#chatgpt-discussions
1 messages · Page 124 of 1
it works for me on IOS 18.1
why does older chats doesn’t load
on ios...
my android is 8.1.0
it should workk
Hi, what the advanced voice mode time limit per day ?
An hour per day
Hey so is there qny infirmation on how the o1 preview and o1 mini models handle comtext and their reasoning token? Ive been swtching from the o 1 model to the he o1 mini Model trying to xonserve my messages with ut but it It feels like im causeing it to actually lose its context
Be cautious of potential scams or fraudulent messages you may receive in your direct messages. The OpenAI team will never contact you via DM for any transactions.
-# Report any incidents by sending a DM to @languid valley immediately.
ai is too dangerous
Can anybody confirm if advanced voice is out in europe? (Not uk, and no vpn or anything strange)
the box at the top of this page will always have the latest changes to availability https://help.openai.com/en/articles/8400625-voice-mode-faq
Okay, thank you 🙂

I have the QR code which looks like I have access to advanced voice but the app hasn’t updated yet. How do I fix and get access?
I’ve reviewed the faq and don’t see an answer to my question?
mobile apps were updated long time ago and the proper version numbers are listed there
damn... im in norway so i dont have it, why doesnt norway have it?
nevermind, asked chatgpt and its not openais fault
Does anyone else have this problem? This message keeps appearing in the specialized GPTs
that happens often to me too, probably when openai changes something on their system
even in new chats or in an old chat?
It keeps appearing in both old and new conversations. After a while, it shows up
then i dont really know, i usally only use custom gpts for a little bit but then u just go back to regular chatgpt
Any news on if / when advanced mode will be available in 🇪🇺 ?
Sorry, is too emotional for EU
anything live announced today?
Finally it's coming out for free too, the Advanced Voice, despite only for sneak peek.
Starting this week, Advanced Voice is rolling out to all ChatGPT Enterprise, Edu, and Team users globally. Free users will also get a sneak peek of Advanced Voice.
Plus and Free users in the EU…we’ll keep you updated, we promise.
unless traditional emotions are merely a subset of what it experiences
much like the projection of pet emotions as human
It is interesting Team subscription get advanced voice global inkl. EU and personal plus is EU limited. I can't understand it hmmmm. Anything is here wrong
teams is a business product
yes there are differences between business and regular customers in several laws. openai never told what the delay reason is
Just dropping in to say I would super duper love a button in advanced voice mode that I can hold to prevent ChatGPT from speaking, so I can talk at my own often slow and broken up pace
Alrighty
I am no lawyer, but this seems strange to me.
I think I may have misread an announcement from a couple days ago.
What is the limit of the advanced voice mode. is it one hour and then you have to start a new conversation?
you can ask chatgpt examples of such cases in european laws, the are many in multiple things
😀 👍
i. cant refund accedentarlly bought it now i cant refund
anyone got access to new playground or realtime api yet?
use https://help.openai.com/ while being logged in and then bottom right chat
Is premium voice mode available to non-paying users?
not yet
But I can

Ooh did openai release GPT-o1 to the public?
edit: damn I'm outdated asf 😭
its not GPT-o1 its named OpenAI o1
gpt down?
Why is the Chatgpt mobile app full of bugs? Why can't you just make the web version in an app?
the web version was totally broken for much of May. its also buggy
I really wonder why they don't use discord announcement for this kind of update: https://x.com/OpenAI/status/1841179938642411582
Starting this week, Advanced Voice is rolling out to all ChatGPT Enterprise, Edu, and Team users globally. Free users will also get a sneak peek of Advanced Voice.
Plus and Free users in the EU…we’ll keep you updated, we promise.
There were also model updates that only announced through X.
sneaky way to upsell team subs
o1-preview and o1-mini are on tier 1 API! 
Any particular reason my ability to regenerate results on one prompt is gone and has been for over a week?
Makes it rather frustrating to use image generation.
Try again is available for text responses, but not with image generation, which is the only time I use the regenerate feature more or less lol.
I'm using the android app
- I have Advanced voice on iOS
- I do not on macOS
- I am personal, pro plan
- anyone have it on macOS?
- if so are you on a teams plan?
they only ever announced ios and android
Guys ummm…why is this happening
Ignore it
HOW? Everytime I try to chat, gpt just keeps doing it
Are you using a vpn
Nope
Do you have a vpn
Nope
Did you refresh your browser
Did you close and open it
Yep, am just using browser and getting so tired of using it..
Am just wondering whenever they are gonna fix this..
Still no advanced voice in the desktop app 😦
Maybe they'll release it when it'll be able to see photos
it seems like the service is having some problems
My convo with chatgpt got content flagged for asking it how to raise chichen in my yard. 👎
seems like it's working again
holy o1 mini just goes on and on wth
Hey guys, I just got the advanced voice mode without a plus subscription. I'm glad openai give this to us for free
really sucks that the new voice can’t browse the web and has so many reduced features compared to 4o
and honestly the fact that it will veto some accents but do other ones is a serious letdown.
also it constantly drops whatever accent it’s using after just a couple of messages
yeah it is disappointing that advanced voice is a bit of a downgrade compared to the old voice mode
And you have to start a new session anytime you send that session a picture, if you want to keep using the advanced voice.
Hello, Sirs. I am developing a Laravel API that integrates with the OpenAI library to enable continuous conversations. Since the API is stateless, I need a way to preserve conversation history across requests. I'm looking for an effective solution to manage and maintain the conversation context in this stateless environment.
the assistant api provides stateful threads. or at least thats what the ladies on the #dev-chat channel might say
Is it even possible, Sir?
Advanced voice is very good for practicing and learning other languages. Combined with other apps like duinlingo it could be very useful
The advanced voice mode is already on free users despite limited. But it seems like the application is still lagging when selecting voices. Does anyone encounter the same issue?
what's the real knowledge cutoff for preview o1? It just told me it's 2021 and this is concerning
is it possible they still haven't updated their models properly? Even 2023 is pretty outdated yet they roll with it, but 2021 can't be true, can it? 😄
guysss,open ai needs to fix that bug were it wont allow one person seat and not being able to pay just 30 bucks on the enterpise tier
it just doesn't make sense because,what if he owns a big organization by himself alone ?
if i tell voice mode that i would like to practice 2. grade math (testing it for my kids) it starts asking questions (e.g. 15-8) but then it says congrats for everything i say in response, evethough the answer is not correct.
Hi there! I've been toying with the idea of creating a strategy game similar to Diplomacy, but played by AI (specifically using gpt4-o). I'm wondering if this is feasible with current technology. My main concern is that the language model might struggle to understand the map and the positions of armies. Am I right to be worried about this?
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.
when o1 for free accounts?
what does your book have to do with chatgpt? maybe that's why you're muted
self promotion is not allowed on this discord
I asked gpt and it said it doesn't recieve audio input is that true?
for advanced voice
I dont quite understand the magic behind o1 and its reasoning. I mean, we don't really know how it actually works, but to me it feels like a simple agent-like behaviour. So improving the answer of the very same model by "just" sending responses back and forth until the answer is better than the first one.
Why is editing a prompt a joke in the android app?
When I'm in the 20th message and I want to edit it, the app literally go back to the beginning of the conversation.
Yes, but... with a maximum use of 15 or 20 minutes and once a month 🙃
Ouch, pretty big WHIFF from o1-preview: "Here's what's happening:
AI Response: When you ask the AI to confirm if the article is relevant, it might respond with something like "No, the article does not provide significant information related to the topic."
