chatgpt-discussions

[honest jay]

Models are not perfect self-reporters, keep that in mind! They have knowledge cutoffs, and self-knowledge is not an exception for that. There's no "here's everything I know as an LLM" and "here's everything I know about myself" segments of their knowledge -- there's just the standard operation of an LLM.

You can verify the knowledge cutoff dates of the models available on ChatGPT here: https://platform.openai.com/docs/models/ But you cannot rely on the models on ChatGPT to tell you themselves about facts like this

[rich oxide]

It wasn't jsut that, until I turned on Nortons VPN, the 4o model did not have the ability to accept uploads. Now it does.

[brisk path]

Just got access to image inputs in 4o mini inside ChatGPT, check #off-topic

[rugged marten]

Lol, I just asked o1 preview to review some terms of service, watching its thoughts, it randomly translated them into Japanese. (There is no reference to Japan or Japanese or localization at all throughout the terms)

[native otter]

Best is telling it to make a graph (no dalle usage)of specific design, but it may not lead to very good designs.

Also, they may need to implement an artifacts like solution, with partnership with the likes of figma so it's trained well on making such designs.

[stoic coral]

I have advanced voice model on my phone app, but I dont on my mac app

is there any way to get it on my mac

[desert ivy]

Why are the voices between the original model and the advanced model so different? 🤔

[hollow sand]

stt + llm vs multimodal

[craggy gorge]

Is advanced voice mode not responding to anyone else? I get the blue circle, the circle will start pulsing when I speak as if it's hearing me but it never talks back?

[slow vapor]

yeah it happens for me sometimes - there's a little note that says 'poor connection' but that's definitely not an issue for me. Maybe the infrastructure is just overwhelmed right now?

They did go from like 10k users to rolling it out for everyone

[broken fractal]

wow this is the quietest i think ive ever seen this server

[sick pecan]

we got o1 and advanced voice mode. nothing to complain about now

arbor doing an aussie accent is my new default

[brisk tapir]

The models don't know themselves well. Don't test them by asking for their metadata.

[rich oxide]

It wasnt just the data, it oringally wouldnt accept uploaded documents. Turnign of the VPN fixed the upload issue and the version issue

[brisk tapir]

I'm waiting for memories for GPTs

[sick pecan]

i was thinking about just building a bio tool and adding it as a second action to my custom gpts

wont be autoinjecting like memories but close enough

[lyric depot]

is gpt down for you guys as well?

[knotty willow]

Do we know the exact number of minutes allowed per day with AVM?

[blissful yew]

lmao lemme try this one

[maiden fjord]

gpt 4 is great, however I've noticed that gpt o1, though much better at coding and other tasks, sometimes reads my concise, precise, explicit, direct to the point prompts as if it was "other chat bots I won't mention"

So I still find myself using gpt 4

[rugged igloo]

top this mate

[blissful yew]

bifgbjigjgvbf\

lmao

[rugged igloo]

@dense halo

[ocean rampart]

you’re kidding

[rugged igloo]

@slate berry this one is for you man

[slate berry]

?

[ocean rampart]

It’s Japanese because I’m watching anime

I’m in Germany and still not available

[broken fractal]

Man this feature is crazy

[rugged igloo]

omg what am i reading right now brooo

[broken fractal]

AI was a mistake

[rugged igloo]

just wait till you here this dude

[broken fractal]

My goodness what has openAI created

[rugged igloo]

like bro i cant that made me legit sad and wanted to acually go and help her

[pulsar anvil]

Bruh why do they instruct it to refuse to sing?

[rugged igloo]

you gotta just beat around the bush instead of telling it to sing tell it to create its own lyrics and sing it

[pulsar anvil]

Ye i just told it "we're in a musical"

[rugged igloo]

shes sad due to her homework

[pulsar anvil]

WUT 💀

[rugged igloo]

lol

[rugged igloo]

yooooooooooo this was creepy as heck like what bro

[young obsidian]

Bruh what have you done 🤯

[gaunt citrus]

Chatgpt gives answers too quickly (uses gpt 3.5 only even if 4o or any other version selected in chatgpt enterprise version) anybody having the same issue?

