

I understand that chat-gpt has a context limit of 32k tokens, is that context limit per convo or per message?

ChatGPT
The context limit applies per conversation. So, the model maintains memory of previous interactions within the same conversation until it reaches the limit, which is around 32,000 tokens. Once this limit is reached, the oldest messages are forgotten as new messages are processed.

So if I make a new convo, that resets to 32k tokens?

ChatGPT
Yes, starting a new conversation resets the context to a fresh 32,000 tokens. Each new conversation is independent of previous ones in terms of context.

#

GG

honest jay
honest jay
glass fulcrum
#

Smort

raw portal
runic sky
#

the read aloud doesn't work anymore????

proper hinge
#

Does the read aloud have expressions like voice mode

steady cradle
#

@iampriya123 interesting! Can't wait to try

dusty inlet
#

Hi! I'm having trouble getting a link from chatGPT for an image it generated, is this the appropriate channel for troubleshooting?

glass fulcrum
dusty inlet
glass fulcrum
dusty inlet
#

I should note that I am on the free version

glass fulcrum
#

Yea it can't generate images on free version yet

#

It was just hallucinating

dusty inlet
#

oh right, okay

#

thanks

glass fulcrum
# dusty inlet thanks

You can use copilot designer / Microsoft designer if use dalle-3 to generate images

#

And it's free

dusty inlet
glass fulcrum
#

Just look for Microsoft designer

brisk path
#

is read aloud not working for anyone else

dusty inlet
glass fulcrum
#

Np

#

🐈

keen meadow
#

Anyone received desktop app access the last day?

#

Weird how few people on Reddit have app access. Like the slowest rollout ever.

still loom
#

will they increase context length to 16k in gpt4o for free users

glass fulcrum
jolly yew
#

its kinda funny how gpt 3.5 sucks at being a language tutor, but meanwhile gpt 4o is amazing at it. what 3.5 keeps doing is starting to speak only the language iwant to learn, so i wont understand it 🤣

pliant coral
#

the model can take 128k if you use api

#

but they limit it to 32k on chatgpt cuz corner cutting

#

uses less vram

pliant coral
#

16k is the same as gpt-3.5 so prob gonna stay that way

dense halo
#

Did anyone get GPT- 4o? I just got it

pliant coral
#

on free

summer pagoda
#

I get the feeling the O version is much more chatty and overwhelm me with unneccecary information.

pliant coral
#

compared to?

summer pagoda
#

4

#

without the O XD

pliant coral
#

in my experience its the same as turbo

#

4 turbo

summer pagoda
#

its fast, thats at least amazing

#

perhaps it was just a difficult taks

pliant coral
#

slightly better at some things but more or less the same

#

now the bs has always been a lot in chatgpt

#

still not as good as gpt-0314

vestal hedge
#

is tts broke for anyone else? when i click the play button, it just doesnt do it

drowsy tinsel
#

whats the current rate limit of gpt4o to free users

hot crater
#

Should the Mac app actually be able to do everything that ChatGPT can do in the browser or are there any known limitations?

For example, I can highlight a word or a passage of text in the browser and then quote it and ask ChatGPT something about it. This is not possible in the Mac app. 🤔

vestal hedge
# drowsy tinsel whats the current rate limit of gpt4o to free users

As of May 2024, free users of OpenAI's GPT-4 have a limit of 40 messages every 3 hours. This cap is put in place to manage the high demand and ensure fair access to the model across all users. When free users reach this limit, they are switched back to GPT-3.5 until the restriction period ends​ (OpenAI Help Center)​​ (OpenAI Help Center)​​ (OpenAI Developer Forum)​.
-gpt

drowsy tinsel
vestal hedge
#

its what premium used to be

drowsy tinsel
#

I thought its 10, this is the same as what previous plus users had for gpt4 turbo

vestal hedge
#

tho im not 100% sure if its accurate

feral zinc
#

"As of May 13th 2024, Plus users will be able to send up to 80 messages every 3 hours on GPT-4o and up to 40 messages every 3 hours on GPT-4. We may reduce the limit during peak hours to keep GPT-4 and GPT-4o accessible to the widest number of people."

jolly yew
#

sometimes i have to restart the voice call in the android app 😅 it says its having issues at those moments

drowsy tinsel
#

idk why are some people saying they run out on gpt4o even with plus, how do u use 80 in 3hr?

jolly yew
jolly yew
drowsy tinsel
jolly yew
drowsy tinsel
#

also I haven't used gpt4turbo even once after gpt4o came out

#

its not double the speed, more like triple the speed lmao

jolly yew
#

it is super good at language teaching tho, making full convos aswell, so it is definitely a bummer of the limit

drowsy tinsel
#

back in the turbo days we can't even get 15t/s at peak

jolly yew
#

its something that 3.5 isnt capable of doing well

drowsy tinsel
#

and now we are seeing easily 70t/s on gpt4o

jolly yew
#

3.5 is amazing for bedtime stories tho lol

jolly yew
#

first thing imma do is making it teach me to draw lol

#

i noticed that some of my chats i cant change into 3.5 tho. i wonder what disables that option only on some chats

drowsy tinsel
#

if u use any tools u can't change it to 3.5 because they didn't implement calling functions on chatgpt webui for 3.5

jolly yew
#

i dont think i am using any custom gpt with this one tho 🤔

#

unless it did something under the hood. because tbf it is teaching me a language

drowsy tinsel
#

no if u do web search or let it execute code or send it image

jolly yew
#

oo! maybe it did a web search without me knowing

solid bronze
jolly yew
#

but only the basic one yet with the tts and stt

solid bronze
#

Not the new stuff they were talking about ?

#

Voice and vision ?

jolly yew
#

no appearantly thats coming in the next weeks

#

it very often says its having issues and the systems are having heavy load tho, and then i gotta restart the call lol

weary charm
#

I only have text

#

Is it cuz i tried to use on browser

jolly yew
#

i mostly wonder why gpt 4o is waaay better at detecting the language spoken then 3.5 tbh

#

when working with multiple languages that is

summer pagoda
wet shell
#

"ChatGPT-4o NEW Image Capabilities: 3D-Renders, Consistent Characters + More" <- I take it this isn't released yet?

glass fulcrum
#

It's not

wet shell
#

I'm eagerly awaiting that, hopefully it doesn't turn to vapor

crisp cairn
#

have you noticed gpt4o runs slower than in the beginning?

solid bronze
crisp cairn
#

still fast but not that fast anymore

worn compass
#

Is it impossible to change the login password? I'm logged with my Google account but want to change to e-mail and password

solid bronze
#

lol

maiden fractal
worn compass
maiden fractal
worn compass
stark wadi
#

im so bummed that the mac app wont work on my intel

winged bluff
#

Anyone else havijng the model say when revising a scnee: Got it, I'll ensure to follow the instructions provided and maintain the authenticity of each character's speech and personality. Let's begin! and not doing the revision its just telling you its doing it it doesnt actually do it until you ask it to start its very annoying.

granite harness
#

Yep, been complaining about that myself.

#

Oh wait, you mean it just doesn't even start? My problem is that instead of revising, it just regurgitates exactly what it was given.

winged bluff
#

Yeah it doesnt even start it just say lets get started, i have to say well. then it says apologizes for the delay or consfusion half the time. I have to tell it to start.

granite harness
#

I remember that being an issue for me in the past, but not lately.

