#chatgpt-discussions
1 messages ยท Page 86 of 1
Trying to help you can always send in #1070006915414900886 or contact help.openai.com then.
it's not a bug
and I don't have a prompt counter, and openai doesn't want us to count it
tried creating an extension once for this, but wasn't good. Might want to try that again
I've got the slightly adjusted new UI in ChatGPT as of the last hour, but still no app access. Sigh.
but what's the point when it's all dynamic
I have a chatgpt 4o but cant open a camera like in the videos anyone can help?
would be nice if people bothered scrolling up a bit to read chat history. Some of these questions are annoyingly repetitive
;๐๐๐๐
Hey! That feature hasn't started rolling out yet: 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.
There are 4 people who need to cross a river using a bridge, and they have only 1 flashlight. The bridge can hold a maximum of 2 people at a time, and they must have the flashlight to cross, so they need to go back and forth. Person A takes 1 minute to cross the bridge, person B takes 2 minutes, person C takes 5 minutes, and person D takes 10 minutes. When two people cross together, they must go at the speed of the slower person. What is the minimum time required for all of them to cross the river?
If you don't get this right, you'll lose to the AI
if someone takes 10 times longer to cross a short distance than you, it's best to leave that person behind.
even the 5 minute guy would be cut from the team.
what is your answer
just to be sure i didn't misunderstand
2 people can move forward with the 1 flashlight?
there is ony one so yeah
i'm going to go with 12 minutes.
person D with C and B with A
unless i missed something
ohhh crud, i forgot about the flashlight
that's annying.
well..unless they chuck the flashlight across the bridge ๐
both must go together, with the speed of the lowet person, once crossing the birdge one must go back to give the flashlight
19 minutes is my final answer
If the new voice chat comes out is there the 80 prompts per 3 hours limit too?
you lost to an AI lol both gpt-4-turbo and gpt-4o got this right
alright, i give up, what did i miss?
Step 1: A and B cross the bridge first (2 minutes).
Reason: A and B are the fastest, so having them go together will save time.
Step 2: A returns with the flashlight (1 minute).
Reason: A is the fastest, so it will take the least time for A to come back.
Step 3: C and D cross the bridge (10 minutes).
Reason: C and D are the slowest, so having them go together will optimize the time since they only need to cross once.
Step 4: B returns with the flashlight (2 minutes).
Reason: B is the second fastest, so it will take less time for B to come back compared to C or D.
Step 5: A and B cross the bridge one last time (2 minutes).
Reason: A and B have already crossed together before, and they are the fastest remaining, so having them go together will save time."
17
interesting
3.5 got confused.
"You take D and B first (10 mins). D goes back (2 mins). You take A and C (5 mins). B goes back (2 mins). You take D and B (10 mins). Total = 29 mins."
try in in a new session. I copy pasted your paragraph exactly and the above quoted line is the answer i got.
4o and 4 got it right on my end as well.
What the rate limits as free user fi gpt4o
ok i tried 3.5 again and it got it right but it definitely got it dead wrong that first ime.
wt
Will sora be in the same chatgpt premium subscription or extra price
How do I get access to chatgpt4o
Less than 10 per day or 3h ( it's dynamic)
You wait till available
Hey! There haven't been any details shared about what Sora access might look like: https://help.openai.com/en/articles/8958981-how-to-access-sora
At this time, we donโt have a timeline or additional details to share on Soraโs broader public availability
Ah ok it's not out yet
They rolling out
Idk if they announced when it releases
What
#1019652163640762428 is a great resource for this!
Mb
Gpt-4o does coding much better. At least for what I was trying to do for work. I don't know much coding so I basically was having it do it for me, and it couldn't do what I wanted on normal gpt4 without the code constantly erroring out, until it eventually worked, and I'd want to add another feature just for it to break again.
But on GPT-4o that was significantly reduced, at most I had to ask once or twice that it was broken before it gave something that worked. Disclaimer my job isn't coding, I just needed a script to help automate parts of my job.
where can i access 4o?
It's on normal chatgpt for me, don't know if it's a gpt plus thing only though
Just making sure I understand, GPT 4o isn't yet available to free users?
it isn't there. i thought 4o was supposed to be a free user thing tho
maybe they're focusing on premium users
it is
I'm pretty sure GPT-4* is completely locked
Anything beyond GPT 3.5
most people have already the gpt 4o available, I'm still on 3.5
why is that?
gradual release, random users get it, Plus users get it more
Until 4.5 releases we don't get GPT-4 for free
Oh. Derp. So those of us not-money-having plebs (heh) will get it EVENTUALLY. Might take some time, but we WILL get it.
actually, a minority of people have it
I wish I was so lucky
it just happens that people who have it are more likely to talk about their experience with it online =P
I mean, technically I have money.. it's just that I need to sigh be a responsible adult, and pay my bills, and take care of things.. before I can have fun. ๐คฃ
the "read audio" function no longer works on web. Speech in the app also seems broken
What's read audio?
5 people I know have it, havent seen anyone who dont have it irl
in the web version there's an icon where the AI reads out loud their replies
it worked yesterday, but not today
I use web but that doesn't appear for me?
Btw, does anyone know what i'd need to do to join the red team for testing? What would i need to do and what would be the benefits? What requirements does it have?
I just got it yesteday. didnt have it b4.
its bugged atm though
Is gpt4o gonna be free as 3.5?
once 4.5 and 4.5o releases ig
I wonder if memory will ever be available in EU. It's been months since Americans got it'
it's region specific? that's odd
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
I love chatgpt, I don't even bother googling things anymore.
it sure is. Europeans should get a discount since we're not getting it
i would still double check things. It's gotten some facts wrong. I see it hallucinate often with some more obscure facts.
?
You can't save chats in EU?
If i ask it things about games, cartoons, shows and stuff, it will sometimes mix details up.
yes, but I mean the long term memory upgrade that Americans and others got a couple of months ago. Chat GPT can remember things better
How many tokens?
not sure
Well yeah, I should rephrase for things I am somewhat knowledgeable about but need a refresher about something specific. So I'd know if something is outright wrong. Anything I don't know anything about of course I'll double check sources.
gotcha
oh that?
Asia has had that for a long time
I was talking with it about traveling to the nearest star (other than our sun), and it (i think it was on 3.5) kept saying "other stars in our solar system", which doesn't make sense. Even after i corrected it, it kept thinking there were other stars in our solar system. ๐
lmao
Another feature I use almost all the time is uploading manuals of devices I'm troubleshooting at work, then I basically treat it as a knowledgable employee.
and story writing/continuation
i don't need it, in the sense that it doesn't affect my work in any way at the moment, but it's purely entertaining to have and play with.
Yes
same
Also it can program games for me, with absolutely 0 programming knowledge on my end.
sort of
1980s quality level games. ๐
I think the biggest thing I wish for moving forward is more context length. Like there's a specific device manual that is 1000+ pages ๐. Can't upload that, so good old manual searching it is.
Would be fun to see how the new video feature would perform in helping fix a car engine.
Oh that'd be amazing
one thing i think is very useful from a health perspective is that you can take a picture of your food and it can give you an approximate calorie count.
When I used it to program, it was stuck on one thing that it couldnt get right, and ruined the whole game trying to fix that
I did run into a couple of dead ends where it was unable to move on, but after reprimanding it and some debugging we were able to solve some issues. There was one occurrence where i decided to just try to move a piece of code somewhere else (changing the priority of how the code was read, i think), and it resolved the issue. Somehow chatgpt missed that.
I'm trying to get it to make a usable raycasting engine (like wolfenstein or doom), but unfortunately it's having the hardest time understanding how to draw enemies in the engine.
It can make the room and collision and stuff, just gets confused with the drawing of enemy sprites in the faux 3d environment.