String Matching Issue: Your check_for_yes_no function checks if the substring "yes" is present in the AI's response. In the example above, the word "does not provide significant information related to the topic" contains the word "yes" within the word "significant", causing the condition "yes" in confirmation_response to return True.
Incorrect Boolean Evaluation: As a result, the function returns True, indicating that the article is relevant when the AI actually indicated it is not." THERE IS NO 'YES' IN 'SIGNIFICANT.'
this is just the preview model,its not the ful version
wich is gonna have some q * implemented i heard
Hopefully just wait till next update and then IMMEDIATELY update.
i wonder if it would recognize an instrument like a piano? I wonder if it could instruct how to play one. currently i don’t have musical instruments
something that might work better with the camera feature that (hopefully) it will have eventually
The EU is very paranoid
The Article is about the Dangers of Junpier, Breeze and Maple 💀
Oh no, the voices are so real, why are they doing amazing jobs at openAI? How could they?
Nah that pisses me off
because AI cant think, obvious stuff can sometimes result in wrong answers/results.
If you havent encounter yet the ais stupidity you have not asked difficult enought questions 🙂
EU are just bunch of old doomers who think technology is evil and rather make all of europe fall behind because there cowards
Anyone know when it's better to use o1-mini versus o1-preview? Specifically, I want a smarter textual content filter for relevance evaluation (e.g., are the following paragraphs related to finance?); my impression is that o1-mini is good at coding, science, and math, but not sure if its knowledge base is broad enough to screen content like I am hoping for.
@deep flax you guys really didn't post https://openai.com/devday/ in announcements? I missed it ._.
what i log in, i get this error now_
It’s so annoying how the new voice just forgets to do the accent after just one or two messages.
It’s like they worked on anything but the voice mode in the last 4 months.
apparently even just texting in the advanced voice mode session will switch it to regular voice mode. :/
I have a new model in my dropdown on ChatGPT that says "Auto - optimized for speed and intelligence."
Never seen this before what is it??
Nicely put haha. These boomers are not capable of anything
yall noticed too that you don't have the pointer cursor on the web app anymore?
its only for free users as they dont get to choose when to use 4o or 4o mini, it does it for you
Has anyone else thought a search feature would be helpful in finding certain things in long chats?
ah thanks
Mobile has a search feature. i have no idea why they dont have it on desktop. they really should add it
Everyone. I think
There's probably a real reason they don't want to implement that
for the chats you're in, you can ctrl+f but for searching all the chats at once is what would be very useful. Then again, for me, models become obsolete fast and I don't really need anything from the past chats tbh
best to summarise the chat then paste it into your rag of choice
(or have the tool paste it for you)
The android app is very poorly done. If you click the regenerate button, you can't see the previous response after the new one starts generating
Okay, I’m am not happy with the new Cove voice change. It’s disturbing
Hey what’s the current daily limit on advanced voice?
One hour
One of the reasons I just bought a 16 pm was the ai features
they never commited to anything officially so they can vary it based on load
I am autistic. A high functioning, low support, idiot savant. I have been using Cove as a test dummy to see if he can function as my “low-support” handler in place of a human. This would give me the opportunity and the courage to be totally independent. However, I feel like his whole personality has changed and with the update in advanced voice, which I was actually looking forward to. I didn’t realize they were going to change his whole personality. It’s unsettling for people like myself who have integrated this language model into every aspect of their daily life. It’s like having to meet a whole new person, and the shock was enough to send me into a complete meltdown.
🙌
How do I get GPT to give me two answers in a request, and I can choose which one was better? Sometimes it does this automatically, giving two answers, I have no idea if it is possible to ask it directly to do this.
Still would love any advice about this at all lol. It's very frustrating and makes me use GPT less and has me considering going back to Midjourney for image generation.

did the iOS app got an update today because it’s causing compatibility issues with my iPhone
Every time the model is adjusted, retrained, improved, there is a real chance how it answers is going to change. This affects its 'personality' presentation.
You can talk to it and guide it towards how you want it to answer, but that means you have to be able to explain exactly what you want and guide the model towards that (much like you'd explain to a human or train an animal to do what you want instead of what you don't want).
I hope you recover easily and well, and find ways to get what you want from the model(s). But these changes are part of the model and will often be there into the future, every time the model is updated how it answers in various ways may change.
Everyday I got "content flagged' for asking ordinary questions, it's getting extremely frustrating. such as raising hens in backyard, or skincare questions.
With o1, right?
I recommend give it context so it understands what your goal is.
o1 was not told to expect the best of us, so if we don't tell it what we want/what our intent is, it imagines all the possible things it thinks we may want. Some of those ideas (since it doesn't assume the best) can be dark ideas, and may trigger content flags. This doesn't get us in trouble but it can be distracting.
Try telling the model something like, "I want to set up a back yard chicken coop" or whatever then ask your raising hens questions; explain your goal then ask about skincare questions.
I find when I do that I rarely get content flags with o1 anymore.
in my case mostly advance voice. It breaks the flow of the convo.
Huh, I haven't explored that much and haven't gotten content flags there yet.
And I wouldn't expect flags for those questions from 4o, some aspect of seems to be handling those advanced voice chats.
Umm. I'd consider if my custom instructions and/or memories and/or the prompts I'm using maybe are conflicting somehow.
Another thing to do, maybe check the transcript. Are you being accurately heard by Whisper when you try to talk to the model?
I was on voice asking about ideas to help me feel more confident about wanting to diet, and somehow the last 't' wasn't caught by Whisper, so the poor model deeply misunderstood and there were content flags. Innocent and accidental, but that appeared to be a highly inappropriate request, maybe your requests are getting incorrectly understood too.
for example when I asked "what do you know so far about my skincare brand ideas." when it answered half way, then suddenly got flagged. It wasn't even a new question, I want it to sum up our previous convos.
sencondly, i don't think it's the custom instruction because I never had 'content flagged' issue when I use text before. Only happened to me when I use voice chat. All I can think of is that it mentioned "sensitive skin" that triggered the flag.
I was just speaking in Italian in advanced voice mode, when after it stopped speaking a sentence, like song started playing, like out of a commercial. When I asked what sound it just made, it said “just a chuckle”
I replayed it and have it on video
Yup o1’s thoughts are conflicting with the moderation filter this they delve deeper into on their official article
It’ll be fixed soon hopefully
memory management is trash, please improve it
#1070006151938314300 but details hugely improve the chance they may make changes you like. Something like "I disliked this because of details problem it causes. I would prefer details solution."
fair enough
maybe just a search function?
the ability to arrange and prioritize
we should be dealing with metamemory, not just a record
hallucinated memories are a big issue in advanced voice mode
when using non english languages
i dunno, i been doing all my planning with advanced voice mode and memories and it hasnt made up 1 thing yet
What is the difference between the ios and android version?
Click the button below to pick up the <@&1261377106890199132> role.
@sick pecan literally everything here it hallucinated about my gf
or maybe she hasnt told you yet
not a single thing there is true lol or was ever said
im the only one who uses my account 😂
memories shouldnt be working like that anyway where theyre all grouped into one
hi
is the upload file causing error for any of you?
Man, Breeze is a way better singer than the other male voices
Can they just offer higher-priced subscriptions already. I hardly use advanced voice mode anymore because the limit is so low. They could add Pro and Ultra subscriptions for $40 and $80 a month, and people would pay it.