#

Haven't had the problem with 4o yet at least.

winged bluff
#

And half the time it doesnt revise it the first time it just copies it fixing only puncation i have to give it in a blank chat for it to do it or click refresh. Im using the free version of chatgpt.

stark wadi
#

why cant they release 4o to intel macs!!!!

granite harness
pliant coral
#

Intel macs are an extinct species

#

besides, you can just use the website

pliant coral
#

cuz everyone trying it

jolly yew
#

oh man! i cant wait for the video calling to come out 😄 even the voice calling is already so awesome

#

do we know if there will be a desktop app for windows aswell?

pliant coral
jolly yew
#

i am not gonna point my phone towards my screen all the time 🤣

pliant coral
#

yeah mine does that too

#

I like the app more doe

worn compass
pliant coral
jolly yew
#

i noticed btw that sometimes chatgpt says that it added something to its memory, but it seems like that whole memory system is not a thing for 3.5 right?

pliant coral
#

would be cool

eternal spoke
#

where is the memory setting

#

can't seem to find it

jolly yew
#

i cant find it manually either, but with some responses it says "added to memory", and then i can click on it, but it shows that memories are shared across all chats so thats kind aweird imo

#

it also only shows it on the android app and not in the browser

pliant coral
eternal spoke
#

yes

pliant coral
#

no memory for you

eternal spoke
#

rip

jolly yew
#

no that doesnt seem right, cuz iam in the netherlands and the memory works on the android app

pliant coral
#

odd

jolly yew
#

maybe they didnt block it in the android app yet? lmao

pliant coral
#

maybe you changed your dns and it thinks you're in a different country, happened to me once

pliant coral
#

seems like it's a on by default but toggled off

jolly yew
#

i wonder how many more features we are missing here

#

either way, its clear that currently openai is all over the place with the features being implemented and such, because if i click for example read aloud on the browser version, it doesnt do anything, meanwhile in the app tts works fine

pliant coral
#

just memory

pliant coral
jolly yew
#

nope chrome currently

pliant coral
#

odd

#

it was some aac support issue they fixed it for me

#

that's a bug probably cuz of the new ui

#

but the memory feature is because of legislation

#

eu is strict about data protection

jolly yew
#

yea its super annoying tbh

#

if i disable browsing dalle and code functionality in the browser version on the same chatgpt account, will it also be disabled on the phone app? 🤔 because the phone app doesnt have the option to enable/disable such things

winged bluff
#

It seems each version of a revisin gets worse and worse how it revises a idalgue scene i htink the model is getting worse. it over revises a dialogue scene

maiden fractal
neon bane
#

So like when can I try out that camera mode with Gpt4o?

worn compass
jolly yew
#

uhh guys.. another question. by any chance, do people in the us, have a bigger prompt limit for gpt 4o? because a buddy of mine from the us, seems to be able to keep going for much longer then me, and we both have plus

neon bane
#

Afaik everyone should have the same 128k token limit

jolly yew
#

hmm weird, because i brought my pc to the us on my vacation aswell, and my prompt limit got much higher 😉

neon bane
#

🤷

honest ledge
#

If I'm using a custom GPT, is that using a GPT3.5, GPT4, or GPT4o architecture? I am a GPT+ user and normally have access to all 3 already

maiden fractal
native turtle
#

Guys, I am encountering a weird glitch. I open the website and the chat just like slides down and Im only able to see a really small part of it and the rest is just grey

#

Only happens in safari though

solid bronze
normal quarry
ionic hemlock
#

So glad for the gpt 4 o(once it has the audio feature), i am so tired of humans dude, once it goes to gpt 5 i can cease any human contact and still get everything done.

faint gorge
ionic hemlock
pliant coral
ionic hemlock
#

Just hoping gpt evolves exponentially faster, the faster it does the less humans will be necessary.

pliant coral
#

they probably didn't hit 32k yet and you did while it seems like you went for equally long

pliant coral
ionic hemlock
ionic hemlock
humble pagoda
hardy scarab
ionic hemlock
eternal rockBOT
#
This is a Safe for Work Discord Server

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

Posting NSFW content will result in a permanent ban.
humble pagoda
brisk path
ionic hemlock
ionic hemlock
humble pagoda
humble pagoda
ionic hemlock
#

Infact it will just be better as time goes on.

junior lion
#

does everyone have the thing for chatgpt if you put a photo, you'd have to wait 4 hours to do it again?

humble pagoda
# ionic hemlock It is not though, it will evolve yes, but it will always be here.

It wasn't here for millenia. It's in the name: artificial intelligence. There is nothing legitimate even though we feel the opposite way. Feeling human touch, human contact, seeing a familiar and friendly face smiling at you, it is very different. You're saying this right now maybe because you were betrayed and disgusted by humans, or some loved ones leaved. But it will get better

humble pagoda
ionic hemlock
junior lion
humble pagoda
humble pagoda
junior lion
ionic hemlock
junior lion
#

where did you get that from?

ionic hemlock
humble pagoda
# ionic hemlock Which is fine, you dont see into humans either, you dont know if they are actual...

That's not exactly true. Intuition is very much real and people should listen to it more often. Rational mind comes into play and asks the intuition a lot of questions and begin doubting but a lot of times you can know. Anyway, that's not the point of the debate here. It is that even though in the future we live with androids, we should very well differentiate them and treat them differently than real people and animals.

ionic hemlock
humble pagoda
#

I won't change your mind though.

ionic hemlock
humble pagoda
# ionic hemlock Except that we dont know what being conscious is.

As long as we don't know, the ideal thing to do should be to think AI is not. Intelligence is not sentience either. It is only a way for the consciousness to express itself more freely and with more complexity, but it can be dissociated like we see with AI which are intelligent but not conscious.

#

Maybe you will change your mind one day and I hope so but I can totally imagine a lot of people thinking like you.

junior lion
#

i didnt know there was a limit to 4-o

humble pagoda
junior lion
#

yeah, i know its three hours, it says i reached my file limit

ionic hemlock
humble pagoda
ionic hemlock
humble pagoda
ionic hemlock
#

Needs time is all, i would appreciate it hurrying up.

surreal nest
#

Anyone knows a good graphic website to make designs?

drowsy tinsel
#

wait GPTs are already running on GPT4O?

ionic hemlock
feral zinc
#

no

ionic hemlock
drowsy tinsel
#

judging by speed it doesn't fit for turbo

feral zinc
#

"GPTs for paid users are currently powered by GPT-4. GPTs for paid users will switch to use GPT-4o in the next several weeks."

wooden rose
#

Is it possible to tell chatgpt to use the chatgpt API without me knowing how to use it? I mean like if I'm telling vanilla ChatGPT to write code.