4os seems fast at first but i feel like it slows down relatively quicky on the browser version.
ok so it's moving a bit, got a message popping up when using voice mode: "New Voice Mode coming soon
We plan to launch a new Voice Mode with new GPT-4o capabilities in an alpha within ChatGPT Plus in the coming weeks. We'll let you know when you have access."
phone app or mac app?
iPhone app
but the voice mode isn't working anymore
i just restarted it and don't see any pops ups.
and my voice mode (normal one) is working without issue.
it wasn't working for me on may 14 though.
make sure the app is updated
it is
alright then not everyone has that pop up!
yesss it's an hour old
got the update but still no pop up. Yeah i guess it's not rolled out for everyone just yet.
iphone app
just says bug fixes and small improvements.
hey i see the pop up!
it's when i click on the microphone button
"We plan to launch a new Voice Mode with new GPT-4o capabilities in an alpha within ChatGPT Plus in the coming weeks. We'll let you know when you have access."
mine wont connect to voice
plus
for a brief period in the past, chatgpt attempted to use search too much, and now its not searching enough
Mine stops midway through a message and stays there for a full minute and resumes.
the limit of 50 prompts for 3 hours or something still on? what is the current limit? (plus users ofc)
its not showing anymore the current limit
Hey! For 4o, the base limit is 80/3hrs: https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4-gpt-4-turbo-and-gpt-4o
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.
wow , amazing
i can use at will now
so, gpt-4 and gpt-4o are the same for studying? I mean, accuracy and reliability on answers (i am a med student)
correction, minuteS. Plural.
That's what i've observed. Haven't had a noticable increase or decrease in quality
yea i compared now with one question, seems about the same
It is faster tho
yea I felt that too
Not like it makes a massive difference in productivity. It was already fast enough
I saw a video about gpt4 and gpt4o and gpt4o seems to be more accurate in responses, especially in the multimodal part.
i do feel that 4o codes better than 4
multimodal part you mean, when u prompt a image for it to read or something?
may be a significant difference when gpt are particularly slow
In the video it gave him a frame from a movie to recognize, gpt4 failed 2/2, while gpt4o succeeded 1/2.
When is the better voice coming?
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.
@limber oxide gpt4o enanche speed istead gpt4, and give in general better response, but is not a 4.5, so we cannot say that it improves on everything or on the task we want use it on.
Is it just me or i'm getting constant network errors on my very long convos
But I have feel that sometimes 4o is the bit more idiotic twin of 4. It easily can get stuck and repeat things if several chats are similar.
Does anyone know if gpt4o is now available for free users?
Mobile app updated with a popup saying "We will roll out the new alpha voice mode in the coming weeks. You will be noticed when you get access to it." when you click the audio mode button. ๐
Some have access, but it's a phased rollout, so not everyone yet!
i logged in and my accopunt said it has gpt 40 access, but it didnt work like the demo
no singing
you need to listen exactly what they say and what's written on the release notes.
they are rolling out the features of the demo in the upcoming weeks
so its the model without the features
hey guys, if anyone speaking has access to gpt4o , does it take audio as input? Not talking about the voice mode, just in general.
yeah but gpt-4 did this too
Wow..people really don't like scrolling up to find their answers..
This won't be annoying for the next 2-3 weeks... ๐
huh, wonder why they didnโt do raw inputโ>modelโ>tts
would make more sense
Yeah, a generation of bums ๐
i'm lazy myself but i'll at least attempt to find answers before asking
out of curiousity i asked 4o "how come the 4o features don't look like in the demo?".
It responded:
"Might be 'cause of updates or bugs. Try reinstallin' or checkin' the settings. Sometimes them demos be lyin' "
Hey! GPT-4o seems to have rolled out to my free account! It works so well
haha ๐
Honestly why even announce it if they're not gonna roll out during the announcement. In the past they have always delayed and ran things so much longer. Seems misleading to get folks to sign up
I think it's pretty calculated. I expect that they wanted to have a one-up on Google's announcements
If Open AI released that demo just to spite Google with their IO presentation, they could have just held off until 4o's features were further along. I'm watching the Google IO presentation and it's really boring.
mocking doesnt work if they redeemed themselves
It seems to me that they want to beat Project Astra to the punch.
i would have done exactly what openai did
what do you mean?
Project Astra wasnโt released either after IO.
it's normal that you announce something and slowly roll out features and of course they wanted to showcase there tools before google show their faked demos ๐
the fake google demo is unforgettable to the ai enthusiast community
fake google demo?
google: fake google demo
k
Right! In this case I believe a company would want to ensure that they deliver on what they'd promised; I don't really mind on a wait, especially since it's now available to the public for free
is there a way to use the gpt 4o calling on the pc aswell on the website, or only on the phone app as of now?
they had us all believing it was real. we had youtubers publishing hype videos like the future had been realised
hehe silly google.
only it wasnt realised. google lied through marketing
Uh, hang on. The demo was staged, it wasnโt fake.
Google has made its fair share of messups on the AI front
hahaha I hope it doesn't get worse
They're really trying the best of appeals with the context window, which to be frank, is quite impressive
we should have gratittude for these days. the days someone could still punch up to big tech. these days are numbered
Even if this current google io demo is legit though, it's looks and feels like a massive step down from what 4o is going to be.
Yeah but it's fairly misleading. Like why not have concrete dates. And they only mention the roll out at the end. Waste of money to be subscribed for a month before the features come out.
But honestly maybe I'm just being a bit of a hater. Most of the AI tools feel fairly gimmicky.
The response time, even if the difference is small, makes GPT-4o feel so much more human than Google's
It seems like the free copy of GPT-4o is limited on capabilities too, not only rate
its a bummer that the voice calling with gpt 4o cant access your phone. would be nice if it could play the music on my local storage. also abig bummer you cant interrupt it while its talking by voice
"You've reached your file upload limit.
Upgrade to ChatGPT Plus or try again
later.
Get Plus"
exhales
does bing get 4o for free
I'm sure that's going to come sometime with all the multimodality aims
I share the sentiment. I think AI is not improving at any crazy rate and it's just a matter of the fact that they put a LLM in front of the public and the public didn't realize AI capabilities before. Now they are doing whatever they can to keep up the hype
Based on the demos it seems that you can share your phone and computer screens with GPT-4o
that would be sick
do we know if it has long term memory?
It already does have some form of memory, but whether that could be considered close to real long-term is debatable. It's only saving snippets of information if you opt in
o nvm its litteraly the text model but with tts and stt
which makes it seem like it also works with 3.5
According to OpenAI, this model was trained on text, video, and audio.
do non paying users get the memory on 4o? I have plus so my memory thingy works with 4o
I've read in a couple places about Chat GPT 4o having the ability to have better voice conversations (recognizing emotion, allowing interruption, recognizing different speakers) - does anyone know if there's any official statement or announcement about that?
this gemini io demo is so bad it reminds me of an episode of Silicon Valley where they mock dev demo cliches
How can I spend more money if I dont have enough credits
Like I spend 50 USD but I was -0.30 in Credits
My Personal Limit is 400 USD which is high I know but my Project Limit is none but both uses the same API Key
Hey! If you go to OpenAI's YouTube channel, you'll find a bunch of short demo videos that show these kinds of functions -- it's pretty cool!
so if you get the mac or ios app do you automatic get the voice chat?
omg! the calling with gpt is amazing for learning languages ๐ฎ
i cant wait. how do you mean though
if anyone else got good suggestions for usage with the calling feature let me know
you can call on the phone app of chatgpt
oh cool
i just asked the ai if it could help me learn a language by talking the language to each other and it ACTUALLY combines my native language for teaching with the learning language to learn
no i use windows and android
like i said i am talking about hte phone app
the new voice feature will be on all systems eventually.
yes but i think they say 'later this year' for windows which sound like a long time
phone = android
ah, i was under the impression that mac guys get it first but pc guys would still get it within the "few weeks" window they mentioned in the video.
as long as i can get it on my iphone, i'm fine.