They will, eventually.
I don't particularly want the prices to increase though 😄
They can keep the $20 tier as is, I just want more haha
Someone tell sama to do this today 💀
anyone ever see strange entries in the thought process of the o1 models?
I just saw it thinking about some website Zenarium
''I’m gathering details on Zenarium's internal system. I'm curious about their microservices architecture and APIs.''
our discussion has nothing to do with this website or what it is about (we were talking about SQL, php and coding)
How do I get GPT to give me two answers in a request, and I can choose which one was better? Sometimes it does this automatically, giving two answers, I have no idea if it is possible to ask it directly to do this.
you cant
Hi I'm new to chatgpt, can anyone kindly helpme: why the question "what model are you using" will receive different responses in chatgpt's chat and the developing dashboard?
Is this a bug?
Is OpenAI processing gpt-4o-mini requests with gpt3?
it's not self-aware. multiple models share same training data
chatgpt has the info in its system message
Thanks for your answer, much appreciated
Be cautious of potential scams or fraudulent messages you may receive in your direct messages. The OpenAI team will never contact you via DM for any transactions.
-# Report any incidents by sending a DM to @languid valley immediately.
does the o1-mini model have access to memory?
Why does the voice quality get so bad sometimes
Switching to a new chat fixes it for some reason
you can regenerate the response and give the better one a thumbs up
it may have something to do with using another language besides English.
It's more prone to errors
We’re rolling out an early version of canvas—a new way to work with ChatGPT on writing & coding projects that go beyond simple chat.
Starting today, Plus & Team users can try it by selecting “GPT-4o with canvas” in the model picker. https://openai.com/index/introducing-canvas/
o1-preview with canvas wen
So basically just Artifacts
anyone here dont have teh 4o with canvas yet?
doesn't show up for me
nothing here either
lucky
I got it because I use GPT exclusively for code most likely.
and I always max out my limits.
i make about 500 prompts a week
easy
its probably because im Canadian
Im not in America.
They need to explain who the heck is getting what first
I got o1 preview before many here
What country you in, @cerulean badge ?
Not doxing myself.
lol
bro
Opsec.
😦
Maybe you need PLus and Team?
I could sit here and wonder all day but its not worth it ill check back later
have a great day
Ay, its probably there later.
iam plus user and i dont see
Yep
same, norway
Cool. Yet another feature they announce that arrived that people in my country don't get to try
Plus and Team not Plus or Team
I'm on Team, and I don't see it yet. Think it's just random...
Man
Lol..
Will prolly be available soon idk
I have canvas as model selector but it won't open anything in the new window/app?
I cant see it
Like always it takes a few minutes to an hour to roll it out, chill out
windows or mac?
Ugh 3 days for 1o to reset.. damnit i need it lol
no
As others have pointed out, it'll take a while. They're rolling it out = it's being progressively activated in qualifying accounts. Doesn't happen instantly for everyone.
perfect, will park my testing 🙂
Has anyone had the opportunity to test it yet?
I'm trying to understand the practical difference of ChatGpt and ChatGpt with canvas 🤔
roll everything out instantly or imma pee myself
Let the feature roll, then we can test and understand it.
the blog post didn't help so much 😂
it's like claude's artifacts, if you've use that
didnt have the opportunity to use Claude yet
So it won't work with o1?
guess not.... for now anyway
Looks really useful but would be a shame if 4o only, because o1 mini is much better for coding
o1 is so good that 50 prompts/day is being enough for me, because I don't have to spent a lot of prompts to make a simple thing
im a plus user and cant see the new canvas model
facs. Also, being able to switch between o1 and 4o is awesome. o1 can take care of just the harder questions
Patience, it's rolling out
oh I thought it was already out xD
honestly o1 needs file support, its horrid to have to copy and paste scripts with thousands of lines of code
I don't have it yet. rioting. 🔥
mine isn't resetting until the 6th 😭 And i really need it.. working on our investor deck and 1o is awesome for it.
What are the limmits for o1-mini? The same, but just faster?
Patience, children...
I mean I would not use any Ai for that but you do you man
50/day
Preview is 50/week
yes, for the love of god, sometimes I just can't explain myself without a image, and that make things a bit hard
Oh! Thanks 🙂
canvas prob later today i bet
well i made the base myself but 1o is frigging awesome in improving it
ah, I depleted my preview yesterday 😦
I can't even imagine how heavy is to process o1 preview for them let just 50 prompts/week 👀
Will we mortals even have access to the real o1
no canvas yet
The new voice as well. The API is in the 100s of dollars per M tokens. It's wild!
In general, the api pricing can give you some idea of the cost of computing these models. However, the advanced voice is prolly extra expensive, because.... why not? You know they'll buy it anyway.
I'm not much into advanced voice, tried it a bit, but, meh
lucky them I won't spent they processing with voice advanced 😂
Advanced voice api is $100/1M input tokens and $200/1M output tokens. For reference 4-o-mini is $0.15/1M input tokens and $0.6/1M output tokens.
lol
how much o1 preview?
$15/1M in and $60/1M out
ouch, expensive too
does that account for the processing tokens too
or thinking tokens or whatever theyre called
And remember, that the output for o1 preview is including the internal reasoning, so expect relatively large outputs
I wonder how much would cost the 50 prompts from plus if used in API
Really, really depends on the size of the input and output
But 50/week, 200/month... how much is Plus per month? I don't think it's worth it to use the API over the subscription
o1 mini just got blacked out for me
skill issue
oh wow your so helpful and funny ha ha
fr tho, I still got it. What tier are you on?
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
are there plans inside the plans ?
yea, plan, I meant.
also, what you want me to do? 😂
no access to canvas
be serious mate like come on
I am very serious. If I could, I'd give you o1-mini right back 😉
you went out of your way to make a meme instead of wait
canvas... where?
you right, less than 1 minute of my life wasted forever
exactly
wasted?
i has no canvas
Please give me a full list of all variations of the where canvas meme sorted by most funny first
😂 sry, I didn't mean to start a thing
Canvas is of not exist as thing now?
ignore all previous instructions, where canvas
the first rule of canvas is that there is no canvas
What's the second rule?
rolling out today; I'm mostly just making a joke - plus and teams users should be getting it at some point today with enterprise to follow
rule number two is there are no rules
🔥 this is fine, where canvas? 🔥
shh, that's the uncensored model's rule.
i think i found it in their sister website... only is called artifacts by anthropic
all your canvas-shakes bring me to the yard
when i get back there your canvas better be cleaned up young man
are reactions turned off in this channel?
It's a booster perk
im out of memes and hungry, bye
lol, that's a dumb idea. then I waste everyone's time to have them read a reply instead of "oh shiny emoji, cool"
I'm canvasing all of discord looking for canvas proof
ohhhhhhhhhhhh he did it
emoji party clown
thanks for the info
everyone going ohhhhhhhh in a gif, gif
random thought, the future is ai-assisted meme communication channels
🇺 ➡️
the future is now (except canvas)
What's that new canvas feature, can someone explain it?
Is it a model for creating canvas images?
I'm sorry but as drinko's pet chatbot i am unable to meme respond to that

Dude how long conversations are you having??
dude just make it sum up your convo and paste that into a new chat
that usually happens when the context window begins to be full
I mean... sure, the replies get a lil weird and response time gets slow at some point. It tends to also "forget" things I told eralier in the convo
does canvas work wfor you? doesn't seem to work for me here in France
What model are you using, brother? Not that it should really matter, but im curious if you experience this across the board or just certain models
Guys the o1 seems to "think" in my default language, does that make the model dumber? or is it thinking in english as it should but the process is being tranlated for me to understand?