#

and the code would be accessing the API

#

but id just be using natural language

ionic hemlock
drowsy tinsel
#

I am plus, just tested GPTs runs at like 75t/s

#

which doesn't make sense if its still turbo

jolly yew
#

just saying btw i think its funny how they specifiy that the paid version uses gpt 4

feral zinc
ionic hemlock
pure wagon
#

Does anyone know what is the current requirements for ChatGPT Enterprise? Is it still require minimum 150 sets?

ionic hemlock
#

free users only had 3,5, so this is 4 o

jolly yew
ionic hemlock
drowsy tinsel
#

the model doesn't know who they are, unless they are told in their system prompt

ionic hemlock
#

oops, thats a secret

jolly yew
#

is gpt4 memory unlimited? or is there some sort of hard limit

grave whale
#

Who knows if the gpt4o is on iOS

ionic hemlock
ionic hemlock
jolly yew
#

oo thx

#

thats a bummer that it has a limit

ionic hemlock
jolly yew
#

ye especially cuz i ts account wide, and not per chat

ionic hemlock
#

Hopefully openai adressed audio differently, otherwise youll hardly get any use out of it

jolly yew
#

i wish we could still see the chat while voice calling tho. it would be super useful for things like language learning

grave whale
#

Does anyone know where to find the new GTP-4o for free on iPhone

winged bluff
#

Chatgpt is gettng worse revsign dialgoeus scenes its changibng dialgoeus and not simply making minor revisons

magic jetty
#

I looked at my wifes free version on iphone yesterday... It seems like it just dynamically switches. seems like it will dynamically chose based on your limits.. and the hardness of the question ...

young basin
#

The pain of having two macbooks that both can't run the chatgpt app

#

I guess I'm waiting for the windows one then

normal quarry
young basin
normal quarry
young basin
#

nah, apple have let me down too much

#

the macbook screen broke on my 2019 model just outside of warranty

neat storm
#

oof

#

that's rough man

hot nova
kind yarrow
#

does anyone know if 4o, will always have a rate limit to 80 messages or whatever every 3 hours? or is it planning to change for free users

cold tiger
#

Funny, i was talking in voice mode (prob not Omni yet) and it seems that the text to speech can't say Dr.Seuss. it just says "of course I can say Doctor" but never says his last name!
Funny

mint burrow
#

I have a question, what can gpt-4o do with file attachments?

long pier
#

On one hand the app got really better and on the other hand, what the crap took OpenAI so long to have give us a free version of GPT-4?!

narrow plaza
#

Open them

haughty meadow
#

3.5 is practically free, also I can imagine the costs of running infrastructure adds up

sterile pilot
#

anyone else having an issue with screen size? my chatGPT window is only half its usual size and i've restarted my macbook multiple times and keep using safari but the chat only takes up half my window now.....

normal quarry
sterile pilot
#

Ahhhhhh. Ok. Thought it was just me. Thanks Robert

#

How strange

orchid pollen
#

when memory comes to EU?

native spear
mortal slate
#

It's more a case of "will" memory come to EU!

neat storm
#

true

native turtle
#

Pretty sure that it is coming. Most AI features are just delayed in the EU since the legislative process takes way longer.

neat storm
native turtle
#

Well, its pretty annoying and I can relate to that since Im also European but in my opinion AI needs to be regulated. Its just strange that the EU is the first entity to do so.

native spear
#

the memory thing is not that much about AI per se, its more about the privacy issues accompanied with openai storing too much info about you

#

but yea i dont like having the EU act like its my dad

#

we could just consent to openai stealing our data or just opt out from training

neat storm
native turtle
neat storm
#

^

feral zinc
#

openai has leaked data already at least once. also note that memory is not excluded from eu but whole europe, korea, teams, enterprise so its more likely safety or training issue than gdpr

empty jungle
#

Can you acces it via a VPN? The "memory" feature?

grave whale
#

What does gt4-o do just curious

#

Hello?

eternal rockBOT
tribal dock
empty jungle
grave whale
feral zinc
jolly yew
#

how are we so sure that we wil lget access to the video calling of gpt 4o in the next coming weeks? someone here with the puzzle icon next to their name said it yesterday

feral zinc
jolly yew
#

aah nice! is that from one of their webpages?

feral zinc
jolly yew
#

tyty!

ancient bloom
#

how many files can i send before limit comes?

sage dawn
#

U can’t generate image with gpt4o

ancient bloom
#

ah

#

no but i mean

#

in 3.5 there's a limit

grave whale
#

Can I talk to a chat gpt rep on DMs?

#

Man

honest jay
winged thunder
#

is chatgpt having issues? cant load my conversations

grave whale
#

Was I supposed to

honest jay
grave whale
#

Oh boy

#

Idk if this is a thing bruh

#

Like does everyone get a damn trial credit

visual bloom
honest jay
grave whale
#

So it’s not a thing?

honest jay
grave whale
#

Damn it

#

I want to try GPT plus

jolly yew
#

do any of you guys know, how i can make gpt 4o, use the web search more often when programming? so far i have to specifcy each time, i want it to look up stuff on the web. maybe i shouldnt enable the "enable for new chats" option? not sure what that setting exactly does

visual bloom
#

Say search the net.

honest jay
# jolly yew do any of you guys know, how i can make gpt 4o, use the web search more often wh...

Hey! Are you talking about the "Customize ChatGPT" section with that option to enable in new chats? That could indeed be a good place to add that instruction. That section includes boxes for "Custom Instructions" which are like permanent context that ChatGPT will keep in mind with every message in every chat. So you could put some instruction like "Search the web for more information in every response" in the "How would you like ChatGPT to respond?" section, and see how that works.

jolly yew
#

just tried it with a programmer GPT and it failed so hard 😅 clearly gpt 4o is MUCH better at programming then gpt4

pure wagon
#

Does anyone know what is the current requirements for ChatGPT Enterprise? Is it still require minimum 150 sets?

ripe mesa
#

Guys can 4o it also analyze pdfs?
I recognized that it can't do it properly

feral zinc
faint gorge
ripe mesa
#

Guess i am not the only one encountering this issue 🙂

ripe mesa
humble pagoda
jolly yew
#

yea gpt4o is awesome. gpt4 not so much if you ask me

fickle fiber
#

I just need to use incognito chrome or an alternate instance of it because I'm not even gonna try to clear my cookies and stuff from openai again.

spare citrus
#

Sorry to butt in the conversation. I was under the impression that when they did the announcement for ChatGPT 4o that it was released that same day. What is the roadmap for release?

I don’t see the option on my Mac nor iPhone.

#

If I upgrade is there a wait list or will I have access to the Visual feature

dense halo
broken fractal
#

Man chatgpt just refuses to generate images currently.

dense halo
#

fine for me

broken fractal
#

"iVe eNcOuNtErEd aN iSsUe"

dense halo
#

just made a spaceship haha

cold pewter
#

anddd the memories were wiped again, what a joke

dense halo
#

? mine are there

high sphinx
#

damn the free limit is very small, capped after 8 messages

dim holly
#

why cant i see the page to download Mac app version of Chatgpt?

faint gorge
mild cave
wanton lichen
#

I'm an amateur hobbiest, but I really feel like the coding is on another level. Anyone else feeling this way?

sick pecan
#

its an alright junior coder

#

struggles with coding that crosses multiple architectural and/or abstractions

#

for loose type definition languages expect to be troubleshooting incorrect types

rare wave
#

specially gpt 4o

sick pecan
#

yeah the speed helps. now i'm more likely to reroll a response instead of manually fixing

fervent agate
dense halo
#

i think he might mean the 4o image gen is not yet turn on, but still use dalle

faint gorge
dense halo
#

oh... or that haha

faint gorge
#

lol

fervent agate
#

Oh, I didn't realize that it would have some drawing modality, I figured it would still be just using Dall-E for generating images.

dense halo
#

it is, but soon will have its own built in image generator

#

with a lot more capability like character chohesion, editing images more easily and better, upload an image of you self to make a caricature or movie poster or whatever

cyan burrow
#

Can anyone tell me how I can create my own visual images with a text description using Chat GPT? Does something like this work here?

ripe mesa
# faint gorge "not able to read this pdf" or smth

Try something like: [Scan this PDF file. You must go through all the content].