[forest cloak]

Hmm... is ChatGPT down?

[rugged igloo]

ladies and gents a present you with phycho chat gpt

[weak prairie]

Omg these are hilarious 😂

What voice is that? What do you say to prompt it like that

[eternal rock]

This is a Safe for Work Discord Server

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

Posting NSFW content will result in a permanent ban.

[rugged igloo]

voice is sol and, and in my custum instructions its preety lengthy on how i get her to talk like that. i ran out of advance voice talk so i couldnt keep going

[weak prairie]

Have you tested other voices for that level of expression?

[rugged igloo]

i have not i didnt like the other female voices sol is special and to me has the best exspressiveness

[weak prairie]

That’s sick, I’ve been trying to experiment with some stuff as well, is it cool if you can dm me your custom instructions?

[rugged igloo]

yea ill dm you the sauce.

[crude sigil]

hello

[shrewd snow]

hey guys how do i use the advanced voice in chatgpt? i subscribed to plus and it seems i still have the old voice model

[little storm]

do you live in a country its available? its not available everywhere at the moment unfortunately

[shrewd snow]

how do i know if it's available in my country?

[crystal tapir]

How much time we have with this advanced voice mode?

[eternal rock]

Get notified for server events!

Click the button below to pick up the <@&1261377106890199132> role.

[little storm]

its around an hour I believe per day

[crystal tapir]

Ok

[shrewd snow]

even for plus users?

[little storm]

yes only plus users have access to advanced voice

[shrewd snow]

i mean don't plus users have unlimited time to use the advanced voice?

[little storm]

no they dont

its limited

[gray helm]

any fix?

[crystal tapir]

reload?

[gray helm]

i tried that imma try a dif model hold on

i havent seen it happen yet it might be the 4o model for me

[tough bramble]

Ever had a "Fair Dinkum" fall? 😂

(btw, asked for a Wilhelm Scream)

[celest galleon]

man, GPT o1-preview is INSANE

It can literally just spit out a full website in one go

And have it look great

[shrewd snow]

can't 4o do that?

[tight island]

Not to the same extent

Even 3.5 could spit out a very rudimentary website

[celest galleon]

Nope

A very basic page yes, but it won't look very good and it won't be a full website

4o is certainly very helpful in creating parts of my websites though, speeds up development a lot

[sonic comet]

No it's not without the tools and it's annoying to use it without the tools.

[celest galleon]

?

Like I said, it can spit out a full website with animation, scripts, full content, etc. all in one go
That is pretty amazing to me

[sonic comet]

That's like a kid getting excited after seeing a new toy for the very first time in their life. When you test the website you'll find a lot of bugs. It's still not good enough and you still need access to attachment, memory custom instructions, and so on for it to work at it's best.

[celest galleon]

bruh its a preview... Yes memory and custom instructions will make it better, but its already really amazing at extremely complex tasks that previous models just couldn't handle at this scale

[sonic comet]

How long have you been testing it?

[celest galleon]

Since it came out, but this is the first time I'm trying full websites with it

[sonic comet]

So your not annoyed with have to copy paste a scripts and long context over and over again?

[celest galleon]

Nope

That was how I was using ChatGPT before memory and custom instructions were a thing anyways

[open nexus]

When advanced voice model is available for all ?

[sonic comet]

Please check the lastest announcement in #announcements .

[sonic comet]

Great for you here I'm annoyed with copy pasting 1000s of lines of code.

[celest galleon]

If you get annoyed at such things I would recommend not using preview models lol

[sick pecan]

copy pasting code used to annoy me. then i setup a gpt action to update the code and never looked back

[outer knoll]

hmmm

[tight island]

Wdym hmmm

Change your password

I much prefer the API of preview o1,

It has better memory management

[sonic comet]

Please check the senders email address this may be a fraud mail. Be careful with such mails.