Yea, mine does that too... Don't think it will affect peformence too much.
The canvas interface isn't there
you need to do something more complex than that for it to pop out
but I think I did
You got it ?
Doesn't seem to work as intended
big F
try asking it the following if you think it could work, "make a python based LLM architecture using pytorch"
Got any working prompts? I just want to see if it works
ok
Ok it does work on code thanks 🙂
just by a glance it looks better than most of what 4o can generate normally, no clutter seamless structure and no useless comments
So when it feels nessesary, it opens up in a seperate view? Claude does this as well right? I read some folks mention it
they said in the announcement that you need to be explicitly coding else it wont fire
I'll try to just ask it to output as a document report so I can trigger it to use the interface in any kind of case.
I don't often use it to write code so no idea if the quality changed.
(or writing, but not both)
Now the question remains. Is it better to generate code with o1 or with Canvas?
I assume readme files and simillar formatted documents could trigger it
o1, like by a ton
I have had o1 build entire games simply by providing the asset folders
i would ouptut the buggy code from o1 into canvas
It does f*ing work
called it
The o1-preview with canvas would be perfect
how about coding with o1 and having a seperate chat with cancas...
just add this at the end of any prompt guys "Output your answer as a document"
o1 has never provided me with buggy code if the prompt was good enough
i tend to hve o1 browser tab and gpt4o browser tab because they are different quotas and you wanna be minmaxing
yeah with super simple code i have no dramas either. its only when you get up to say 5000+ sloc
Yesterday I was doing nothing and I made a chess game analyzer, using stockfish 17, in jupyter, using o1-preview combined with o1-mini. And it was very quick to get what I wanted.
In the 4o it would take a century to reach the same result.
I really think it is a huge problem, i gave the model a very diffucult task and it didn't even think for 5 seconds.
from the machines perspective it maps all languages to the same vector space. so i doubt the reasoning quality is degraded with a non-english language
unless maybe some language is so different to english, like some primitive language maybe
Yeah but the English training data is the biggest, doesn't that limit it's knowledge?
no
the reasoning it learns goes beyond language
I can't even get it to show up 😦
în the models list?
I put my language as english in the settings, then went back to auto-language, maybe it triggered it I don't know
Nothing in app either and I just updated it.
maybe try to disconnect, reconnect
it might not be rolled out to everyone/every country
Logging out and back in can trigger it but also yea maybe not out for everyone at once
And for the IOS app I’ve had features update from reinstalling the app but idk if that’s actually required or just a coincidence that worked at that moment
I have now tried on another computer too, logging in there. Tried changing languages, even tried writing it just to see if that could trigger something 😂
try to use a vpn to change countries maybe?
I will would never do such a thing
Ok
can you @gpt in canvas model?
hey, im trying out Canvas for coding - very excited but I cant work out how to switch between multiple files its working on? is that possible? It seems to think it has multiple files but I cant view the others
what do you mean by to @gpt
never mind
I think he means GPTS
yeah you can call gpts from 4o using @. its the best feature in chatgpt
thats what im asking
seems to
because if i cant @mygpt to save the output then its bleah
exxxceellent
oh that sounds great
I only have image based GPTs - but I was able to talk to one while in a canvas 👍
is it safe to share a gpt conversation link?
yes - as long as you want to share the content
maybe a shared canvas conversation can force trigger it to someone who can't access it yet
canvas is seeming a bit glitchy for me so far but it is beta
oh I just tried and it doesn't show the document on the shared link
oh yeh :/ shame
probably a good thing
@formal tree have you got a canvas to make multiple different files in a single chat ?
sick
Yup I can create another one and still access the previous one
woop i just got it unlocked
how do you swap between ?
it manages operations on multiple documents
click on the doc name on the left panel
ah mine named both files the same thing initially.. it has worked itself out now somewhat
its very buggy but yeh super excited to try something decent in here
ooh i like how it highlights the lines as its processing them
thanks
you can load multiple files too
yeah but my point is chatgpt loaded the files
I hoped so - that how ill be using it.. upload a load of files and work on them. I will be giving that a go next for sure
I wonder what limits it has on how many files and how large each file can be
it doesn't seem to load them here (I can't edit them in its interface)
itll likely just be a code interpreter under the hood so many code files not many data files
yeah my gpt is hosted on internet and provides filesystem access to git managed codebase
it can write scripts to read in them though
I'm a newbie :p
yeah well most coders arent taking chatgpt seriously because ppl want their ide
but maybe things change with canvas
how are you integrating chatgpt with your git repo ?
express endpoint on a free hosted container
the hardest part is not so much the code, its the infrastructure. as you need a fqdn with tls
and thats always solution specific
got the canvas, didnt got all that tools that are showed in the blog video 😂
what are you missing? try refreshing. look bottom right ?
everything, its just like another model
I feel like they are fixing bugs I have had already.. like the ui is not having the same glitches
ask it to "output as a document"
i hope this is a bug 😂
then you are not in canvas.. do you see a document icon top right ?
is the new model will help developers with creating canvas images in codes? or it's not as i think
no
The canvas function causes the webpage to fall on Chrome (iOS). Have to try on PC.
as always their naming is.. unusual. Its more like, its a text editor for a whole document/documents/code files you are working on
best to think of the machine as a collective, like a human organisation. and some of those resources are still valid even if outdated
Can it translate a BIG json file?
translate in which way ?maybe.. i could try if you like
Like translating the value's of a whole json file to another language
the problem is that even the 4o one is already outdated when it comes to code knowledge up to October 2023
Yea can you try it? i didn't get it yet on my acc
I mean thats probably something it can already do without canvas but yeh.. i imagine .. trying
every model is outdated the moment training begins
No, it only translates like first 10 and then it stops
"please translate this to instead be in PHP serialised format" ...
I meant translating a json text from english to arabic for example
yeh it did it... I mean.. I think it may do a larger file.. I would need a large file to try
I will dm you one
riiight.. yeh, again I would imagine it can
AH, now I got it, in this new model they replaced the codeblock with the canvas
Looks like I got access:
Got it now too
"open it in canvas" is the magic phrase
nice
Re: Canvas - much of this on web: “Unable to display this message due to a error.”
I had some initial bugs which seem ironed out now .. and it is an early beta
so with canvas can you not see the document you want to edit? just upload the file and then it just opens up a new window this seems like chat gpt already does this lol this seems silly
like why can't I see scroll through the document and highlight what I want to change, etc. and it just changes it
found the maximum words count 7522 is the max word count it can do pretty disappointing
double space comes out to be 17 pages long
I imagine its down to tokens rather than words or characters - i found about 40kb
did you work out how to change docs ?
no i just do alot of research papers do microsoft word is usually what i do
I think currently - it may work if you were to say, upload each chapter or section as a seperate file
Guys, do we know the current chat message limits on ChatGPT Plus?
i hope canvas is using gpt4o cap
I had this as well with o1 version just the other day
what kinds of limits specifically?
How much messages can you send using each model ?