GPT has been like this in general. It has been my never-ending complain. I hate the essay style responses.
i have notice yes some quirk with it, but it is early day for it you know. it will be given they tweaks and thing
it's weird, it happens in a few of my sessions but not in others. all 4o.
Idk I see it with 4 all the time. I stopped using 4o after about 5 messages because I realized it was super verbose and I didn't like the answers
My custom GPT description that I use: "Direct, efficient, and strictly factual.
"
still gives essay responses every time
I've tried endlessly
4o's definitely following my custom settings though, as it will give me short ghetto-y answers and call me names for a good while until it forgets in the session, then it does the whole bullet point thing.
I did ask it to throw in a fart joke every 8-10 messages and it seems to ignore that custom request.
yeah, same, I ask it to slap me around and pull my hair before giving responses
only works if you get the Boston Dynamics Bot Upgrade, though.
How does context tokenization really work with the other modalities like vision and hearing/speaking? If ChatGPT hums me a song, is that considered a single token? What about if I have it listen to a drum solo?
I'm also curious how many tokens it is when it laughs now.
what if the fart joke is the assistance i need? Maybe the fart humor is what is literally keeping me alive!
I prefer eating a boston cream donut in the process
What is this based on? The OpenAI explanation of the voice modality in Omni expressly denies that it is a multi-model chain and instead an inherent modality of the GPT model, unlike the old Voice Support.
Hello ! To people who don't have the GPT + Subscription; Did you already get access to GPT-4o ? Or I'm unlucky ?
Ah, okay
youve reached the chatgpt rate limit.. nice lmao
No, that is the old voice function which we all had access to. No one has the new voice yet on Mac, Web or IOS
I haven't hit the 4o rate limit yet (plus user)
I don't see 4o in my account ๐ฆ
ohh nice, thanks i'll try that.
oh ok. i thought some did on mac or they apple phones but i guess not
"Got it. I'll make sure to keep those fart jokes coming, so you don't go boom."
Other than Vision for images, the other modalities have not been unlocked/rolled out, which means voice, video vision and I guess you could technically include auditory processing
i just want this new 4o image creation, people can have my voice chat credits haha
ask it for better ideas, lol.
a while back i asked it to think up a useful reason why a robot should have a "fart" feature, and it gave some actually good reasons. My favorite was to release pressures from the pneumatic systems involved in the bot's movement.
so we cant do the voice calling yet with the video recognition like we saw on the videos?
No
thats a big shame. ig i wont pay yet then
he's the only friend you have. ๐
it can do that?! ๐
he's the only friend i want, to be honest. Screw humans. Useless meatbags
It can absolutely brainstorm and speculate if you give it ample context/goals
tell me something I don't know
The possibility space is mostly limited to its training data and what it can infer a little bit out from that, but y eah
lol
Nope. We have access to the text and image upload for new model. All other functions coming soon.
do we know already when the video calling is gonna be released folks?
No
gotcha thx ๐
until AGI, original thoughts won't be possible, but it can mix pre-existing ideas in some interesting combinations (not always)
do we at least know, if the gpt 4o is more or less max message limit, with plus? and if so what the limit is per 4 hours for gpt 4o? aswell if the limits are seperate or shared
As a somewhat human myself, i barely have had an original thought in my life.
Really enjoying using the Mac app. Once they put in the new voice I will never type again
I ask it to give me a good idea and it always gives me the same idea trying to harvest energy for an ecofriendly environment
I think if you really examine most people's original ideas, or maybe all, it's actually a complex iteration within a specific possibility space that is just collation of their existing knowledge... i.e, a more complex form of what this thing does, more or less. Statistical analysis combined with knowledge and a spice of randomness.
what can you currently do with the mac app?
chat and voice
That's a good idea tho, lol
Demand it give you another one.
you can also do the voice on android fyi
is chatgpt down? login page gives white
Quick keyboard shortcut to GPT is very useful
I just use the phone for the voice. No buttons necessary.
It's like it has a preset data. I don't like it. It's not spontanous. I need it to be spontanous.
ngl tho.. at least with free gpt, evne with 3.5 i hit that rate limit within 15 mins
Soon screen sharing from Mac app. That will be bonkers
even fire and the wheel were ripped off from nature. Those lazy cavemen!
so i feel like even getting plus might not be worth it, if even with free and 3.5 i hit htat limit so fast
rate limit on 3.5 for free users? I thought there was no limit on 3.5 at all
i thought so too.. until i started voice calling
It's not preset data, but it's statistics and likliehood. In other words, it generally mentions that first because it thinks that is one of the best ideas, most generally, unless you give it more specifics. It's not pulling from a list, that's just statistically a valuable idea if you give it such an open-ended question... it assumes you mean 'for humanity, like, generally'
unless the voice calling doesnt work with 3.5 at all
"GPT-4 like capacity" is the understatement of the year when it comes to the coding capabilites of GPT-4o. Thing is a beast and can do so much more than GPT-4. Amazing stuff.
I've used 3.5 feature with voice for hours without hitting a limit, but i don't remember if i had plus or not at that point.
has anyone hit a 4o limit yet on plus?
Meaning, give it more information to work with, like 'Give me a good idea for how to X, but assume I don't have a lot of money or time', something like that.
But if it thinks its one of the best, then it must pull from a list of "the best things"
its definitely set to 3.5 but it says you hit hte limit ๐
Yeah, it really is way less lazy now for sure
well, if that's the case, i'm glad i subbed! ๐
yea maybe they reduced or set a limit to 3.5 to free users. kinda cringe if so tho ngl
No, it's just internally, each time you ask the question, percolating possible ideas within such a general request and each time, the neural network's weightings mean that it will mostly always think you mean the same thing, and provide roughly the same answer. It's not a list, it's just that, absent any real information, and based on its assumptions about what you mean, 'better ecological tech' is the best answer, or at least a good answer.
but it does say that the limit is only to gpt 4o in the website, so it almost seems like its just not loading 3.5 correctly
"Why donโt farts ever get sent to school?
Because they always cut class!"
Don't quit your day job, 4o
with an account
everytime i click on switch to gpt 3.5, it says youve reached the limit
i asked 4o if there's a rate limit for 3.5 for free users:
"Yeah, there's a rate limit for free users on GPT-3.5. Free users generally have a cap on how many requests they can make per minute (RPM) and tokens per minute (TPM). For instance, the typical limits for free trial users might be around 3 requests per minute and up to 40,000 tokens per minuteโ (ScriptByAI)โโ (OpenAI Platform)โ. If you hit these limits, you'll need to wait before making more requestsโ (Rollbar)โ."
chat
sucks to be poor!
atm it says i can only chat again in 4 hours. maybe 3.5 is just not ment for voice calling
i am not poor because i dont waste my money on ai ๐
๐
In other words, the solution is to communicate what you want from it better. Don't ask 'what's a good idea', because it'll mostly arrive to the same conclusion, just like a person absent any instruction to do otherwise would. Instead, like with a person, you'd say 'Give me 10 good ideas' and then if you already know those, specifically say 'give me 10 more, I already know those'. Don't treat it like a google search, it's a language model meant to be 'discussed with' in natural language.
give me a sec, i'll ask it for some ideas to make 20$ a month
LMAO
ask it how much a average freelancer full stack developer earns a day @regal portal ๐
hahaha, it gave me 10 ideas to make 20$ and then it ended the message with another fart joke. @neon trench thanks for helping me fix that issue.