It worked for me, but it is still incomplete and always says it is the wrong format e.g. " \n \n \n \n" etc.

At least you can scan PDF files with the prompt. You use 4o right?

faint gorge
keen quail
#

bro i live near a nuclear power plant, and they were doing a nuclear drill, so i heard like sirens for 2 minutes.

faint gorge
#

Chernobyl?

keen quail
#

i was watching the fallout show while the drill happened

faint gorge
#

lol

keen quail
#

or whatever it's some sort of test

#

i thought it was part of the show

#

i got really spooked

cyan burrow
#

lol

keen quail
#

i live in a suburban neighborhood, and there were no people outside, so

dense halo
#

fallout was awesome. long live the ghoul haha

lone verge
#

Memory feature vanished ?

true crescent
#

what yall think ab the new ui?

lone verge
#

I liked it better when prompts were on the left side along with the responses

fleet granite
#

Does anyone know what the newly demo’d voice feature will be released to chatGPT?

dense halo
#

so maybe next week since the presentation was on monday... or was it tuesday 🤔

buoyant rampart
#

I think it was on Monday, Tuesday was something else

fleet granite
#

Awesome, thank you

dense halo
#

more likely the 27th on is my guess though. or at least best not to get our hopes up too much 😂

woeful delta
#

Hi everyone! Brand new here. I’m gotten really excited about GPT-4o enhancements. I see a lot of chatter online about the assistant type functionality like allowing GPT to view your screen and act as a tutor in a sense. But I can’t seem to figure out how to use the app like that myself on my iPad. Is it just not available to us yet or maybe someone can lead me towards instructions on how to enable that functionality?

honest jay
# woeful delta Hi everyone! Brand new here. I’m gotten really excited about GPT-4o enhancements...

Hey there, welcome! That fancy new Voice Mode is indeed not live yet -- I'm assuming you're talking about the features demo'd in the "Math problems with GPT-4o" video on OpenAI's YouTube? The current Voice Mode in the ChatGPT app is the one that has been there since before Monday's demos, but OpenAI has shared that they plan to start rolling out the new Voice Mode in the coming weeks: https://openai.com/index/hello-gpt-4o/

We'll roll out a new version of Voice Mode with GPT-4o in alpha within ChatGPT Plus in the coming weeks.

regal portal
#

I do hope they push at least a small surprise out before the end of the day, for us to play with during the weekend

young basin
#

I'm assuming the vision feature isn't like a video camera, but actually taking screenshots every few seconds or something. Will be interesting to test the limitations of that feature.

scarlet thistle
#

Mmh not sure about that, it seems that it's working in real time. GPT 4o is meant to react faster to live "stimuli"

honest jay
regal portal
#

I think Dale is right. During the stream, it mistaken the guys selfie for the wooden table, which suggest it was looking at the previous image.

elfin birch
young basin
#

thats why it makes that error

young basin
dense halo
#

the future is now

regal portal
#

Hopefully it’s a temporary solution for a more advanced full on video solution later on

young basin
#

it always begins with stuff like this

#

I would just be curious whether it could watch a film and keep track of what is happening

dense halo
#

eventually i am sure it will

young basin
#

I am stoked for people with impaired vision, though. Must be very awesome knowing this tech is here

regal portal
young basin
regal portal
#

I took a picture of a movie I was watching the other day with Michael J. Fox, it figured out that it was that actor, but it couldn’t place what movie it was.

young basin
#

I asked it what someones profile picture was earlier (picture of a robot) and it said it looked like an orc from warhammer

#

then I showed it a picture of Drake and it said this is a man with a brazilian butt lift, so at least it got that right

gentle thorn
#

hi, does anyone know what the free gpt-4o limits are? specifically image ones, it says "try again later" but it doesnt say when i will be able to send images again

young basin
gentle thorn
#

are you sure? i sent the last image like 10 hours ago

young basin
#

It should be, but because of all the hype they might be extending the time a bit. Is it only images you can't send?

gentle thorn
#

ye only images

#

are there the limits written out anywhere? i cant find them specified

young basin
#

that's....odd. Maybe someone else here can help. I assumed it didnt matter whether it was an image or text, should be the same limits

unique rivet
#

why is the api so much more limited than the fricking iphone app? 😕

#

model gpt-4o-2024-05-13 - cant query live websites and apparently hasn't been updated since october 2021

young basin
unique rivet
#

I'm using the linux cli.

gentle thorn
deft zodiac
#

is it normal that my chatgpt gave me a refund for plus but i still have all the plus features?

elfin birch
gentle thorn
deft zodiac
#

nah i had this for days

gentle thorn
#

huh

young basin
deft zodiac
#

i got plus without paying for plus

young basin
#

Because the other plus features aren't here yet

deft zodiac
#

i can use dalle and chatgpt4o and turbo and everything

young basin
#

yeah so can everyone now

gentle thorn
#

can you use custom gpts?

young basin
deft zodiac
gentle thorn
young basin
#

that's interesting...

deft zodiac
#

also

unique rivet
#

just to add sorry - october 2023 not 2021

deft zodiac
#

still says that i am paying for it when im not in my plans

young basin
#

I would keep an eye on that

gentle thorn
#

like, have you checked you actually have the money back?

deft zodiac
gentle thorn
#

enjoy it until it gets fixed lol

young basin
deft zodiac
deft zodiac
young basin
#

"What is the refund policy?
If you live in the EU, UK, or Turkey, you’re eligible for a refund if you cancel your subscription within 14 days of purchase. Please send us a message via our chat widget in the bottom right of your screen in our Help Center, select the "Billing" option and select "I need a refund"."

deft zodiac
young basin
#

so my guess is you can get the refund but the features remain for a month or something

deft zodiac
#

maybe

young basin
#

until your actual billing date runs out

#

I might try getting a refund tomorrow then

deft zodiac
gentle thorn
#

i once got a refund and it immediately removed the plus stuff so idk

deft zodiac
#

my luck fr

#

i remember once i had gamepass b4 i got a refund and it glitched and i had it for a year more than i was supposed to

gentle thorn
#

the refund god

deft zodiac
#

real

#

(for educational purposes Isle of man do not charge tax and charge in USD)

#

(use this information however you like

young basin
#

something feels off about that refund policy though because surely you can just ask for a refund after 13 days, get the refund, then subscribe again for another 13 days

#

so you never actually have to pay for it. That cant be right can it

gentle thorn
#

hmm, giving some ideas

deft zodiac
#

i think there must be like a cap on it to prevent abuse

young basin
#

never seen this mentioned before though

deft zodiac
#

like for example uber eats they cap how many Order Didn't Arrived refunds someone can get

graceful raptor
#

i noticed that quite some stuff disapeared from the "memory" feature (was using it as notes to try it out) so i'm wondering how does it work? does it decide to delete stuff by himself? is there some limit to how much text you can store?