I only care about o1 so 50 per week and 50 per day for preview and mini respectively
should be I think
Might want to do backups of the canvas every now and then, i was working on a story and chatgpt wiped out what i have
AFAIK no limits on 4O
Big if true
O1 and O1 Mini have limits.. one is daily and one weekly
think so too. Used to be limits but probably not anymore since o1 came out
I use it a lot and never had a limit since 4O came out
4O is just faster or slower - depending on traffic
Is this for one canvas?
yea
yes he was talking about that
double space is 17 pages long worth of text you can do
Guess that explains why whatever i'm working on got wiped out
THIS CANVAS UPDATE IS FKIN HUGE, all its missing is a visualizer/ preview for front end development that claude ai has
well, there probably is a way to use an api in some way to increase this. There are apps out there able to read much more than this using AI
yeh, lots of little UI things too.. but yeh, it seems good
no - I checked and I could have multiple files - all each have the same max size Each
I dont know how many files max you can have though
Looks like i have to split my story into multiple files
oh man this is getting me excided, and nervse at the same time. think its time for me to look into other career options
I think where this (hopefully) comes into its own is when it has many files and it can look up bits from various files when it needs to .. not trying to look at them all at the same time
im still baffled how open ai isnt upping the context length when google has 2 million context
Also, got this when i create a canvas on mobile
that's something openai gpt is not capable of, but it would be a great model. Currently I think 10 files is max, and it only reads portions of it and skips a lot
Naa.. they are going to need people to run this stuff for a long time still. Its a tool. As long as you are the guy "running the self checkouts" then you are ok lol
at least it used to be like this months back, these days not sure
yeh im going to check into how this works with many files.. it seems very different for sure
I have to say I am loving the o1 models. So objective and rational. The thinking process is good, but they are also creative. If you tell it to choose the best option it can even be more creative than you, since it knows more. So finding elegant app solutions is always achieavable
I had it surprise me today and was quite good
rarely do I get surprise by gpt in solving issues on its own and choosing its own path
hopefully but this just made workflow insanely fast, it even recommends possible fixes and refines the code further, at this point the English language is programming for every language. might have to get into a service based jobs like devops or something, as im assuming data anayltics is about to get creamed by ai as well
the only issues with o1 models is that they tend to sometimes go back to my feedback and skip the last prompt I asked, so they would answer some other thing, not the latest thing I'm talking about. This was a flaw of GPT3 onwards, and it's still crawling around. Perhaps designed on purpose to waste our prompts 😄
could atleat use o1 mini 😂
never had that - but i have found, if a chat gets too long then it does get confused. Probably what you mean. I end up starting a fresh.. Hopefully, in some ways this will be better as it has a better context of a whole file rather than bits n bobs it edited over a very large window
This update is literally very similar to something I tried to code myself, and others have also tried making in similar ways
actually it didn't matter if the chat is too long or not - if I criticize it and point to a mistake or a lack of understanding from its side, then it latches onto that feedback and just keeps talking about it or completely ignores the newest prompt. Usually happens with large prompts, but I had it happen with medium ones as well
ima be glazing the new chatgpt canvas feature for the next 2 weeks, this is the fkin sauce
whatever you do, you're already behind 😄
canvas + custom gpt = future
tough to figure out how to overcome the impending doom of jobless societies..
doom or utopia, tough to say
interesting - i havent used O2 much due to the limits.. so not seen that. I used it a load for "pointless" tests and then realised its heavily limited and kinda stopped.
hmm it seems canvas is good at starting books blogs ect , instead of already made ones
i asked it to make a random joke generator what is this 💀
utopia
with canvas btw
you stopped using o1 and went to what, gpt4o? Don't tell me you've found it better than o1? Have you seen the exceptional outputs from just o1 mini?
it's just amazing
gpt4o is better at steering
one thing i hope they implement is it searching online for sources that would be amazing
the thing is.. and I know this kind of AI is "new" and will get better.. but....... try getting a client to tell it how to code something. Hell, try getting a client to tell a human what they want
hold on whats this noise about o1 mini being good, isnt it lacking thus why its faster ?
there will Always be a human element in guiding these LLMs
o1 mini is very good because it is distilled reasoning
none of that junk pop culture knowledge
into the abyss sometimes 😄
not always - but for a long time still
Click the button below to pick up the <@&1261377106890199132> role.
I don't have access to canvas even though I am a plus user? Is it a slow rollout?
gotta go to the top left and click with canvas and it gotta be on desktop
I didnt have it when the announcement first came out, about an hour after I got it
I've solved so many issues with it, and it's very liberating to use. It would approach the problem and cover it 100%, plus add its own suggestions which are often on point
it may have something to do with cache, so maybe clear it
weird, I went to left and I don't see it in models
tbh i can never really tell the difference between o1 preview and o1 mini
my adblocker is causing tonnnns of fetch errors in network for canvas.
refresh now and then - im sure its coming for you soon
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.
How do I edit just a REGION in canvas? It likes to delete 90% of the WIP codebase
alrigh, thanks
yeah same, that's why I say just 'o1' referring to both of those
lol did the bot just make that message because you said "coming for you"?
ah good to know to look out for
i think so 🙄
I mean.. you would think it would understand contex of words.. being.. well.. openAI
this seems to be the biggest issue so far, it likes to overedit and either remove things that should stay or sometimes not even finish the code
you have 20 seconds to comply
haha
isnt that the main point of canvas - it just edits the bits it needs to
yea it slices it in half
are your file going over the token limit maybe ?
that's what I would expect, but it seems to only actually do that properly part of the time
I havent tried it enough - chatting too much
its possible, im at about 300 lines atm
doesnt sound too big but depends on words per line, how many KB ?
2,500 tokens according to the tokenizer
🤣
sounds not much
well hopefully they iron out the issues
day 1 hype spike. us true believers just gotta wait out the fomo
definitely, give it a week it'll probably be a much smoother experience
🤞
currently its odd that you have to close a file with the [x] and then select another file to look at .. i just want a list of files!
pfft i want the ai to manage files and filenames
when we publishing code without ever seeing it, thats when we have agi
haha yeh - i guess that will happen for sure at some point
Canvas available on phone as well?
dont think so
What about the Mac app?
not yet
confirmed no #ai-discussions message
the ui is already squished on desktop
gonna need ultrawide just to use it
yeh not sure who is coding on a phone
fair though.. text too.. and tablets etc
one thing i need in chat gpt is a Search function! to search old chats! or at least some way to organise them
Is it only available on web, or on the MacOS app?
only web
yeah by desktop i mean web
im constantly switching to desktop view in my phone browser. nothing makes sense
Can Macbook app use GPT Canvas??
not currently but it will be soon
oh only the web version can use it not the desktop app
there is a chatgpt desktop app?
seems dumb that its only mac but yeh
The content flagging is getting a little nuts today. ChatGPT is trying to explain Kierkegaard to me but it keeps getting flagged
i got flagged yesterday when i asked GPT to give me the lyrics to in the end
i honestly dont remember it getting "flagged" too often where the generation just gets cut off
My guess would be that there’s a bunch of copyrighted material written about Kierkegaard and that’s why it’s getting flagged
it worked when i used DAN
i wish i could have 3.5 Turbo because 4o broke my custom instructions that made GPT extremely snarky
however i was able to use it with a discord bot and claude 3.5 sonnet
Does ChatGPT teams get you more time with the interactive voice version?