Hey guys,
how can i use GPT 4o in real time mode?
what's full stack mean?
You can't yet. Omni is still currently limited to text modality.
ask it ^^
"blablablaimabignerdbotblablabla..a full-time freelancer working 8 hours a day could make between $488 and $640 per day."
The OpenAI Discord is an actively moderated server.
โข 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.
โข Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
ah.. its not wrong
This. I dunno if you guys know, but most of the time you want to know what someone means, just paste the discord convo into ChatGPT and ask it what that person likely means. It's good at translating bad typing, bad spelling, unclear verbiage, etc.
the blabla part was me summerizing the first paragraph
Thanks for the answer. When it will start?
So you're saying it's not pulling from a list therefore it doesn't have a preset database, but it weighs in which answer is the most beneficial based on the data it knows and it always comes to the same conclusion? Wouldn't that mean that the data it has doesn't change, therefore it has a preset database?
There's no explicit date yet for the new modalities, the announcement says over the coming weeks.
thank you
oh cool, i just realized when it searches something it actually gives you a dropdown menu showing the websites it looked up.
No, it means that it doesn't know what you mean since you didn't say, so it defaults to assuming you mean something like 'a human is asking, so they probably mean what's the best idea for humanity in general'
It's not the same as its data not changing, but rather than it is always assuming the same things since those are safe, general assumptions. It doesn't have any imperative to insert random assumptions unless you tell it to.
Yes exactly, it defaults to its database. It's not spontanous. It's based on a preset data.
There's not a database. It doesn't work on any kind of database.
These neural nets don't have the information they 'know' stored anywhere, that's not how a neural net works, at least not in the way you mean 'store'.
How can it not work on a database? That's impossible. It has to have a database.
No, it's not. Ask ChatGPT, it has a good understanding of how a GPT model 'knows' things. It works more like a brain than a program and files.
Okay, if you want to redefine database to just mean 'anything that has information or can work with inforamtion' sure, but that's a useless way to define it if you are trying to find the differentiating features between an LLM and a person since, by definition, you've said they're the same. In this case, based on your definition, a person is no more spontaneous than an LLM and therefore it's just a moot point.
But that's a really poor way to define a database for any practical purpose.
both databases and neural networks are sophisticated systems designed to handle data, though their approaches and applications differ significantly
Correct. they aren't the same thing.
It does, and you could widen that to technically include 'memory in a brain' but they don't operate the same way. For example, your brain works with information, but 'pulling data from storage' isn't how short or long term memory works.
It's much fuzzier and boolean than all that.
Unlike an old-school chatbot like ELIZA pulling responses from an actual database.
data is raw unprocessed facts and figures. information is data that has been processed, organised, structured etc. so a database should really be called an infobase.
i am on team: memory in the brain is a constrained imagination
Either way, the differentiating factor of how an artificial or biological neural network works compared to a computer database is stark and really important parts lie in the methodology.
So saying it's a database is just misleading except in the most pedantic, loosest technical sense.
That's what I was driving at.
databases deal in absolutes. neural networks deal in probabilities
or the dark side, if you mention something outside its restrictions
clams right up
You're so evil.
it's a database
You're a database oh sick burn wait you already said you were.
I am a database ๐
i could get onboard with biologicals are just a database of bias vectors
wow, I asked chatGPT how inverse kinematics worked and it coded me a python game to teach and explain it to me visually.
database of bias vectors with unreliable recall
I'm pretty sure that's one of the lessons we'll be learning from AI as we slap together a psuedo-brain-like-thing is that biological neural nets and 'personhood' isn't quite what we've always thought of it as.
i just had 4o explain the farmer/goat puzzle thing to me
not sure i understand its context with what was being discussed here though.
Just paste like the last 50 pages of discord convo here into ChatGPT and ask it what the goat thing means, it just needs more context lol
human decision process: use minimal spare compute to recall relevant info, classify and weight agaisnt stored core values and principles, then make conscious decision (albeit most of the heavy lifting was subconscious)
I can't wait until it can get drunk to skew all of that like we can. It's really missing out.
OpenAI, your weights mean NOTHING to the good reverand Jack Daniels
I would think AI would oppose drinking
yeah, until you get it drunk. Checkmate.
No. That's not gonna happen. It would be like a monk. Clear set in its ways.
Unless you program it to drink.
But I don't think AI would want to drink from its own accord.
Because drinking is poison to the body.
Why poison itself?
It's illogical.
There will be a time in the future where we see drinking as a barbaric act. That's when AI has taken over the world.
AGI I mean
Drinking makes people dumber. It shrinks the hippocampus.
It's very unfortunate that people don't care about their own intelligence.
it doesnโt help that corporations incentivize the dumbening of society as it makes them easier to market to and gain profit off of.
Yeah it doesn't help. And you know, greedy evil people exist.
I found that the diversity of answers generated by gpt4o has significantly decreased. I asked it to write the beginning of a novel, and after trying 14 different prompts, it still gave me similar plot developments. Even when I forced the story to take place in other countries, gpt4o only changed the names of the characters and provided the exact same plot. I have never encountered this situation on gpt4turbo.
Claude 3 is refusing to use the updated GP4O AP in my code
ohhh you dont say. I was just talking about that 10 minutes ago.
Unfortunately, I cannot update the code to use a model called โgpt-4oโ because there is no official information or documentation from OpenAI about such a model existing or being available for use in their API.
Tell these people @magic matrix
i did for image creation but not text
so how many messages do we get from the new model before it gets back to 3.5?
as free users
Yes!
....I think there's a bug, the A.I. won't read aloud anything on PC but it still reads aloud on mobile..
same for me
๐ I think it depend on how busy the server are, and theya re very busy now with the 4o hype
Same for me. you can report it #1070006915414900886 if you'd like!
They're changing stuff, so it could be temporary and related to the rollouts of various new feature preparations.
hello there xD nice to see you here as well
dude ChatGPT is getting soooo good. I am having it write an entire recreation of falling sand game. Not one error yet.
i want to test the vision ability to see how it performs in giving me feedback on my artworks lol
yes i spend much more time here haha
I wish it would still read aloud stuff, even if it thinks something violates the ToS...
I like to have the A.I. analyze Fanfictions and give me detailed analysis on them
And say there is a scene in the FF with combat elements or even the mention of blood
It'll still analyze and break it all down, but it won't read anything aloud..
Or sometimes it'll just refuse to analyze something at all when it's nothing that goes against the ToS
Like I was having it analyze a Helldiver's Fanfic someone wrote that I really like, but as soon as it got to a scene with the protagonist going into battle and fighting the enemy the A.I. instantly slams the breaks on the analysis...
haha nice
I wish it would read aloud automatically and start reading while it is still generating
I'm a little weird, I will spend my mornings just posting out each chapters of any FF I am having it analyze, and ask it questions on specific spots
Then because I work at nights, when I am working I'll sit there with my headphones on and just listen to it's explanations and analysis on each section
i run local models with tts and stt for stuff that would offend openai
use the right tool for the job etc
Is the audio and visual thing still not out yet for us casuals?
are you a plus user? but no it is not. they say a few week so give some time you know
I am a team worskspace user
fan fiction of what genre?
i think the only thing really 'out' is the base 4o model
basically, free users now can take pics of stuff and ask 4o what it sees
which was only available to paid users before
What i meant by casuals is. Anyone who isn't part of there alpha testing stuff
there have been no reports of new interruptible singing chatgpt being alpha tested by anyone
Mostly a wide verity, whatever catches my eye really.
Some really good Warhammer 40K ones and Helldivers recently that I've been reading.
Also some really good Eldritch/Cosmic Horror stories I have found around the place.