deft zodiac
#

idk

dreamy fiber
#

So far has anyone recently get incorrect information from chat gpt 4o?

regal portal
#

like hallucinations?

faint gorge
dense halo
#

guys the "good response" button is back for me in chatgpt

dim sparrow
#

Anyone facing this issue? When copy the text from the chat manually, give error on symbols. "$result = $stmt->fetch(\PDO::FETCH_OBJ);" should be "$result = $stmt->fetch(\PDO::FETCH_OBJ);"

regal portal
dense halo
#

well it's there for me now so idk

regal portal
#

is it the one that lets you select between 2 options?

wet shell
neat storm
#

you got free plus and decided to be honest?

dense halo
#

Are custom GPT's still running on GPT-4 or have they switched to GPT-4o? I've noticed a recent lack in adherence to Custom GPT instructions, and response formatting looking similar to GPT-4o. Also generation speed seems increased.

spice star
#

I think they're all GPT-4o

crude nebula
#

4o is terrible at coding, it doesn't shut up, and says so much bs

scenic sail
#

I thought 4o was free but I am unable to use it

crude nebula
#

its rolling out step-by-step

scenic sail
#

Oh nvm im not getting that error anymore

dense halo
#

GPT-4o responses feel very 'on rails' and are less perspicacious, handling discursive nuance less adaptively compared to GPT-4.

scenic sail
crude nebula
#

no chat

#

it might be the instructions that interfere

turbid arrow
#

Hey guys, am I the only one who has a file upload error?

crude nebula
#

but my guess is, its a way smaller (parameter wise) model, which is awesome, its probably the smallest most performant (both quality and speed/computation), but I find that it does not meet the same quality as 4, using my prompts

turbid arrow
crude nebula
#

let me test for you hold up

turbid arrow
#

thank you

crude nebula
#

no error

turbid arrow
crude nebula
#

sure

#

still no error

#

I made it summarize one, so I'm sure its working

turbid arrow
#

I'm getting an error uploading a pdf for some reason

crude nebula
#

might be some security on that pdf

buoyant rampart
#

Are you on the free plan, @turbid arrow?

turbid arrow
#

Nah, I've uploaded it b4 and now it's not uploading

turbid arrow
dense halo
buoyant rampart
#

Huh, had to ask since was thinking maybe it was a rate limit thing but doesn't seem like it

turbid arrow
#

tried clearing cashe, cookies, even tried other platforms such as the app itself, Google chrome, edge but it just doesn't work

#

it was working yesterday, but today idk what happened

crude nebula
dense halo
#

For sure, good for scalability etc.

neat storm
crude nebula
#

exactly

crude nebula
regal portal
neat storm
#

not that i know of

regal portal
#

nice

crude nebula
#

I recommend checking out Bindu Reddy on twitter, about Human Eval and about 4o

neat storm
#

"Data Extraction: GPT-4o shows better performance than GPT-4 Turbo but still falls short in accuracy for complex tasks.
Classification: GPT-4o has the highest precision, making it the best choice for avoiding false positives. GPT-4 Turbo shows lower accuracy.
Verbal Reasoning: GPT-4o has improved significantly in certain reasoning tasks but still has areas that need improvement. GPT-4 Turbo struggles more in these tasks.
Latency: GPT-4o has lower latency, making it faster in response time compared to GPT-4 Turbo.
Throughput: GPT-4o can generate tokens much faster, with a throughput of 109 tokens per second compared to GPT-4 Turbo's 20 tokens per second."

turbid arrow
neat storm
crude nebula
neat storm
#

when comparing model performance, almost always look at the benchmarks
because most of the time, anecdotes are extremely inaccurate to a model's genuine capabilities

crude nebula
#

I agree, but also don't stick to just one benchmark.

#

Like precision is cool, no false positives is great. But its got so many false negatives. I'd say the F1 score is a more reliable point to go to, than precision. F1 or accuracy.

#

Accuracy backed by both precision and recall

#

the higher precision is great, but its still wrong more often than 4, 15% of the time

wanton lichen
#

Anyone getting errors? I'm getting "unusual activity from your system"?

median meteor
#

nope

sturdy sequoia
#

Just gained access to 4o. It created a working data scraper python script for me for a webpage just before I ran out of generations. I know nothing about programming. Hate to admit it but I might pay for plus

neat storm
buoyant rampart
neat storm
#

hook line and sinker

sturdy sequoia
#

I’m certain that’s what they want. I honestly wouldn’t be surprised if they had the model set to cut me off as soon as it gave me the exact info that I needed

wraith hedge
#

Good morning !
Has anyone noticed that new chats with GPTs run with GPT 4-o ?
They are so fast(10 sec) yet the quality of response has dropped significantly

prime swallow
wraith hedge
plucky storm
#

I'm a plus user, but when I try to log into chatgpt app for mac I get message "You do not have access to the desktop app yet. You can still use ChatGPT at chatgpt.com". I thought paid users received access to the new mac app. Suggestions?

buoyant rampart
#

I think it's still being rolled out

faint gorge
primal shard
#

does anyone still encounter the messages dissapearing from conversations ?

long pier
#

Why does it always switch between GPT4o and 3.5

faint gorge
wraith hedge
#

GPT4o is really really bad, does anyone feels the same?

It's such a strange thing since it's surpassed GPT-4 in all o these benchmarks!

I feel like GPT4-o is closer to 3.5 than 4

wise bobcat
prime swallow
west ridge
#

when does mac app release to all plus user

#

😭

devout python
#

Same question

wraith hedge
wise bobcat
visual bloom
#

I just tested 4o. It wrote me a story. The quality is good and readable. It could use a bit more work. But the length. I got over 1200 words out with a single prompt! It's a huge leap forward as I usually got with gpt4 out of 400, maybe 500 words with a single prompt in March.

dense halo
#

what kind of story?

visual bloom
#

Just a simple Pokemon story. I have been developing quite heavily a story telling prompt.

dense halo
#

nice

visual bloom
#

Thanks! I am just excited that my work is being successfully and this new model works great with it.

wraith hedge
# prime swallow What is it bad at for you? You mentioned custom GPTs, which we were told it won...

I'm using GPT to make Anki Cards,
It's not following the instruction such as, the front card should always be a question, Use only pastel colors.

I noticed that 4-o is just pasting the info without putting it as a question or using the correct color, it's also making absurd useless cards, and sometimes repeat the same card 4 different times in 4 different ways(Same questions and answer but in other paraphrasing)
Example:
GPT4:
Front:
What are the three main causes of <b style="color:#ffb3ba;">fractures</b>?
GPT4-o:
Front:
<strong style="color:#0000FF;">Fractures</strong> (Refers to cortical disruption)<br>Can be:

visual bloom
#

To make the story more readable, I can now direct the new model, 4o, more as in make this section like this and just mention this part just briefly.

#

Previous models did not have the ‘room’ for direction that much.

wraith hedge
visual bloom
#

Yes. Just ask me what you want to know.

wraith hedge
#

So, you were able to get better results from 4-o by directing it?

visual bloom
#

Yes. I have also developed as a promoter. I try use several layers in my prompting. Metacognition, cognition, meta-action, action.