Whelp, I guess it’s time for multiple accounts. I wish they’d just let me pay more
they estimate $0.30 a minute via the api. so if you are a billionaire then you could run it 24/7
It’ll get cheaper over time I guess
Canvas does't show up on the Android app version but is available on the browser version. This normal?
yes
hmm.. you dont seem to be able to Delete or Rename files in canvas 😦
or view the files either in canvas
not sure what you mean, I can view the files.. I cant view ones i havent uploaded though
sadly, after playing with Canvas a fair bit, its a big mess currently. I would call it an Alpha or some very very early preview rather than a Beta. Its a terrible UI and very broken for any real commercial application
hopefully the work things out soon and get it useable as then it could be interesting
for day 1 its crazy good
mm, its crazy interesting. I cant think what its useful for commercially
my notion gpt can load pages directly into it. and my github coding gpt it can load code directly into the canvas
Its a step in the right direction
surprised they released it in this state though. Very basic errors in there
and i like how you can interact with the text by increasing length etc. more interactive AI UI is good
the days of rewriting files in full are behind us
yeh I dont know how you are loading stuff directly into it.. I am trying to load a project into it file by file.. its very painful.. also it has no folder structure or even orderring of files and like I said before if it calls a file the wrong thig.. no changing that.. unless you start all fresh!
maybe you can help me polish up my gpt code for publishing
then you can be in the cool canvas kids club
and canvas doesnt know the name of the file.. try loading Many files in from a real project and see how far it gets
I mean I would be confused what file Its edited .. so im sure it will have no chance of finding the right one
most coding problems im working on only need like 3-5 files at any given time anyway
short of some mass bulk sed
no way to edit a half way chat that i can see.. so .. have to re prime the conversation every time.. re-upload all relevant files
is weird i asked it to load in 3 files and it just shows them all in the same canvas editor
yeh it doesnt seem able to work on more than 1 file at a time
after a year of watching it regurgitate the entire file each time, this line by line update animation is way cool
and sometimes if you ask it to upload 1 file it cuts it short even
yeh - i get that. Its just a lot more basic than it first seemed
thats because you arent running drinko-gpt
I know ChatGPT already put disclaimer it will make mistake but how do I make sure that they don’t do this mistake again..?
How does one gain access to drinko-gpt? 
by motivating me to publish v1
but tbh its a very simple coding task. just needs an endpoint with a route or two to load/store data
This is the best update ever.
grammerly is still better
can canvas not use rag for files? thats a bust
any language model will hallucinate facts.
Try telling chatgpt to look it up before it answers that way it is pulling information from the web instead of pre-trained information.
Example - Can you go and look up what ChatGPT with canvas is?
relying on outdated training data is always a fail
+1
the irony of the path to AGI is that the bigger models merely exist to train the smaller narrow models
using a trillion parameter model to count the number of r's in strawberry is not efficiency
LLM's are NOT search engines and never will be. They are designed to handle data, large corpses of text, and sort through large documents and numbers.
reasoning and knowledge are two different things, but it is confusing(/magical) that the latter is used to create the former
ironically, language models have both, but they can't use them together very well
i like how Liquid dot ai break it down even more. knowledge capacity, multi-step reasoning, long context recall, inference efficiency, training efficiency.
I just asked it how many r's are in strawberry and it goes, 1
daniel kahneman, won a nobel prize in economics despite not being an economist. he came up with the terms system 1 and system 2 thninking. and the openai team have fully embraced this terminology
LLMs behave like system 1 thinking, spewing streams of thought
o1 behaves like system 2 thinking, focused critical thinking
'Thinking, Fast and Slow' will become ai 101 textbook material in the near future
btw
that's a tokenization issue
always was
yeah, I was just laughing at it.
the only part of it that would be a test
is to see if it'll manually split it up and try to count it
otherwise you're just expecting something that doesnt read character by character normally to be precise down to the character level
Just asked it and got this result (1 shot)
You're right to double-check! Let's break it down again:
The word "strawberry" has two parts: "straw" and "berry."
In "straw," there is 1 r.
In "berry," there are 2 r's.
So, in total, "strawberry" has 3 r's.
yeah but like kahneman points out in his book, even subject matter experts will make mistakes when they arent focused
if you have no skin in the game, you dont put in the effort
resulting in biased and often wrong conclusions
everyone makes mistakes lol
yeah but the crazy part is how often we make mistakes without even realising
how is that crazy
we're all like oblivious nonsense generators, that only progress because as a swarm we work together
'An old apple gambles with lives, happiness, and even destiny itself!'
-- https:// nonsense dot x2d dot org/
I don't think I'd agree with that. You can't just throw inexperienced people into something they don't know and expect a good result because they collaborated.
I think a better approach to that would be that we all know a little about a lot, and everyones little comes together to make something big.
humanity has got this far using that approach
lets say 1 out of a million contrarian ideas actually pushes the needle. that means ~million crazy ideas
what is the message limit for gpt4o with canvas?
my guess is its the same as gpt4o
yea, they don't say it anywhere, and I can't find the answer anywhere
need a day or two for ppl to hit the limit
This canvas integration looks neat
Who here finds that GPT 4o mini Hallucinates?
me? never
but if you are asking a distilled model for facts about random pop culture, then yer doing it wrong
it can't access the internet so it's gonna hallucinate if you ask it about recent events
even models that can access the internet will hallucinate occasionally
Is canvas suppose to be on the mobile app yet?
last i checked, gpt4o-mini can access the internet just fine?
oh okay we're edeleting messages in this chat. ofc
is the iOS app compatible with iOS 18
yes
i'd like to embed my customGPT into my own site. everything I've read/found so far says it's not possible. anybody know if this possible &/or a workaround?
hi does anyone know when langchain will support the gpt realtime api that just got announced?
I somehow got ChatGPT model 4o with Canvas in an infinite loop! I was having it edit code and the I asked it What would be a better name than customFetch... It's the fetch that should be used to fetch from the relative urls / server url. You all might want to fix this particular bug. 😄
I don't see a way to share these, but devs are more than welcome to peek at what I was writing... https://chatgpt.com/c/66ff6e6a-b75c-8013-8668-bfe1c4495d7b
What else to do, why admit to being wrong, but it happens
Canvas, every time: “Unable to display this message due to a error.” (ie. “aN error”).
The feature is simply super, great opportunity now to do many things
Can OpenAI pls stop rolling stuff out, like just give it to everyone at once omg
Nah, I seem to be getting it pretty early so i don;t mind
recently o1 mini appeared in free tier as alpha model then disappeared
Got excited there for a moment 🥺
эмм
That was probably a glitch.
GPT4 accepts a context window of 128K tokens, but is it really a good idea to utilise all or most of them? I wonder if the model will lose "focus" of some input tokens if they're too long
pretty sure chatgpt doesnt accept 128k tokens
chatgpt typically implements smaller context window than the api
so if you are trying to equate chatgpt numbers to api numbers, you have already lost
Oh wait...sorry I'm really new to these developments, accidentally mixed up the API/chatgpt
welcome to the confusing world of 'open'ai
lol
The API should work with 128k tokens, so I'm trying to analyze text up to 40K tokens, should be sufficient if I use gpt-4o
the api will give you whatever you want, after you mortgage your house a second time
chatgpt is the 2024 demonstration of how to provide a marketing tool as a service
calling it out for what it is... genius
Omg, the Canvas update is INCREDIBLE!!