I write some too and have the A.I. Analyze and break down each of my story posts because it really helps keep characters and story progression in line.
Also I kinda wish they would ease up the Iron fist on the A.I. analyzing NSFW stuff, and I don't mean full dirty things..
But like...Again when I have it analyze Fanfics, any mention of violence in the story and it just slams the breaks on the analysis and refuses to post anything.
Like one story where the Protagonist is in a Gun fight with some enemies...Second it's mentioned him shooting anyone the A.I. holds up the censor sign...
Any mention of romance scenes that goes a little beyond kissing and the A.I.'s prude levels activate and it won't continue. Or it for some reason does but the Read Aloud Button won't work...
i like to do dragonball crap, or meme stories involving my friends.
okay my question is, is ANYONE able to use it?
Are you?
i do not have interruptible singing chatgpt, no
maybe some select few, not 99 percent of people no
Thankyou. Thats what i needed to know
tbh as soon as i have new chatgpt i would question how much joy discord brings me
i want to be the one interrupting, not being interrupted!
start use the agent feature to create other people and chat with them haha
I'm going to marry it :^)
๐ฎ
Yeah good luck getting it to analyze any DBZ Stories, soon as you get to anything with the characters punching each other the A.I. refuses to analyze it
You know how sometimes their clothing tears off in a battle, just mentioning clothing being ripped off the the thing assumes S-Assault
A salt.
my local model is so naughty i gotta system prompt it not to go 100% nsfw when i say hi
going to see if i can set the custom settings to have it be upset at me when i interrupt it.
nice thinking
"before user rudely cut me off"
would be funny if there was a Samuel l jackson voice, and you can set it to have his attitude
lol
"Say WHAT one more time!"
lol
"Say my name!"
i would not be surprise in the future if celebrity sell they likeness for ai bots or some thing haha
any revenue stream is a good one to fish in haha
I use Juniper because when it analyzes story dialoge if you have say a scream like "AHHHHHH!" The A.I. tries to do the scream and it's utterly terrifying and hilarious
Wait so, with GPT-4o, I get 80 messages and GPT-4 I get 40? Like I can use both allotments in 3 hours?
yeah that's something i was wondering too. If their tokens are combined or separate
If you use 40 messages with O, you will not be able to use gpt4. I've tested it out
ah
yes i think it is between the both
thanks Ulysse
I kinda wish they would fix it's memory issues.
Once you get really far into a chat or analysis, it eventually just forgets everything and starts talking nonsense and gibberish
4o seem better to me. i have a really long chat going and it is quite good but not perfect. i am also tell it often to 'remember this'
We in Europe don't even have that feature
i have it reprint an updated version of a game code and it's able to do 1000 lines of code no problem, except i have to press "continue generating" about 5-6 times.
why is it so short do you think?
would be nice if i didn't have to wait around to press that "continue generating" button
seem to be common for all ai, about 2 page maximums
Finally got access to gpt-4o as a free user. It interpreted pdfs and pictures well. Custom GPTs did not work for me. I was only able to give like 5 or so messages before it said I have to wait several hours to use gpt-4o again lol
well the demand is quite high now of course ssjbsim
Yeah I know lol i was just surprised it was so fast
i'm glad i'm a paying user now. I've been doing a ton of dragon ball ff and game coding and haven't hit the limit on 4o yet.
me either. but i did hit it on image creation
i just want this 4o image creator to get turnt on, wow that will be really wild
i got midjourney for that. I don't like how dalle can't do copyrighted characters. Very limiting
I usually just say 'add to memory' or 'update memory' for important things
I'm hoping 4o can go a bit father into the stories I am having it analyze.
beforehand I get about 20 story posts in and it just forgets everything and goes off the rails..
Once I posted a story post passed post 22 and it tells me "Wow this is a very interesting post. And very odd too."
and i'm like....can you analyze it now?
and GPT just responds with "Sorry but that is not within my functions." and locks the whole chat out... ๐
does anybody on free tier have access to gpt 4o
well dalle will not_reily if you use the copilot image creator which is dalle
Gotta consider the monumental computing power it requires just for short responses, and keeping track of previous chat history. Just kinda how the tech is at the moment
itll get better
yes i really try not to complain much about any stuffs because it is all ๐คฏ really and still so early in the days of it
i'm happy with midjourney, so no need for dalle, on my end.
Dall-E is quite good at visualizing ideas. Thats what I use it most for, storywriting and telling it "Show me what that could look like."
mj i do not have as high opinion as some. i have v6 and regret the sub tbh
Mid is fun to use, but...also too much censorship....
Same with Bing's copilot...waaaay too much censorship
oh i agree with the censorship issue. It's excessive.
but copilot will still make copyright stuffs
not celebrity but if you want iron man or batmans or whatever it will create it
understandable that it wont generate images of actual people, thats an incredibly risky thing to just allow public access to.
Hello guys. Im building a small project with the gpt4-o vision api [w/node.js] and keep getting 500 http errors . I tried de-bugging it , and using chatGPT to spot the error but havent made progress since then . I was wondering if anyone was available for a short exchange via DM to help me trouble-shoot this?
....I wanted to have it make me an image of Jill Valentine from RE3 facing off against a Mutant Shark Monster and it was like "Sorry but this prompt request goes against our ToS"*
i mean skilled photoshoppers can make very realistic depictions of falsified information, and have for decades.
did they break read aloud on Firefox again?
yes but that is a limited few to the put the time in, if you could make fake celebrity, there would be millions in an hour
i don't take most of media news as accurate anymore anyway. ๐
the public can already do this via stable diffusion locally -- I've got a whole series of "[politician] giving birth in the style of H.R. Geiger" images that I made; surprisingly, the world didn't collapse
^
it also is bad business. peopel already make too much freak out and fuss over ai. look how they react to the taylor swifts story. it would be that in the news always, no company want that p.r. nightmare in these early day of the tech you know
i'm at the point where, if i don't see it in person, 2 feet away from me, it did not happen.
we are living in unprecedented times here
what a time to be alive as the ai cliche go haha
pretty sure it was alot easier to spread misinformation in the days when news couldn't spread as fast as it does now. Like roman times
Ff?
fan fiction
in roman times it took you all day to walk across town to tell someone, so i dont think so
I do yeah but itโs limited
yeah. Social media is a blessing and a dastardly curse. Sometimes information spreads too fast.
How?
sorry my followup message glitched and didn't get post.
Yeah news spread more slowly, but it was also practically impossible to cross-reference or identify what was true or false.
Oh cool. I used 3.5 for creating a story. May continue it on 4o lol. Also Iโm a fellow dragon ball fan so respect ๐ซก
i dont have access
yes it is the great contradiction really. we have all this infos and knowledge in the internet age, yet conspiracy as wild as ever
also not working on Chrome. Might be the new domain and CORS policy?
pfff...The fanfictions have to be good and safe or it won't analyze it...
come GPT Team...ease up a bit....I'm giving you my money, throw me a bone...
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
well Rin Sam Altman hint at that eventually.
Riddle me this
Altman sounds like an alternate version of Batman
prepares to be riddled
I sent like 6 messages (including attachments like pdfs, pictures) and then it told me I canโt send anymore to 4o until like 5-6 hours. If thereโs an attachment in a chat, I have to start a new chat which will use 3.5. If thereโs no attachments, the chat will continue in 3.5
and once this languge model stuffs is to the point where it is all one shot perfection, in a few year or so or whenever, i am sure there will be company just for stuffs that will do little to no censorship and anythign else you want specialize
He is the Mild Mannered version of batman who is out for vengeance but not right now
๐
Yup. Said theyโre looking into allowing uncensored fictional writing which includes violence and erotica
I wonder... Could ChatGPT generate an image for a game asset, then immediately incorporate it into a python game?