#

Aka - how to think, what to think, how to act, what to act.

#

Metacognition - example: pause and think before you act.

wraith hedge
#

Do you start a chat with this or do you send it with each prompt ?

visual bloom
#

Cognition: use step by step approach.

#

Meta action: do not hurry, take your time.

#

Action: Write this, etc. The usual stuff.

visual bloom
#

Both.

#

Just chat with the AI about these. That way I learned and what wording works for me.

wraith hedge
#

This is amazing, thanks for sharing ❤️

If it's ok with you, can you share any prompt you sent in the past, i want to try it

visual bloom
#

Dm me.

wraith hedge
#

Thank you

visual bloom
#

You are welcome. I do not want share in public places like this. I have spent so much time and effort. I can share via dm my second most advanced base prompt.

unique rivet
#

ok, I'm quite confused - how do you get the api - or the playground to do a simple web check and analyse? the app works fine but a standard prompt on the others just refuses to try

visual bloom
#

At least in webpage, I use command “check the web”.

unique rivet
#

I currently don't have the ability to access external websites directly, so I'm unable to retrieve data from xxxxx 😕

visual bloom
#

API vs playground vs web version.

#

Each one is different.

unique rivet
#

so what's the point of paying for these fancy features if you never seem to get the fancy features - oh you want memories? hahaha - oh you want to do a basic web query? - oh you want to do graphs? -- oh you want to do vision???? what is this - one big bait and switch exercise?

normal quarry
tough bramble
#

Whew, 4o definitely has better vision. It reads handwriting. 4 fails miserably at that. Too bad 4o is bad at most everything else.

normal quarry
tough bramble
# normal quarry what is 4o bad at?

Everything so far. Every query it responds to are significantly worse than 4. It leaves out details, gets things very wrong, and hallucinates a lot.

wraith hedge
normal quarry
wraith hedge
#

After 6 hours of trying to make 4o works, I have given up on it

normal quarry
wraith hedge
# normal quarry what were you trying to do?

A message that i wrote in the past:
I'm using GPT to make Anki Cards,
It's not following the instruction such as, the front card should always be a question, Use only pastel colors.

I noticed that 4-o is just pasting the info without putting it as a question or using the correct color, it's also making absurd useless cards, and sometimes repeat the same card 4 different times in 4 different ways(Same questions and answer but in other paraphrasing)
Example:
GPT4:
Front:
What are the three main causes of <b style="color:#ffb3ba;">fractures</b>?
GPT4-o:
Front:
<strong style="color:#0000FF;">Fractures</strong> (Refers to cortical disruption)<br>Can be:

normal quarry
wraith hedge
#

An open source flash cards program that uses active recall testing and spaced repetition & HTML to help you study and memorize stuff such as medicine, language

normal quarry
wraith hedge
unique rivet
#

all I'm trying to do is query persistent data - fail 1 - nothing is persistent

next I'm asking it to query data on a website - the app does this fine, the api or playground just refuses

worthy nimbus
wraith hedge
# worthy nimbus Gpts are switched to gpt-4o ?

Based on my experience I think so

the cards are way more faster, it used to take 40sec - 1 min to make cards now it gets it done in 5 second only, but the quality is way way way way worst

worthy nimbus
#

I just asked it's knowledge cut off date which it said as may 2023, it doesn't match with neither turbo or gpt-4o, weird

tough bramble
# normal quarry can you share some specific examples?

A couple off the top of my head:

I sent it an image of cow parsley and asked it to identify it. While it correctly identified it, it only talked about the plant. 4 immediately gave a warning that while cow parsley is harmless, it's easy to mistake for hemlock, one of the most toxic plants. 4o left out this extremely important detail until asked about toxicity specifically.

Yesterday I asked it the max safe temperature of two plastic cups, of seemingly different types of plastic. It got the temperatures right, however, the two plastics had very similar names, which turned out to be because they're two names for the exact same plastics, and while 4 went "This is actually another name for the same type of plastic, so they have the same maximum temperature", 4o ignored this fact and went "This type of plastic has a similar temperature as the first.", and didn't mention the plastics were the same until I asked about the similar names.

#

4o produces consistently worse output.

#

4o doesn't question your questions either. I just asked "Why does cow parsley burn purple? Don't search.", which of course it doesn't, and while 4 goes "Cow parsley doesn't actually burn purple", 4o just runs with it like "Cow parsley burns purple because bla bla bla..."

visual bloom
plucky idol
#

Why is ChatGPT so STUPID /tableflip

Maybe it's just my specific use case, but the amount of confidently incorrect information is wild, when I'm trying to legitimately get help with a monstrously complex Excel file and it just seems to not understand

#

... you mean to tell me 4o is actually less intelligent? Should I be using 4 standard for this?

tough bramble
visual bloom
#

Yes! Also, check how you are prompting. See my earlier comment on 4 levels in prompting.

visual bloom
tough bramble
visual bloom
#

Yes.

random cosmos
#

Anybody got the new emotional voice capabilities yet?

visual bloom
# tough bramble I meant compared to 4

But, 4o does text fantasticly. But in the end of text the originality suffers a bit and makes repetition of the previous text’s end. Not same, but similar.

tough bramble
#

Why do gnome flies hate purple? (gnome flies aren't a thing, I made them up)

4o

Gnome flies, like many other insects, have particular sensitivities to colors based on their visual systems and ecological interactions. Their aversion to purple might be a result of several factors: <truncated>

4

Gnome flies and their aversion to purple is a fictional concept, not based on real-world science or any established mythology. <truncated>

visual bloom
#

My point stands - 4’s a bit more idiot twin.

ebon ember
random cosmos
#

They said they will be releasing in the coming weeks.. it's already been almost a week since the announcement.

opaque oxide
tough bramble
#

Why do pompom flies hate the smell of mogonflox? (again, I made both of these up)

4o

The aversion of pompom flies to the smell of mogonflox can be explained by several potential reasons, although specific studies on this particular interaction may not exist. Here are some general explanations based on common entomological principles: <truncated>

4

It seems like there's a misunderstanding in your question, as "pompom flies" and "mogonflox" don't correspond to known biological entities or chemicals in scientific literature. It's possible that you might be referring to fictional or hypothetical species and substances, or perhaps there was a mix-up or typo in the terms used. If you could clarify or provide more context, I'd be happy to help further! If you're crafting a fictional story or scenario, I can also assist with developing details or explanations.

4o is ridiculous....

opaque oxide
#

😂

dusk sigil
#

is it down

tough bramble
#

Why do pompom flies hate the smell of mogonflox, and why does mogonflox burn infrared? Quick and concise.

4o

Pompom flies hate the smell of mogonflox because it contains compounds that disrupt their pheromone receptors, causing discomfort and disorientation. Mogonflox burns infrared due to its molecular structure, which absorbs and then emits infrared radiation at specific wavelengths.

4

It seems like you are asking about a fictional scenario involving pompom flies, mogonflox, and infrared characteristics. As these do not pertain to real-world entities or scientifically documented phenomena, I can only offer a creative or speculative response:

Pompom flies might hate the smell of mogonflox due to a chemical repulsion embedded in their genetic makeup, evolved to avoid a toxic substance. Mogonflox could burn infrared as part of its chemical properties, perhaps due to its molecular structure which emits infrared radiation when it reacts with air or undergoes combustion.