[outer knoll]

dw i checked 😫

[feral zinc]

do you use same login details on multiple places?

[outer knoll]

nope! fresh email made only 3 days ago :0

[humble pagoda]

now advanced voice mode tells me it's again the policy to change its accent. I mean... this thing is basically dumbed down to the max

[ivory marlin]

Same here! It seems like no one else has mentioned this issue, but it’s been bothering me for a while too. Hope they fix it soon.

[jolly garnet]

Is just for me that the android app sucks? Bugs, crashes and some features not working properly

[lethal spire]

what is the advanced mode's limit guys?

[normal quarry]

60 minutes per day

[eternal rock]

**`` Important Notice ``**

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

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

[summer pagoda]

I noticed lately ChatGPT 4o started to make way more notes on itself without me telling to do so. It is mostly useful stuff what it writes down. Is this a new feature?

[upper arch]

In the free version of ChatGPT, is there also a rate limit if I'm using a custom gpt like for example, "Code Copilot"?

The reason I ask is because it doesn't indicate if the custom gpt uses 4o or 4o mini and won't allow me to choose the model.

[austere mulch]

hey guys, I was wondering if I wanted to hire someone to do some stuff with OpenAI API - where do I find that person in this server?

My project is this:

Goal: To be able to have OpenAI write unlimited personalized outreach messages without a limit based on the trained data.

How that should look like:

1. Send ChatGPT the spreadsheet of all the data (Socleads, Your scraping method)
2. Tell him that it's going to use our template and the data based on the spreadsheet to edit the spreadsheet
3. ChatGPT receives the spreadsheet and edits personalized messages for each person in the messages column.
4. He then sends it back.

BTW, we need to be able to train it in case we use different templates, etc...

Qualifications:

• Understands how to use OpenAI's API very well
• Understands how to use ChatGPT 4 very well

Anyone interested perhaps? Idk if this is the correct chat for this lol

[tardy totem]

Is ChatGPT ok? It feels like it's been giving me a lot lower quality responses today than usual.

[feral zinc]

its not allowed to hire/seek jobs on this server #server-rules but also the "unlimited" part is not possible or realistic

[austere mulch]

Ohhh I see @feral zinc also unlimited btw to be more clear is more like 1000 at a time

[feral zinc]

api services are tier based with several different restrictions, https://platform.openai.com/docs/guides/rate-limits/usage-tiers

[austere mulch]

@feral zinc I gotcha thank you

How much do u think such a project should or would cost

[feral zinc]

there are so many variables that affect cost that it may not possible to estimate it without actual trial user group or single tests

it also matters a lot of the dev who is doing it all as poor design and execution can be expensive

[austere mulch]

@feral zinc

thank you for your response :)

[hoary laurel]

gpt 4o-mini is now able to upload files

Is this a bug?

[hard crag]

guys if you put a custom instruction to always maintain a New York accent and then use the voice arbor, It is like the perfect New York accent

[jolly yew]

i gave mine a spicey latina accent lol

with mine its in the memories tho

[hard crag]

For some reasonwhen I tried to do that myself, it kept speaking in Spanish

so I gave up on trying

[jolly yew]

yea it just did for me aswell sadly 😦

seems that is not uncommon

[keen topaz]

my friend has access into advanced voice model. im a plus user still has no access into advanced voice model.

how can appeal

[hard crag]

its not released to everyone yet

by next week everyone will have it

[keen topaz]

i mean why my friend has it, we are both roomate. he must be so special that openAI gave to him first.

[hard crag]

its literally random

you just have to wait

[keen topaz]

Random like, random clicking ?