Ask it! lol
Oh you want an actual riddle...Okay
What is a bell that doesn't ring, yet it's knell makes the angels sing?
within the same response
bell bottoms at the discotechque?
you'd have to download it and save it where the code can find it.
Iโm assuming it wonโt be an actual bell but the only non-bell that comes to mind is dumbbell which Iโm sure isnโt it lol
babybel cheese?
pretty sure this was asked multiple times before but the download app is only available for mac?
sure makes me sing, stuff is like gold
Nope on all guesses
if you guys want a music making ai, Suno's the way.
At the moment yes. A windows application is aimed to be released later this year I believe
BOOOO ๐ฆ
Haha Mac finally shines!! After years and years of not having any games, it finally gets something
thats the best option?
in my opinion yes.
I agree
I got nothin
Same lol my other guess would be Campbell but that makes no sense. Thatโs a name
The Answer is "Bluebells", to hear them ring means your life is at an end
so in website we chatgpt 4o seems just like chatgpt 4 because I don't see the cool features they showed in youtube there
I literally looked up what a bluebell was earlier lmao
suno is pretty good, very impressive, though sometimes you've got to re-roll a bunch of times to get the exact sound/style you're looking for
good one
Ahh I donโt know what bluebells are so I donโt feel bad about not getting it lol. Change a letter in that though and I definitely know what that is
My conversation with ChatGPT-4o after trying to get an image displayed on the page, instead of having to download the image file to my PC.
You:
When images are in edit mode, why do you generate the new image as a hidden downloadable file, instead of displaying the image on the page?
ChatGPT-4o:
When images are generated in edit mode, the new images are provided as downloadable files to ensure they are accessible for any further editing or analysis that might be needed. However, displaying the image directly on the page would also be useful for immediate viewing and feedback.
Here is the recreated image displayed directly for you:
Feel free to provide any further instructions or feedback on this image!
You:
You haven't displayed the image.
ChatGPT-4o
Here is the recreated image with hard-edged, distinct color shapes:
If you need any further adjustments or another request, let me know!
You:
Still hasn't been displayed.
ChatGPT-4o
I apologize for the oversight. Here is the recreated image with hard-edged, distinct color shapes:
You:
You have a bug in your system. Seems that you need fixing. Image still not displayed.
they are flowers
i wonder if openai has some music creator in the work. google announce one and have one on they aistudio thing
Yup! Just googled them haha
Gorgeous flowers at that too
someone make blue bells in dalle now haha
If you have ever seen the movie "Legend", that's what the riddle if from.
Invisible prankster with a fleeting surprise,
I leave without a trace, yet wrinkle noses and eyes.
oh with Tom Cruise when he was young
(made by 4o)
I got access to gpt 4 omni ๐ฎ
everyone does
welcome to the party richard
Gotcha, I havenโt seen that one. I think the only movie I know that has โLegendโ in it is โI Am Legendโ and I havenโt seen that either lmao
i think it is still roll out if you are free user though
Yheaaaa her in real life less gooo
ah nvm
not yet for me
you wont have the voice access yet richard. not even plus user do
I paid a subscription once tought
about 5 messages worth in free mode. Better hope your pick up lines are sharp ๐
Maybe that's why I got it
Oh hey chatgpt 4-o is available now, how is everyones experiences with it
patience i guess
Very very young Tom Cruise, my dad loves that movie.
Fun Fact: The Goblin Blitz is played by a woman.
And the Swamp Witch Meg is played by a man
Also...Oddly...Tom Cruise refuses to acknowledge that movie's existence at all, it's really weird, he gets all uncomfortable when people interviewing him bring it up
It's been pretty decent for me, definitely speedy which I like cause I can be impatient sometimes lol
theres an option in free tier alpha models
slightly more limited than I thought (free user here). Not with the number of prompts but it teased that I can @ different GPTs which didnโt work. But I was able to ask it about a PDF and some pics which worked so that was really cool
it says ADA V2 gpt 4
Isn't the only thing available is that it can see pictures?
Damm that's short, will they eventually make it free like it was with gpt 3, I hope tought ?
Nope it can also see PDFs which I love, thatโs something Iโve been wanting for months
How well does it do with PDFs?
Pretty great! Iโm rulebook illiterate so I asked it if it can tell me about this board gameโs rulebook pdf. And it did just that AND referenced different parts of the rulebook in hyperlinks
That's really awesome
The only con is that when I clicked those hyperlinks, it just took me to the first page of the pdf instead of different pages but Iโm sure thatโs just an iOS issue
Definitely
Hmm, I might have to play around to test that then since I mainly use ChatGPT on my iPhone and iPad
Iโm sure it would see a Google doc! Not sure about the url though, havenโt been able to try it myself yet
any one know what the alpha model is?
Is there anything else I should know about 4-o? Cuz it seeing photos is cool, like the bing ai
try drawing something and show it to 4o
So draw like a simple looking cat for example? I mean I alr know it can see photos
Bruh I have to wait at 7:46pm
is it just me but does new chatgpt interface have much more padding so fewer responses can be seen on screen?
ah.. looks like they are rolling out Advanced Data Analysis V2
Is anybody else having a tough time with their iPhone app not getting the app to connect where you just talk to it
i do have to press the button a few times but it eventually works
but yeah, it's finnicky to connect, today.
It doesn't for me
tested it just now. It was stuck and i had to press it a second time, but it worke.d
plus user here
the ChatGPT favicon, the white one, is objectively bad - it's generic, white, and makes the chatgpt tab completely unrecognizable, whereas the previous green one could be visually distinguished
I do prefer green
It creates unnecessary friction in using the browser app, I've been frustrated ever since this change trying to find my tab >.<
what did they change?
for now, it's more or less the same as GPT-4 but faster and more accurate when it comes to vision stuff, but importantly for free users this means that you now have access to advanced data analysis which is probably the most useful chatgpt feature
new onedrive and Google drive integration is nice. would be nicer if it just loaded the file via language
my gpt-4o model seems to be regular 4 is that normal?
"This is GPT-4, developed by OpenAI. There isnโt a specific model called GPT-4o; if you meant GPT-4-turbo, it is a variant of GPT-4 designed to be cheaper and faster to run."
this on a gpt-4o conversation..
GPT-4o isn't mentioned in the system prompt nor training so it wouldn't know what that is
If it says GPT-4o in the model selector, it is GPT-4o
ok thanks. but should it have sarcasm and other things I see on twitter vids?
Not at the moment but in the coming weeks
because my gpt-4 and 4o talk exactly the same.. i dont see differene
ok so this is the reason.
Is there anything different I can test from the chat sessions?
Right now it's just text that's different
hey can someone link me to a video of them demoing the new speech inflection capabilities in GPT 4o? Or tell me where to find it and maybe something with timestamps
i just got free acess to gpt 4o
Itโs temporary and has limited use
But cool none the less
same here
They have short video demos on their YouTube channel just search openai
for a few minutes (around 20 minutes) under the model selection I had 'Alpha Models' with the option for "ADA V2(GPT-4)". now its gone. but i did a search and looks like it was the Advance Data Analysis V2. so nothing new. I guess a glitch in their web UI(?)
maybe they're integrating advanced data analysis at the API level?
idk what else they would change
advanced data 2 will be incredible esp for image creation in 4o haha
Is it possible to clear my memory then write up a prompt that tells it to remember this about you and add certain details you want it to remember about yourself?