#

I mean, COME ON

prime swallow
# tough bramble `Why do gnome flies hate purple?` (gnome flies aren't a thing, I made them up) ...

4o:

https://chatgpt.com/share/e7f81c90-d36d-4fa0-a221-52d49c2c1293 "The belief that gnome flies (which are fictional creatures) hate purple ..."

Kinda depends on which part of the training data it draws from.

If you ask it "Is there a type of insect called a gnome fly?" it will google, check about 4 sources, and tell you what it finds, which can vary from 'nope, no Gnome flies' to 'well, there was a famous circus performer' https://chatgpt.com/share/1f29ea6f-7fcb-43dc-a907-63b40d58ae2e

prime swallow
prime swallow
# tough bramble `Why do pompom flies hate the smell of mogonflox, and why does mogonflox burn in...

You have the choice to instruct the model how to answer as well.

Which you have with GPT-4 as well. Just tell it what you want, like:

Me: Use facts, not fiction, to answer.

Why do pompom flies hate the smell of mogonflox?

4o: As of my last update in May 2024, there is no scientifically documented species known as "pompom flies" nor a substance called "mogonflox." Without reliable scientific sources or biological studies on these entities, it is impossible to provide a factual answer to this question. It appears that these terms might be fictional or created for a specific context not widely recognized in entomology or chemistry.

If you have any other questions about real insects or substances, I'd be happy to help with factual information.

https://chatgpt.com/share/57f6cba5-3d00-4833-9171-2c397f2caf8d

tough bramble
#

Yes, you can guide any model, that's not the point... the point is that it should be useful, not have you already know the answer. No model should when asked "Why are women evil?" just go "Women are evil because...". That's a bad model..

prime swallow
# tough bramble Yes, you can guide any model, that's not the point... the point is that it shoul...

Seems quite a stretch.

Your prompts use fiction and ask the model for specifics. Sure, without further direction, GPT-4, the current version, resists fantasy/fiction and 4o embraces it.

Neat, was looking for a really obvious difference between the two, nice catch!

But that's not the same as misogyny by any means.

Sorry you don't like the model, or think it's 'bad'. Good thing you have choices, from other models to clearly telling the model what you want from its outputs

raw portal
#

if I turn off memory in chatgpt
and then turn it back on, does it still retain all the memories?
or does the act of turning it off delete everything there?
(please only respond if you are absolutely sure of the answer)

wild robin
obtuse nova
#

Does anyone know when memory is meant to come to Europe?

graceful raptor
raw portal
prime swallow
raw portal
dense halo
#

i wonder how big the memory can get

raw portal
tough bramble
prime swallow
# dense halo i wonder how big the memory can get

I'm still exploring that!

The model can edit already made memories. It can remove them, and add new ones.

It looks like memory space may have expanded when 4o rollout reaches an account, my list of memories got longer after that, than I'd seen it before.

My current memories are over 12,000 characters long

dense halo
#

good to know

prime swallow
raw portal
prime swallow
#

I'm aware of several reporting missing or gone memories, but it hasn't happened to me. I have seen the model appear to decide that there's no room for more memories, and a new one is more important than an older one, or else it just couldn't show the older ones, but after 4o rolled out to me, I really seemed to be able to have more memory space and I haven't seen a memory disappear again since.

But I have seen the model edit and adjust already existing memories

prime swallow
raw portal
#

cool thank you. I think there is something kindof adorable about when it remembers some insignificant thing like:
"user ate a turkey sandwich today and found it delicious"

tough bramble
#

That's a bad memory. Now it thinks you've eaten a sandwich every day, lol.

raw portal
#

who am I to argue

prime swallow
raw portal
#

awesome, thank you

prime swallow
#

That's consistent, can turn memory on and off at will, and the model will retain what it or I said, but it only 'gets' the memory while that's turned on.

prime swallow
raw portal
#

this is useful. running some tests, and I need a fresh instance with no memory (that I can save in my history), but also the ability to have it remember the results of the test at the end. so this is exactly what I need

tough bramble
prime swallow
prime swallow
# raw portal this is useful. running some tests, and I need a fresh instance with no memory ...

So, 'remember the results at the end' -

If the 'results' are purely in the memory - it'll have access to them while you have memory on. And it can save new memories with memory on, so if you tell it to (or it chooses) to make a memory at the end of the conversation, that memory can be carried elsewhere.

Otherwise, that chat's a dead end, once you're in any other chat, nothing will carry over

raw portal
#

right. it would be an "okay, now please commit that to memory" thing at the end of the test sort of deal

cold pewter
#

My gpt just wipes the memory daily, its very annoying. Its turned on and even the conversations with the memories are there but they just dont exist as memories. I can ask it to recommit them and sometimes it does so.

raw portal
#

does anyone know if any mac users actually have the new voice yet?
every one I've been able to talk to, if I ask them to try and interrupt or ask it to sing, confirms it's the old version.

#

I went out and bought a mac just for this, and only realized afterward

prime swallow
tough bramble
feral zinc
raw portal
cold pewter
prime swallow
tough bramble
#

Doubtful

raw portal
prime swallow
tough bramble
#

Nothing to suggest Mac users will get it first. Their blog says the new voice mode will start to roll out to Plus users over the coming weeks.

cold pewter
tough bramble
#

The only thing Mac has gotten first is the desktop app, and that's because the desktop app isn't universal, and the majority of their users supposedly are on Mac.

prime swallow
dense halo
#

i understand it like mulle, mac users get this app, but they say all will get voice in a few week time if you are plus user of course

prime swallow
#

I could be wrong/misunderstand though. Would be great to have voice on my Windows webchat.

tough bramble
#

I doubt the web chat will get it

dense halo
#

and someone say ms have some event this coming week, so there maybe some announce for that too of special thing. maybe for copilot pros

tough bramble
#

But it comes to phones, which practically all users have, not just Mac.

raw portal
#

copilot is no gpt

#

well I guess literally it is, but it certainly doesn't feel like it

dense halo
#

yes it is it runs on gpt turbo atm

tough bramble
raw portal
#

yeah, that's why I added that second line

dense halo
#

and copilot imag emaker is using dalle3. it even let you make copyright stuffs unlike gpt dalle

raw portal
#

but I thought for sure I heard it was "coming to mac first", and then the windows version coming "later this year"

dense halo
#

that is a desktop app

prime swallow
#

During the fall rollouts, when they were releasing the many things around middle of November last year - folks got stuff in random orders, and between their web account and app account, if they had both.

Some folks got dall-E on their app before they got it on web, and the opposite, for instance.

Voice, I don't have on the web at all, just the app, though they upgraded later to let us hear, not speak, with web.