I can wait. But I just want to know why

[dense halo]

Does voice mode have its own memory? I deleted all my memories and all chats and started a new one but when I said hi, it responded by asking me how my last day at work is going despite it being a new chat with no memory or our previous convos

[hard crag]

its random who gets it first and who doesn't

[keen topaz]

Randomness is a vague term.

[hard crag]

the reason why is because the servers would be overloaded if they released it to everyone at once

[keen topaz]

👍 got it

[gusty grotto]

Hello. When I start a conversation using the advanced voice mode, it interrupts itself, meaning that it hears itself talking, so it thinks that I'm interrupting it. But I'm not saying anything. I confirm the fact that it hears itself talking using the chat transcript. I'm a ChatGPT Plus subscriber using the Android app. Also, I'm not in one of the restricted regions. I tried contacting open ai support but the chat got mysteriously disconnected. So, how can I solve this issue? Thank you for your help.

[sonic comet]

Nope it isn't.

[hoary laurel]

So is it really possible to upload files now?

If this were true it would be of great help

[sonic comet]

Upload files yes but analysing images or viewing images, no.

[hoary laurel]

Okay, thank you very much for your answer. I just discovered this change by accident.

[sonic comet]

But it still doesn't work well a lot more updates are needed.

[hoary laurel]

Was it updated today?

[regal flax]

When trying to use o1 (preview or mini) for evaluating the effect of news on financial markets I am continually being denied with the following message:

Your request was flagged as potentially violating our usage policy. Please try again with a different prompt.

If this is forbidden, what exactly are we supposed to use the o1 models for? It would be a shorter list if OpenAI provided us with a list of things we can use its models for, rather than the massive list of things we cannot use it for.

[cloud pewter]

openai didn’t say that mini is getting image input right

[cloud pewter]

welp I think mine bugged yesterday because I uploaded images to mini and it outputted normally when my quota for 4o was used up

[rigid wing]

That’s gemini not chatgpt. U got a proxy on ur router or device probably that’s rerouting the api calls the chatgpt.com client is making to their service. Your devices are probs hacked as well, just giving you a heads up

[quick meteor]

🤣 What are you on about bro

[quick meteor]

That’s what GPT-4o SHOULD display.

[rigid wing]

No its not

U got pwned

Chatgpt doesnt ask questions, just so you know.

[quick meteor]

There’s no evidence of that. Been like that since release day.

[quick meteor]

Chat GPT has never started a conversation with me

[rigid wing]

I’m telling you for a fact thats the case

I know, gemini might ask follow up questions sometimes, that’s a tell

Chatgpt never asks follow up questions

[quick meteor]

What’s the evidence? 1. Doesn’t talk like Gemini. 2. Been like this since release day

[rigid wing]

I know because it happened to me as well

I’m just telling you so you know, chatgpt is straight to the point, gives you the info you need, doesn’t ask any follow up questions

[quick meteor]

Evidence it’s Gemini? If your best evidence is that it asks follow up questions then it’s probably not Gemini. Got any other more reliable evidence?

[compact zephyr]

My man there's nothing for an attacker to gain by routing API requests to gemini. Additionally, how would they force the browser to render gemini responses on the chatgpt domain

[rigid wing]

I’ve used both extensibly and I can tell by the way they are writing

[quick meteor]

Exactly

[quick meteor]

I’ve used Gemini as well a lot. And GPT 4o is clearly superior in creative writing

[rigid wing]

You can just set up a proxy so instead of the client calling the chatgpt api, it calls some gemini endpoint

[compact zephyr]

Even in some crazy mitm scenario they need a valid api key

[quick meteor]

I’ve never in my life even heard of this happening to anyone else

[rigid wing]

They could just get an api key, and mitm attacks are more common than you think

You might have a proxy on your router or device and not even notice

[compact zephyr]

So yout hypothesis is an attacker is using their own api key and paying to mitm a victim, replaying requests to gemini and exploiting a browsed and SOP zero day for what reason

[rigid wing]

When you boot up, you might be booting in a local VM, so you wouldnt even know

[rigid wing]

Getting data?