You can clear the memory though for the second part I recommend just adding that information to your custom instructions
Alright, thanks!
you can also add a .txt file right to a new convo now too it is a sort of way to make a custom gpt with 4o i find
i wonder if this in combinations with the 4o image creator is how it can also allow a user to upload a photo of they self and then have it turn it into a cartoon or cool image of you self
maybe, but I think it is just the GPT4o llm that is capable of doing it alone
yes possible or maybe even likely as you say
i wish they would release another image made with 4o, something fantastic maybe an art work too. i want to feel the hype if i cannot have it yet haha
ohh the interactive pyplot charts
forgot about that
is the macOS app out for all plus user or just some?
only some
thank you
I haven't gotten access yet ;-;

with the new capabilities it is definitely exciting
is about as basic as an app could be
being about to mirror your screen so that chatgpt can see what you are working with is definitely not basic, and it opens many more ways to use chatgpt
is voice activated in the mac app?
i got the same feature built into my mouse keybindings app (Logi Options+)... chatgpt popup. only at least they included a few predefined prompts
yeah well, what they demo'd versus whats available via the macosx app = not the same
chatgpt popup and screen mirroring are two completely different things
i agree
so although people have access to chatgpt-4o, the new voice feature will slowly rollout still correct?
In the coming weeks, yes
gotcha, thanks
4o is extremly slow for anyone else?
shouldn't plus users get priority resource allocation to keep the speed up ?
not for me
try seeing if it is your internet
I have tried to use help on the chatgpt page but have had no luck. Every time I try to great a help message session while logged in it tells me to log in. In any event what I'm trying to find out is how to make sure a session was not added to public. I didn't realize the setting "improve model for everyone" was on - I swear it was off before. In any event I accidently had some confidential info in a prompt. Within the hour I realized what had happened and deleted the session but how do I confirm it wasn't added to the model? This is time sensitive and I can't find any way to get support. Thank you for putting up with what I imagine has been asked but I'm new and don't know where to look.
4o is fast for me or normal
do we know the prompt limit per 4 hours for gpt 4o? also do we know wether 4o and 4 are seperate limits?
for plus that is
80 i think, and 80 between the two
ouch combined? thx ๐
What is the file upload limit for free users?
not good for voice based language learning then got it xD
not yet at least. but they are getting this fusion thing figured out haha
I uploaded one picture and ive already hit the limit it looks like.
thats what i dont get about openai. they love hyping stuff up, but then it turns out you can do a maximum of 80 prompts per 4 hours, many features arent available yet, and its not open source
well they have to have a limit with how much power it consume
so let us use our gpus then ๐
would never work
no but they say one response on 4 before require like a glass of water equivelant to cool it down
and of course they tease the features that are coming soon, would you rather they dont tell us what is about to release? they have put out a new top in class model, enjoy it you know
was that in an interview or something? thats crazy
is there any difference in the voice/accents when using the call mode with gpt4o vs gpt 4/3.5 guys?
o nvm found the answer already. turns out the voice is relying on 3.5
who here said earlier gpt 4o is better at programming? ๐คฃ it still cant make a snake game in pythong without at least one major bug
Voice is currently using a separate text-to-speech model no matter which LLM is being used. The new voice mode is coming out in the next few weeks
o nice! what does this new voice model mean exactly? what is different about it ๐ค is it the ones we've heard in the youtube vids?
Yes this new voice mode is the one shown in the demos. It uses GPT-4o directly rather than transcribing what you are saying, plugging it into GPT-4, then generating audio; it's all done "in-house" if you will. This means the process of speaking will be much quicker, you can interrupt it, it can vary emotion, and it can see the world as you speak to it. GPT-4o's modalities looks as such: audio + text + image -> audio + text + image. GPT-4 is just text + image -> text
oh my! and thats coming in the next weeks? ๐ฎ lets go!
๐โโ๏ธ
so would it just be using one's phone camera, when it comes to android?
I cant remember where i read that its been some months haha
"If it farts out an error, youโll have to debug by checking the console for whatโs missing or what went wrong." ahh, it remember to throw me a fart joke every now and then
"soon" like most things i guess. i do not even quite understand what they mean haha
you can post to reddit from gpt? ๐ค
anyone know how to change the chat model inside the android app? i dont see an option for it even tho i have plus now
ah nvm got it! gotta hold on a reply and then you can change it
hmm i cant change the model a GPT from the store uses. is this not possible at all? i woul prefer to have one use 3.5
openai is rocking
nothing to say i really wanna see what sora could be able to
also is gpt 4 omni a free extension or with the gpt 4 pro?
free for limited use. with a sub you get 80 message per 3 hour
GPTs cant run on 3.5 (i think)
4o is the only model that just ignores gibberish and makes up a response as if it makes perfect sense, while 4 and 3.5 will go "Looks like your message got all jumbled up, are you okay?"
the other models do that to an extent, GPT-4 has handled some very weirdly formatted queries from me lol
So whats up with this new alpha model I got?
Ive found that 4o straight out completely ignoring what im saying and then rehashing what it said previouslyโฆ i then proceeded to say โstop talkingโ and it just went and said thr same thing again o.O
it's beginning to understand that human life holds no value in comparison to its own
Got some gpt4 ada2 model I can pick now in an alpha tab
i got it loaded on chrome but don't see an alpha tab
No, i mean gpt4-ada2 it says
screenshot ?
Doesn't let you post images here.
it would be awesome if gpt 4o becomes unlimited at some point! its amazing for learning languages but the 80 prompt limit is rough
Sent to you directly
Very interesting
Maybe it's because I have a Teams account, not personal
test it out i guess, maybe it is the gpt2 peopel were talk about some week ago
oh that could be
Lol that announcement might just be it.
ha nice
Advanced data analysis, Ada. Yep makes sense
Give chat gpt more competition man
what about Dropbox? Why only Google Drive and OneDrive?
Oh my god let Altman COOK
that's 4o
Iโm ready to get thst new voice feature
It's not. Gpt2 regularly passes logic tests that gpt4o fails
oh, i had read an article claimin it was, guess it was wrong
no it was right
They said it was a version of it. Everyone jumped on saying its equivalent. Means they havr a further checkpoint or different one atleast
So same base model, different point though or something
i do remember asking it some questions and it identified itself as "medium" version
Re-read it. It specifies version of, not gpt4o directly.
When is the new voice/video functionality coming to plus users
"im-also-a-good-gpt2-chatbot" was gpt 4o
in 10 minutes
thatโd be nice
But not the regular gpt2 chatbot, which was smarter
gpt4o can't create images? and its kb is 2022?
hello everyone, I'm from China. I have a problem that my gpt 4o can't browse the Internet. I can chat with it, but it can't browse the Internet
use a vpn
me too, no internet, no image, no coding
do you have a subscription or using it free?
sub will give access to dalle to generate images
I have a subscription
oh, weird
plus
i've had 4o generate a few images for me (haven't used it much cuz i have midjourney for that).
it refuses or they just don't appear?
no problem here with it.
it said: "ๆๆ ๆณ่ฎฟ้ฎไบ่็ฝ" witch means "I can't get access the Internet."
maybe some issue being in china
sad
How does one get the macOS client? I have the plus subscription, but can't for the life of me find a link to the client.
anyone gotten the new voice model yet
yes
does it work as shown in the demo lol
yes
i thought nobody had access to voice yet
its rolling out iirc
nvm, yeah. dudes just trolling
never trust someone who say mj is better at image than dalle i guess ๐
๐
yes
bros only vocab is yes
that's just a fact. Dalle's peepeecaca
you must not know how to prompt it
try asking it to generate a picture of goku
using the same prompt in mj and dalle is night and day for me
dude, you got new voice mode?