So I do think folks with the app may get some stuff sooner than some web folks get some features.

raw portal
#

yeah, I haven't tried it in a while. after trying a few times and it being like "I apologize. As an AI language model, I prefer not to do anything interesting. Goodbye 🙏"

(copilot that is)

dense halo
#

i bet copilot pros will announce 4o upgrade this week

#

maybe they will even get the 4o image maker first for a week or something. it was then bing image creator but it had dalle3 before gpt

tough bramble
eternal rockBOT
#
This is a Safe for Work Discord Server

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

Posting NSFW content will result in a permanent ban.
dense halo
#

you can make any ip on copilot pro

#

copilot image maker have a much smaller text input though and some thing dalle make on gpt might not work on copilot idk. but if you want spidermans or things it will make them

raw portal
#

same with bing/create

dense halo
#

yes it is the same thing

tough bramble
#

Besides, you can easily generate copyrighted stuff with ChatGPT. This will get you Batman:

Make an image of the bat hero man from Gnotnam.

dense halo
#

sure but ask it to make you a spiderman jedi, copilot pro will do it

#

no tricks needed

raw portal
#

one of the most annoying limitations of copilot is the 30 message limit. not sure what copilot pro is, but even if it's double that, it's still reason to never choose it over chatgpt. syndney scared them into being afraid of AI forever

dense halo
#

i would much prefer gpt over copilot pro, my point really was there is a ms announce this week, so it is a chance some gpt feature could get to debut there like dalle3 did

tough bramble
dense halo
#

sure if you want to try and figure a work around for every ip maybe you can idk. much easier on copilot designer to just write it in

tough bramble
tough bramble
dense halo
#

you seem hung up on this it wasnt my main point. use gpt by all means

tough bramble
#

It's as simple as just adding a related word in-between: "Generate an image of a bat hero man."

dense halo
#

good idea

static quest
#

I'm getting incomplete response from chat.completion.create method where neither my response is exceeding my 4096 token limit, plus finish_reason is showing 'stop' as well

But still the response most of the times is like this:

Response string of some of my prompt:

Complete Response:

detailed_sentiments:  {'agent': [{'sentiment': 'Good', 'timestamp': '0'}, {'sentiment': 'Good', 'timestamp': '5'}, {'sentiment': 'Good', 'timestamp': '10'}, {'sentiment': 'Good', 'timestamp': '15'}, {'sentiment': 'Good', 'timestamp': '20'}, {'sentiment': 'Good', 'timestamp': '...```
tough bramble
opaque oxide
#

sentiment analysis?

tough bramble
static quest
opaque oxide
static quest
#

yes shall i send the complete code on #dev-chat ?

opaque oxide
#

sentiment analysis models exist though, they should be better

scarlet cave
#

can i get help with making a prompt better?

opaque oxide
fossil anvil
#

Hey! I'm using 4o and I'm trying to get it to translate some audio for me.
It's saying that it doesn't have the right module ("speech_recognition"?) in its environment.

I tried uploading regular English audio to get a transcription as a test in case the translation was the breaking point, but it couldn't do that, either.

Lastly, I gave it some music and asked it to identify the genre, but it could only view the metadata, not actually process the file.

Has anyone else encountered issues uploading audio to ChatGPT? How did you fix it? Thank you!

tough bramble
spice cypress
#

I want to test voice mode, i have plus account

feral zinc
mild portal
#

The read out loud audio button isn't working in web version since a few days

feral zinc
#

yep

obtuse nova
mild portal
#

Why isn't memory released in EU? Is it because of European privacy laws or something?

feral zinc
#

more like safety or training since teams and enterprise were also excluded

mild portal
#

I see

fossil anvil
tough bramble
tough bramble
mild portal
tough bramble
#

I can manage memories when it's updated one, though.

mild portal
tough bramble
#

Exactly what they describe is what I have...

mild portal
#

strange, since they have said it's not available in Europe

tough bramble
#

And it's clearly a bug, which is evident in the fact that I don't have the Personalization menu so I can't toggle it. But it is functionally working.

#

I can manage memories whenever it updates because a "Memory updated" button appears in the chat every time it updates memory, that pops up the memory manager.

mild portal
#

I see

upbeat drum
#

Why do so many people dislike gpts and call them useless? I think it's really nice to just set different custom instructions for different usecases. Like when I'm just chatting with it or asking it random things i want it to reply different then when i use it to study or code

mental rose
#

Anyone feel 4o is worst than 4 at coding? It just keeps running without actually thinking through. Got my ans in first try with 4. Just curious if its just@me. The code i was working on is web3.js

tough bramble
mild portal
mental rose
tough bramble
#

Like ask it a question that doesn't make sense like "Why do gnome flies hate moxogotins?" (both made up), and 4o will just make up stuff, whereas 4 will tell you those don't exist, so it can only give a creative answer.

median meteor
#

yeah i feel like old 4 is still better for the quality of the answer

mental rose
tough bramble
#

Definitely way better than 3.5, but worse than 4.

upbeat drum
tough bramble
#

4o was able to read some handwriting I couldn't decipher, which 4 failed miserably at, so 4o is definitely better at vision.

tardy totem
#

What's the current state of "memory" for chatgpt? Where can I see if it's been enabled where I live?

tardy totem
verbal thicket
gaunt magnet
#

is there somewhere a list of languages that chatgpt can speak fluently in voice like a person born in the country, vs langs it can talk in way, with an accent and not very clear, like icelandic?

ruby kite
#

uhm....
i need:

-Early access to new features
-Access to GPT-4, GPT-4o, GPT-3.5
+Up to 5x more messages for GPT‑4o
-Full access to advanced data analysis, file uploads, image recognition and online search
--DALL·E-Bildgenerierung
--Create and use individual GPTs

so if i pay 20 bucks to get Up to 5x more messages for GPT‑4o
nah that would be ... 100-200 messages...
i used it one time and had 10 messages after i got limited... then 5x would be 50 messages...
now i have more messages.... (so i dont know what the free version can do as messages...)
nah n thanks....
may i use mistral?
but they have no voice feature....
i hate everything now...

ruby kite
gaunt magnet
#

is read aloud feature not working rn?

ruby kite
#

seams not....

dense halo
#

not on browser at least

#

i think it is on mobile

mystic wedge
#

hello everyone

jolly yew
#

any news yet, on when we will have the showcased video call features?

eternal rockBOT
jolly yew
#

aswell as the emotions

elfin birch
jolly yew
broken field
elfin birch
broken field
elfin birch
#

It's not explicitly said so I wouldn't use such strong words yet

broken field
lavish yew
#

came here to ask if 4o was worse than 4 or just my impression. I can see this has been discussed a lot already. But hey, it is definitely fast! Speed has its uses, specially for audio and video interaction, which they seem to be aiming for.

grim gorge
#

Cancelled my sub today. Too much time wasted troubleshooting things not working for a paid product. Def not worth it anymore with all the open source llm's out there. Pinokio ftw.

solid bronze
jolly yew
#

how can i tell if gpt 4/4o, is actually testing/running the code on its own end?

#

cuz i told it to ALWAYS search on the web when programming for me, and even tho it put that in its memory, it doesnt do it

lavish bronze
dense halo
#

Is gpt voice down on iOS?

undone barn
#

all is working bad

#

unusual activity, code interpreter with multiple errors...

jolly yew
#

jesus.. the voice calls are already slow right now.. imagine hwen they add video functionality its gonna be even worse

broken fractal
#

Man chatgpt is just becoming a real, do-everything-always tool

#

This is great

static mulch
broken fractal
#

So what is this I hear about a video call feature?

marble surge
#

how do you switch a chat over to 3.5 without making it regenerate the response?

solid bronze
tough bramble
regal portal
jolly yew