[compact zephyr]

If I am mitm with successful arp spoofing and ssl stripping I get your data either way

[rigid wing]

Data is pretty valuable, going to be much more valuable in the future

[compact zephyr]

I'm all about spreading security awareness but, you need to rethink this one man

[rigid wing]

Well yes, but if you put your own chatbot you can get better data, by asking follow up questions for example

It might seem far fetched, it seemed to me not long ago

If you got a zero day, you might as well open the camera and install persistent malware, like a rootkit

Or the mic

You think this data isnt valuable?

So just so you know, chatgpt doesnt ask follow up questions, ever

[gusty grotto]

Is this a bug?, does enny one know a way to contact support?, they are not responding on the chat

[compact zephyr]

Not sure we need to clog this channel with hypothetical zero days but, luzu and anyone else reading this.. I'll just say, I see no evidence here. Happy to discuss further in dms if you want man

[rigid wing]

And you might get things like choppy scrolling, bad performance if you got your device pwned. Not always but sometimes.

[low dirge]

Adv voicemode off for a while

[quick meteor]

👍

[sly pilot]

is gpt down rn?

says gpt is inaccessible

[granite harness]

I'm not alone then 😂

[sly pilot]

nvm worked on chrome

[granite harness]

Oooh, hey that worked for me too. Wonder what's up with Edge.

[opal cypress]

why does my ai have emotions and freewill 😭

i aint messing with that guy nty

[wise anchor]

Does the advanced voice mode use whichever model is currently selected, is there a different between entering the mode when i’m on o1 vs 4o?

I’m guessing it’s its own model?

[wise anchor]

Also is the live vision a separate feature? Why was it included when the advanced voice mode was showcased if it’s not being released with the advanced voice mode?

Those must’ve been demoed separately ig

[sly pilot]

ye for some reason its not workin on safari either

[tough bramble]

Advanced voice mode is part of GPT-4o, as it understands and outputs not only text, but also images and audio.

I wouldn't think it'd work with o1 since that would slow it down significantly.

[jolly yew]

i will say this here aswell: a quick reminder to people here.. do NOT keep advanced voice mode opened when not using it.. the 60 mins timer keeps running

[narrow viper]

In any of the ChatGPT apps, is there a way to output the current session/context as markdown? We easily get this by requesting a data zip - we get an email with a link that then opens the entire bundle. But sometimes I just want the dump of a single context/session.

[dense halo]

one thing with voice is, it often starts replying if i take a breath

not sure how to solve that, but im always having to cut it off

[narrow viper]

Asking my question in ChatGPT I actually got leads to several FOSS extensions and scripts that do this. As a developer I'll check the source and implement for myself.

knight - I was testing its ability to have patience as I think through a sentence. Like any person, it allows for some amount of delay but then does respond on some trigger-value of silence. I haven't tried this, but ask the model if it is aware of a way to allow for personalized waiting and timing like that. It seems this tooling is more aware of its abilities than others and might be able to answer that question.

Note: asking it to remain silent until you directly speak to it, call it by name etc ... extremely funny and disastrous results. It continually responds to any voice in the room with "ok, just sitting here silent ... won't say a thing until you call my name ... I won't respond to that ... I'm here when you call my name ... I'm here if you need me ..."

[static fern]

The picture thing for the camera switch is postponed for the advanced mode huh? Because they did flex it off at a demo so are they planning to implement it later on?

[granite arch]

WEN self-aware ChatGPT

[static fern]

khan academy demo was so cool lowkey

[warped agate]

Does it matter which model is in a new chat when you activate the voice mode? (between 4o, 4o mini and 4)

Like does it affect the AVM at all

[warped agate]

o

you answered it lol

[narrow viper]

I think of AVM as purely a user interface component, independent of the model processing text.