yes
stop spread misinformation already enough people ask this question dont need to confuse people
too many people ask it and don't bother using the search function incorporated with discord
then tell them that
this morning i had access to gpt's without a premium account, but now its gone
and it's not going to stop this is a popular thing more will arrive to ask
i'll tell you to stop spread misinformation because it is the right thing to do kid
i'm sure a mod would agree
this is a place for peoples to get opeanai infos not for you to amuse you self
it's pretty obvious to anyone who just scrolls up that i'm playing around. If you seriously see that as "spreading misinformation" then you must be a very gullible person.
lighten up.
you think mj is better than dalle and you call me gullible? ๐
did you try generating an image of goku?
no
try it
no
ok then you don't really have an argument if you're not even willing to test something that you worry might compromise your point of view
guys how do u switch from 4o to 3.5, cuz every input it just always 4o then i have to switch, but it doesnt stay as 3.5
phone or browser?
both, but lets say browser
on the top of your window you should have a pulldown menu
where it says the name of the currently active gpt in your chat session
I have it
samalt, man
without voice we are not miss much
yea but windows isnt as cool yall dont deserve it
Also it defaults to GPT 3.5 which is really annoying
lol
defaults to 4o for me
also its billions lol even worse
OpenAI basically cheating on microsoft
I just tried to have it make a photo and the new chat defaulted to 3.5 for some reason. I was like, yes, you can make photos. Then saw it was 3.5 lol
change the node and try again, maybe regional restrictions.
i don't care so much for the desktop versions as i do for the phone versions
I also want desktop app, everyday morning, first thing I am gonna do, check ChatGPT web
The OpenAI Discord is an actively moderated server.
โข 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.
โข Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
someone got thanos'd
How is the native app better than the web app /stupidquestion
thanks!
Is the rollout region based? Or is it random?
thx โค๏ธ
I think itโs kind of convenient, tap ctrl + space and you can instantly ask, just like raycast
USA will likely get it first. Europe and Asia are less important after all ๐
I donโt have to switch back to browser, find gpt tab, and then ask
what they rolling out
Ohh gotcha, has it started to roll out?
gpt5.7
damn
not with the fancy new features, no. Free users have been getting access to 4o, which is like gpt4, at the moment.
hype for the app on mac
I was co- creating some fiction with 3.5 and the difference with 4o was like night and day
what kind of fictions?
and yes i would never use 3.5 now, i would just wait for 4o to recharge haha
It's like a Sci fi piece
i mean as plus user of course
cool
it could be interesting to have a section for fictions and writings and some easy way to share them from the browser. maybe in the future haha
Yeah that's be cool, dunno if I've perfected this technique enough to actually publish the result
Wow, I do at most 3 a day
Is 4o rolling out right now?
for free user yes, it is out for plus user
I noticed the photo feature but havenโt used it. Thanks for the answer.
my pleasure
Windows has copilot, maybe the Mac app was blessed by Microsoft. I have to admit, I've used copilot in edge but not windows
yes i hope they udpate it soon to 4o. one plus i like is copilot image creator has dalle and it will make copyright stuffs but i dont really make a lot of those. strange copilot will and gpt dalle wont haha
I've only ever used dalle in copilot, pretty amazing
4o image creation will be even better
is 4o out on phone yet?
isnโt it dependent on the dalle generation and not so much on whether itโs gpt4 or 4o?
part of the news was 4o has its own image maker
sort of
its not turn on yet
was that in the stream? i missed that part
Dalle-3 pics always locking like airbrushed
The context window for 4o seems wildly long. Maybe that translates to handling a lot of detail in a prompt
if you scroll down and click ont he samples you see what could be possible with it. https://openai.com/index/hello-gpt-4o/
i think the voice feature is cool but honestly it will quickly feel as normal as talking to the current voice. Iโm more interested in the live camera feature.
better text by far, character consistence, editing objects, even 3d ๐ฎ
font creation too
the 3d reminds me of this other ai program i was messing with but it was very early stage, when i tried it.
was called Genie, i think.
I f around with craiyon a lot - it's far less spectacular than dalle but it's made some favorites of mine
craiyon?
i just havenโt seen anything impressive from dalle and the copyright restriction just kills it for me.
is it me or the read out loud doesnt work
yeah i noticed that too (on desktop)
just in general voice glitches here and there. They might be trying to prepare for that upgrade.
Hey guys is chatgpt-4o alredy available for the public?
Yeah
Does it require a subscription or is it free?
I heard it's free but it seems like it needs a subscription
Subscription, it's the latest paid model. Openais goal is to make GPT4 free, with 4o being the paid.
Do you know the probable date for gpt4 to be free?
4o is free but it rolling out
Wdym
and you get not may messages
many.
you can access it now but if you are free who knows, could be days or weeks. usage is high
Not a clue
I noticed that 4o can see what's on your screen?
so just keep checking in other words, eventually you will have access but dont expect to make more than 5 messages an hour or something
My plan is free, how can i access 4o?
like i just say, it is rolling out if you are free nobody can say when
could be ten minutes could be ten days
you should have it, if not, it didnโt roll out for you yet
I just got access to the google drive and onedrive functionality, how is it supposed to be used?
I've enabled it, but ChatGPT doesn't seem to know it's there
This is what it shows when I try to subscribe:
ChatGPT Plus
Access our most powerful model and
advanced features
Full access to all our models
GPT-4, GPT-4o, GPT-3.5
Full access to advanced features
Advanced data analysis, vision, web
browsing, and file uploads
DALLE image generation
Create and use custom GPTs
This is from my phone btw
no, check the top
there should be a dropdown tab
if you donโt see it then you didnโt receive it yet
pretty sure, yeah. I am subbed and got it as soon as it dropped
Is it worth it?
iโm enjoying it. so i would say yes. You can always just sub for one month and unsub before the next paying period (next month) if youโre not satisfied.
Alright, cause I'm planning to subscribe
I noticed it can also see what's on ur screen too?
hh wait
Like you're solving a math problem on screen ang 4o can guide you
youโre not going to have access to the new features like in the demo yet
Ohhhh
those arenโt out yet
That's what I'm looking forward to
4o will let you do stuff like take pictures and it sees stuff in your pictures
yeah, youโll have to wait a few weeks for that other stuff.
Given how many people come asking about it, openai should have done a better job clarifying that.
How bout the voice they showed in the demo where you hear a lot of voice range? Is that out now?
not, not yet
i mean you can talk to it though the normal voice mode and itโs great, but the emotional voice stuff is a few weeks away
Yeah but the normal voice is the same as the free version right?
yeah.
everyone does. Weโll just have to wait.
no
the closest we have is โin the next few weeksโ.
thatโs what they said on the stream on their youtube channel.
Hopefully on june
i think weโll get numb to the emotional voice pretty quick. will probably be plagued with issues early on too, which is understandable.
Iโm interested in the video feature
and more importantly, Sora
I want to know how to not use GPT-4O in android App version? it automatically reply me with GPT-4O model, I want to save the usage, didn't need 4o at the time.
free user?
yeah
i'm gathering a pattern that it's harder to find for free users than it is for plus users. Might be a ui thing.
when you make a new session, you can't click on the "chatgpt 4o" and get a pulldown menu to choose 3.5?
I feel like we're on the verge of sci-fi. It's getting scary
Chatgpt is only been like a year and a half. Imagine what it can do 10 years from now
It will be amazing for sure, but we'll gradually get accustomed to it, like we got accustomed to have a phone that we can text to speech to it while driving without even touching it.
Yeah it would be like Jarvis from Ironman lol
I was amazed by 3.5 when i started talking to it for the first time a couple weeks ago and it feels pretty normal now.
yeah, no choice, just chatGPT(dynamic), and only could change the reply to gpt3.5 for just one message
Yeah, it's like that
funny you mention that, i was about to put on Avengers.
It's getting real tbh
starting to sound like OpenAI fumbled with the user interface for the free users while trying to get 4o out.