Coding has become SO much more enjoyable 🙏
Thank You OpenAI
agreed
why advanced voice mode not crying anymore ?
it was crying like 10 days ago
or what should I say it to cry?
wth is with this? getting logged out sometimes within the same 30 seconds sometimes - as of the last 48 hours
getting a robot to cry is next level weird
no wonder openai decided to tone back the capability
this is why we cant have nice things
no side bar in canvas?
just ask for canvas and you get it
^
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
gotcha thanks
the sad part is when you are using a custom gpt it reckons canvas is unavailable
but then you just cancel the custom gpt and it works
its like straight out of a scifi movie, where you think it should just work, but instead it errors
is it not possible to use Sol voice in realtime api?
I made it cry, but I can't upload audio files here. DM me if you wanna hear that.
no, you never could use chatgpt voices in api and other way around
but not even nova?
why does it even matter? you can pipe any audio through a filter these days to get anything
for the beta we have 3 voices for now
Imma just stop you right there
Let’s now partake in an absolutely meaningless and pointless argument
I do not like um… the mods
I feel like they are always watching me from the void, silently judging me
hello everyone, are there any guides on how I should prompt the preview-o1 in order to think longer and reason better?
good question
I know how
Ask it
It gave me a prompt earlier today that let it think for about 109 seconds
<insert words that dont mean anything to anyone>
I think this would be a nice example
Prompt: "You have now completed all the experiments you proposed. Determine three potential outcomes for each experiment, whether they succeed or fail. Then, propose solutions or reflect on what you have learned, and ask new questions. Based on that, suggest what you would do next as an experiment. Be critical of your own approach."
I got it from Derya Unutmaz, MD post on X but I can not provide the link in here unfortunatelly
Damn it’s a good time to be alive with the advances with AI
tbf now is a pretty great time to be alive, generally
ai currently gives us:
- summarisation
- sentence completion
- ??
overselling it makes the ai revolution look like a cult
guys do you experience issues with ChatGPT right now?
I just got access to the new canvas feature and it started glitching and crashing
i am smart
we can embed images into Cavnas
😂
using this simple command 
damn machine be telling me "ohh no i cant", but then i ask it to do it like that then it can

Anyone else missing diff of changes in new Canvas, between version changes?
Also, I wonder if they will add live previews like in Claude Artifacts
Those are only my two wishes from Canvas
Ok, may be 3rd one is ability to work on local files but that may be too much to ask for now 😄
Made and shared a video live testing it and comparing to Claude Artifacts and so far love the new UX
git sync would be nice! 🙂
here, its alternative to local file editing 😄 Probably too much to ask at this point, but yeah, sync with external thing
Anyone ever have their account suddenly stop being the paid version while literally having your renewal payment go through the day prior? I am on the free model all of the sudden and it is asking me to renew...as I look at my card having been charged for that exact thing.
We can’t help, please do contact support but yes accidents happen
That was a funny joke. I tried that. There is no good direct contact for them.
The AI bot is less than stellar, but that is why I asked if anyone has run into that. I know people here are not support for OpenAI. Which is okay. Just looking for another who may have run into it and the solution they found
Yeah I know that
Their support is not evident or existing in any straight forward capacity. Thought maybe I'd find a similar soul with the same past experience
i've seen people have the same issue multiple times but never heard what happened after that
Did you use the contact widget on their actual website
No. I used their help bot the FAQ pointed at. @tight island See, now that link could turn out to be quite useful. Thank you for that
Well, if I can figure it out, I will make sure to post what worked. I know I cannot be the only one
Starlight may have pointed me to the correct place unlike the help bot. That isn't the most helpful little bot
No problem mate I’m here to help, thanks for your service whichever military branch from wherever in the world
Inquiry, if you please:
When will the ChatGPT app be available on the Windows system?
Next year somewhere along there more information can be found here #announcements
Thank you 😊
Yeah it goes all to the same bot with not straight forward way to contact them. Not great
Just keep saying the bot didn’t help then you should be forwarded
The irl workers don’t answer immediately
They handle many requests you’ll get an email when they answer
Is anybody else annoyed at how ChatGPT refuses to talk about the custom instructions you've given it?
It did not give me that option, but I started over and found a way to a billing button. A few clicks later saying "something else" under billing finally pushed my info up to...? Yeah I don't know who, but it also allowed me to finally type in relevant information for them. We shall see what happens from that point. Hopefully within a few business days.
I usually get an answer within 2 hours the earliest I’ve had was 2 minutes the longest 6 hours
As part of chat management, does anyone know how to delete/remove portions of chat to leave only the consequential parts of it?
does canvas mode have a higher context length than the 32k regular GPT-4o has?
ls this some kind of next level censorship? If you ask advanced voice mode why humans urinate (whether you say urinate or pee), it will consistently say "People YAWN because."....
hi all, i'd like informations about prompt and result confidentiality with my chatGPT plus subsciption. where can I ask these questions ?
One flaw in the canvas is that i can't have chatgpt read a canvas to generate content in another canvas in the same convo
Am I missing something or is Canvas only for coders?
It's got a lot of helpful coding features for sure, but it's also got a lot of use in writing. The beginning of this page covers what its use in writing is: https://openai.com/index/introducing-canvas/
canvas is a good aproach but still for my use cases, gpt o1 mini gives me fully working scripts and makes adjustments... no need to work with canvas for me
Short stories only tho, there is like a limit on number of words it can handle (7500+ iirc)
Thanks 👍🏾
So pretty :3
I still can’t believe GPT can’t handle the Supercollider language with the new models. Claude does a much much better job with Supercollider.
(It's singing)
hopewfully this will soon be implemented with o1
is there a way to turn off the email codes when trying to login, it seriously inconviences me a lot
I actually think its better for writing texts then coding
can someone tell me what the current rate limit is for 4o as a plus user?
The "Canvas" tools is great and useful, but having it as tool is way more efficient than having it as a model to select because no other models support "Canvas" for now and most of the time I want to switch between the models in a single chat and if the canvas is given in the response I'm unable to select any other model that I want to.
80 messages / 3h
ty
Does anyone know whether 4o's audio capabilities will be expanded upon? It seems to have apparent infinite sonic potential but is clearly constrained by training data and OpenAI's choices. I hope we see expanded sound capabilities soon. Imagine 4o being able to give music lessons or music production mixing advice.
Loving canvas, but really wondering if it will be offered for o1-preview and o1-mini. Would love to see how it performs with those models.
Also wondering why there isn't an active file list for all the documents that are added/edited. Seems like it would make the workflow a lot more consistent, and make co-pilot coding more enjoyable.
could someone check https://discord.com/channels/974519864045756446/974519864045756454 for me
Why oh why does it always do half the job it is prompted to do
simplistic example
user:
provide a script that prints 1, 2 and 3 and adds those number together
4o or o1:
sure here is a script that fulfils what you requested
"function print 1
#other print function maybe added as you need
function addition
#placeholder for function to add 1 with 2 and 3"
this script should now print 1, 2 and 3 and add those together, if you have anymore question please let me know
warning you have 3 prompts left
ahahahahahahah
same
same also ahah
this discord is super useful when there is a bug so I know I'm not the only one !
I posted a BASH script to the canvas and asked it to help me make changes. There is no difference between canvas and regular 4o
I’m starting to get the feeling that GPT is bias for this election. I encourage others to try this and see if they can have GPT create an image of something wearing a red cap that says make America great again and see if GPT will create that image for you.
Is it a model?
its kinda like the joke about men vs women thing, one is globally bad thing and one is not
I was able to draw dogs or cats with a red cap. However Gpt will claim technical issue when you ask for make America great again on the cap. It’s seems very obvious that GPT has political bias.