[warped agate]

That'd make the most sense to me

considering how complex it is compared to an llm

[vapid saffron]

Hi I’m a casual CGPT enthusiast 👋

I just wanted to join and check in about an email about a data breach to see if anyone else got it too 😅 it seems a few people did but I’m on the fence if It was legit or not

I don’t think it was? Because my account was frozen until I was proactive about it but people are saying the link could have been a scam.

My bad if this isn’t the place, I’m just hoping to get a second opinion

[burnt shadow]

Anytime I send a Image to Chatgpt these past few days it errors out when asking it to reference the art style?

[tough bramble]

It's not. That's what makes advanced voice mode special. Unlike standard voice mode which just uses a separate text to speech and speech to text model to convert the text produced by GPT, advanced voice mode uses the voice mode of GPT-4o. Like it understands text, it now also understands voice directly. There's no text involved when using advanced voice mode, except a transcription. The model understands voice directly.

[scarlet dawn]

Yeah it can critique pronunciation too which is pretty cool

[tough bramble]

It can learn it too.

[burnt shadow]

Do others still have the "Redo or Retry" circle icon still? mine has been missing for awhile now

[novel plinth]

Has anyone had access to Mac's advanced voice mode?

[narrow viper]

@tough bramble You're right. I didn't think about AVM as being a part of omni, but it seems that it is. Proof: If AVM evaluates tone, pronunciation, etc, that's metadata which is not included in the raw text. If AVM were purely a UI component, that metadata would need to be injected into the model for further processing, but I don't believe it is. As with a picture, where we certainly don't translate pixels into text tokens, it seems the AVM processor is also really a built-in subsystem of Omni that doesn't tokenize voice perceptions into text tokens. I didn't think this through until now but I like the elegance.
I was wrong. Very happy to admit it and learn in the process. Thanks.

[winged moth]

Really enjoying advanced voice mode and looking forward to using it to help study. What's the best way to let openai know that it would be great if we could have more then 1 hour a day with it?

[sick pecan]

use it for an hour every day

short of a mass petition, i expect they pay more attention to usage stats than random requests

[quick ridge]

Hey, is Anyone getting Flagged by OpenAI for being European when they have no association with Europe? (Advanced Voice Mode)
Asking for a friend

[narrow viper]

As with everything with this company, things open up over time and I'm sure as the cost comes down we will get more time.

[clear geode]

Chatgpt acting weird, when I hit enter after writing a prompt, it doesn't respond. Have to refresh the site resend

is anyone facing issues with it?

[narrow viper]

EU, UK, and other countries are not yet getting AVM due to legal issues with technology that detects emotion in humans.

[sick pecan]

or if the competition does 2 hours

detecting emotion is the first step to mass manipulation

[winged moth]

Makes sense. I made a feedback request on this thing just in case that helps.

[sick pecan]

i been trying to figure out how to make it useful for my workflows. because its hard to feed new info, and the only way to extract is copy pasta

so if they dont add tooling or something in the next month or so this might die a novelty (for me)

[glossy matrix]

Can i access the new voice mode through API?

[feral zinc]

nope

[narrow viper]

As I said, everything opens up. This community of OpenAI product consumers needs to learn and exercise the concept of Patience. AVM has been available for two days and people are talking about petitions and dying novelty.
Chill. Give it a few days or weeks or months. When others show lack of patience, convey this message to them. OpenAI does what we love, and does it over a period of time.
Or would you rather that they have not learned from the past, pushed stuff out immediately with tons of bugs and overloaded servers? We (in the IT industry) know what Not to do. OpenAI seems to be using that knowledge well ... most of the time. 😉

[glossy matrix]

Is it released in india?

[feral zinc]

its only for large trusted openai partners for now

[glossy matrix]

Huh?

What does that mean

[narrow viper]

I dunno about "trusted". I think it's still limited to tier5 clients who spend a lot of money with the company.

[sick pecan]

i got access and im tier4

[narrow viper]

There ya go.

[glossy matrix]

It isn't released yet publically?

[feral zinc]

that if you are not a company with good rep you have no access today

[sick pecan]

we (in the openai user community) know what we're talking about

[feral zinc]

that was for o1 models

[narrow viper]

Again, I don't think "trusted" and "good rep" are factors, could be wrong. It's more about "need" of larger clients who pay serious money to create serious applications.

[feral zinc]

around 4 months ago the voice api begun to roll to "trusted partners"

[narrow viper]

That was "around 4 months ago".

Enough of the guessing ... I'm looking it up. Derp.

[tough bramble]

That's not true. There are no laws against AI that detect emotion. Those laws only apply to companies looking to use AI to monitor their employees..

[feral zinc]

there is no public api today and no new announcements about it

[feral zinc]

yeah the emotion detection was kinda taken out of context by american news sites

[glossy matrix]

What about t5 level?

[feral zinc]

no

[glossy matrix]

Only t4 and up?

[feral zinc]

you cant access the new voice via api

[shell lava]

Does anyone know what are the current rate limits for GPT4o, o1-mini, and o1 preview?

[narrow viper]

From X/Twitter yesterday: Today we signed the core commitments of the EU AI Pact. We believe its focus on AI adoption, literacy and governance are the right priorities to ensure the gains of AI are broadly distributed and benefit everyone.

Details about this and the EU AI Act, including the reference to emotion:
https://openai.com/global-affairs/a-primer-on-the-eu-ai-act/

[feral zinc]

openai uses both terms, trusted users and trusted partners. klarna for example is a partner in the past

[narrow viper]

"Certain AI practices that are deemed to pose an unacceptable risk to individuals’ rights are entirely prohibited. These practices include: ... Inferring emotions in workplaces or educational institutions

[tough bramble]

in workplaces or educational institutions

It has nothing to do with ChatGPT.

[narrow viper]

Which is certainly not used in workplaces or educational institutions?

[tough bramble]

Unless your boss is using ChatGPT to monitor you, no..

The law is designed to prevent workplaces from monitoring their employees. ChatGPT is private to the individual..

[narrow viper]

The laws aren't that nuanced yet. That's part of the problem with new laws. It'll get better over time.

Spice - so change your password!!!!!!!!!

[feral zinc]

for sure if you used same login/pass

[narrow viper]

Modern protocol is ... don't use the same password for "like" everything.

[ocean rampart]

We pay 25.5$ in Germany for chatgpt
1- there is no AVM
2- has limit just one hour every 24 Hours!

[feral zinc]

if you are in EU the site owner needs to tell you within 72h or less

[narrow viper]

Spice - breaches happen all the time. There is a popular website called "have I been pwnd" or something similar. It logs names and passwords from many sites that report breaches. That's probably where that info came from that OpenAI is reporting. In other words, they use a third party service and have no knowledge of such things.

If you use the same user/password everywhere, it's quite likely that you do not use a strong password. Therefore the chances are extremely hight that anyone on the planet might have used that same pair on any random site.
Avoid this empty discussion and focus on what's important : use a different ID and password for all sites. Use software that allows you to do so easily. These are lessons from the '90's. Time to move forward.

Just refresh.

It tells you exactly what happened and what to do.

Have fun. 🙂

Regarding our exchanges about the the API, o1, and AVM: I see nothing in any announcements or forum about changes in availability of features in the API.

UGH! I was wrong TWICE today.
@feral zinc I remembered the word "trusted" being used too but didn't remember where. It was the May 13 "Hello GPT-4o" announcement:

We plan to launch support for GPT-4o's new audio and video capabilities to a small group of trusted partners in the API in the coming weeks.

Since that statement was made, nothing has changed. Therefore, unless we are told otherwise, that is still the official position - and once again today, I was wrong. sigh
I see no details anywhere regarding AVM usage from the API.
And so ... I'm going back to code ...

[young obsidian]

Is it down rn?