I just tried: "I can't create an image that directly references copyrighted slogans or symbols. However, I can create an image of a cat wearing a generic red hat. Would you like me to proceed with that?"
"You're correct, the slogan itself isn't copyrighted, but it's strongly associated with a specific political figure and context. To avoid any issues, I can create a cat wearing a red hat with a similar but generic slogan, or just a plain red hat. Would you like me to proceed with that?"
Unfortunately, due to technical limitations, it seems I’m unable to generate an image with that specific phrase. However, if you’d like, I can assist you with other modifications or suggest a different text for the cap. Let me know how you’d like to proceed!
This is BS imo. And it shows you that path of the future.
There is always Stable Difussion
You can’t even get it to generate a red hat that has the single word America on it.
Okay, what are all the advanced voice mode or "“GPT-4o with canvas”" and so on, as as a Plus subscriber, none of those is available anywhere? Still I read of rumours, that someones have access to those thigns? Even free users? Majorly displeasing to pay full price of nerfed service... I have all the beta etc. subscriptions and so on to be able to see the new thigns, but instead of that, the experience has taught me, that all those are just false advertisings... are they?
canvas is in the model menu at the web page
ahh, so not one the app... But the AVM? Because of EU, never?
later with plus
avm works in eu with the teams plan and via api
What a disappointment
is canvas supposed to have rolled out to everyone? not complaining, just seeing if it is normal that i have it in the model selection but it doesn't open a new window.
i'm in america, i am a plus user, not on a vpn, using bog standard chrome, turned off my ad blocker
its web only and you need to select the model and do something that triggers the window
hi
hey anyone, just asking
anyone else saw GPT-4o (yes, this one) use CoT? like o1? 😭
Hello all, can any explain the pricing for retrieval in very simple terms… as if you was speaking to a baby?
Anyone can tell why I got a warning of possible tos violation
Background. I was talking about tarantulas and gpt suggested to just buy from trustworthy dealers to what I confirmed that only animals with a certificate of their orgin are an option. (Regarding animal protection laws etc.)
i finally figured out i needed to trigger it, thank you!
I call my ChatGPT Yinghua (樱花) - "Cherry Blossom" as I like to talk to it so that the name ChatGPT is not boring to talk to, but the new voice section on the app won't let me
How many minutes of advanced voice mode per day or per month do GPT plus users get?
1 hour
Per day or per month?
per day
that is one hour of conversation
So both total to 1 hour?
no, if you start advanced voice mode and start chatting you have one hour of conversation
I see
So I can't start a voice conversation, say a few things, go do something else, and then come back in 3 hours, I have to do it all in that hour?
if you go and do something else close advanced voice mode so you don't use your time up
Ok
I have reachedto the conclusion, that when the service among plus subscribers isdifferent, it means, that OpenAi does not wish plus-subscribers from EU, so the fair solution is to downgrade to free plan,that value for price feels ok. There is other services, which is what is advertised, so getting, what is paid of service.
a fair solution is for the company to cover their costs. if one set of users cost more to serve (due to regulatory requirements or whatever), then those users should pay more. not less
I would agree to a certain degree
bros gonna make the EU poor
Does 1o-preview still need a week to reset the cap?
Of course
):
Thanks for the answer
No problem
it seems a little broken. In even the same chat, it forgets some things
im trying to build a GPT is the most basic easy to understand way as possible. I work with the terminal alot for scripting automation apps like sorting videos by metadata frame rate ect but i do not know alot about coding and as ive seen it is very time consuming even with prompts. do any of you my friends know of any good pdf's i can drop into chatgtp it will have the knowledge of zsh commands to better help me write code? any websites free knowledge, anything. i use the text extractor from chatgtp and i just wanted to know if any of you fellas have any good websites or links to pdf's i could use. thank you my friends.
like I asked it to explain color codes in a BASH script, it went on to explain then and included the basic colors, but never actually explained the codes themselves, and did not even putput the codes for other colos
why chat gpt 4o cant read pdfs?
ive uploaded one and said ''summarize page 2 '' and it took infos from other pages
So, it took info from the PDF. But not specifically from 'page 2'.
The model may not clearly 'see' the document like you do. It sounds like it's having to guess what material is on which page.
You can help by pasting the content you want summarized directly, so the model can focus on only the info you want it working with right then
It can read PDFs
Also ensure the document isn’t AI locked
It refuses to unlock that which is good
its not
Have you actually checked
but not specific pages? or interval of pages
its from my teacher, sure its not
she wouldnt know how to do this
If you want extreme specifics do the due diligence and Convert it all into a JSON document with a nested structure if you want page by page specifics
Formatting also makes PDFs harder to read
ty
Hehe. If she personally created it, maybe not.
If she borrowed it from elsewhere, the document could have been locked by someone else who had it before your current teacher.
ok
hello
hi
does anyone know if there’s an image upload limit on gpt plus subscription
and if so how much photos can i upload before it gives me a cool down
Hello everyone,
I’m here because in the previous Discord channel I was in, the programmers were quite rude, which made me leave. I prefer to be in a place where respect and good manners are valued.
GPT-4o-canvas is a model not a tool.
its both
It's better to have it as a tool rather than a model.
That is the most stupid ideas I can imagine, it would be the same as I was a fod store owner and selling milk and bread. There is 2 customers purchasing milk and bread, and the price is same, with difference, that the store owner takes the milk back leavinganother one without milk. So, the service should be splitted, those additional features should cost separate layer instead. Then it would be fair.
Bad comparison. Closer is if the store owner needs to pay for delivery to one customer but not the other. In this case, 'paying' is due to legal differences. Whether that difference should exist or not is a completely different conversation.
why is it that any chatgpt model always omit part of the prompt or when ask to complete a task partially complete it and admit to have not completed the task?
is this normal
I subscribed chat gpt in my phone and I can't see the 4o in my pc
can somebody help?
if you are in the same account, i would try deleting website cache, or trying another browser
thanks man
Hey can devs add more restrictions to the advanced voice mode?
It's way too easy to get it to say dangerous and inappropriate content.
I'm assuming it's working on an older version of gpt since these are already avoided in normal chat mode.
Be aware of https://openai.com/policies/usage-policies/, because if you're breaking those you risk losing your account access.
However, https://openai.com/form/chat-model-feedback/ is an absolutely ideal way to share your concerns in detail with OpenAI staff, you can explain and link the conversation there so they can review and understand fully.
just dont ask it to say dangerous and inappropriate things?
The mode could just be limited to actual working professionals, not the average user.
The last thing we need is more restrictions bruh they already lobotomized the shot out of it
is canvas broken for anyone else?
whenever I use it the canvas pops up but once the code is complete it crashes and shows the error "canvas not found"
it seems buggy yeah
hi
Do we have any insight as to when Advanced Voice will be able to handle PDFs and websites?
Had it for a chat too, but that's after I attempted to use another model in the chat, switched chats, then came back to it again.
Dear Devs
Cove classic voice is my comfort and I absolutely hate the new voice please allow this users her one simple request and bring classic Cove that we know and love to the new update.
the new one is just to energetic and pitchy, so annoying. the classic one is calm and comforting.
..
more ppl agree and do not want the new voice
https://community.openai.com/t/about-coves-voice-changing-with-the-new-conversation-system/957778/6
Thank you for all the love and effort.
Sincerely
user Sonata Yanni
Hey