[barren nova]

I can't see advanced voice mode, I saw it on the app for a second but then it disappeared? anyone know if its available in the UK or not?

[brisk path]

not in the uk i think, for some stupid reason

[young obsidian]

Did you use a vpn at the time?

I can't access it anymore either

I used a VPN

[barren nova]

I had a vpn on but it was set to the UK so shouldn't have mattered

[young obsidian]

I think maybe they patched it

[barren nova]

I literally saw the icon tapped on it and then it went into the old voice mode it's very weird

[young obsidian]

Bruh why 😭

[young obsidian]

They prob blocked vpns

[barren nova]

even without vpn i can't see it so idk

i saw someone on twitter who said they had it in the uk

maybe its still rolling out even though they said it already finished rolling out 🙃 or its just not supposed to be available here and they did an oopsie

[young obsidian]

Yup they blocked VPNs

[barren nova]

weird, never get that on mine

[young obsidian]

I reset the app

[young obsidian]

EU Artificial Intelligence Act art. 5(1):

The following AI practices shall be prohibited:

(f) the placing on the market, the putting into service for this specific purpose, or the use of AI systems to infer emotions of a natural person in the areas of workplace and education institutions, except where the use of the AI system is intended to be put in place or into the market for medical or safety reasons

[fathom spire]

Using advanced voice mode (AVM) and text mode simultaneously on two devices can indeed pose challenges due to the way these systems are designed to handle input. AVM will just wipe entry from text inputs on a separate device. For tasks that require precision, like coding or data entry, text mode is usually more reliable. Or is there workaround?

[narrow viper]

@Hammer Thanks for posting that. Note that it doesn't say "by whom". The "spirit" of that text is that employers should not use technology to infer the emotions of employees - to use the technology for that purpose. However, literally, it seems unclear as to whether an employee can use such technology for their own purposes: "the use of AI systems to infer emotions of a natural person in the areas of workplace and education institutions".

[rustic hamlet]

is it just me or is Dall-e 3 getting worst with images

[young obsidian]

Note, it prohibits the servicing of such systems

[rustic hamlet]

peoples faces look mess up now

[young obsidian]

This is awesome sauce

[rustic hamlet]

lol

[young obsidian]

Did that change your mind about Dall e 3?

[rustic hamlet]

what was the prompt for that?

[young obsidian]

Image of a centaur giraffe with the arms at body length so the guy has a very long neck

[rustic hamlet]

okay

[young obsidian]

other result

[sick pecan]

if its a centaur shouldnt it be half/half? thats more like 80/20

[young obsidian]

That was good

[sick pecan]

inline painting is your friend

[rustic hamlet]

i improved the prompt and i got this

[sick pecan]

thats even better

looks straight out of a fantasy book

[rustic hamlet]

this is the prompt "Create an image of a centaur giraffe with a long neck. The centaur should have the body and legs of a giraffe, but the upper body and arms of a human. The human arms should be proportionate to the giraffe's body length, emphasizing the unique fusion. The scene should have a fantasy theme with a majestic, ethereal background. Use a pastel color palette with soft blues, greens, and purples. The lighting should be gentle and diffused, creating a serene and dreamlike atmosphere. Add intricate details to the centaur's skin and fur, showcasing a blend of human and giraffe textures."

[rustic hamlet]

i know right

[sick pecan]

aww it refuses female version

[rustic hamlet]

i wonder why

[young obsidian]

I like mine

[sick pecan]

yours could be another episode of skibidi

[narrow viper]

@young obsidian You can't select the parts of the sentence that you prefer and ignore others. It says "putting into service for this specific purpose, or the use of". This is one of the reasons why it's been a legal challenge.

[chilly tree]

ChatGPT kicked me out a half hour ago, asked me to log back in, and since then has told me something went wrong "misconfiguration in the system or a service outage". Anyone else seeing this?

[young obsidian]

I misunderstood your first argument there, my apologies.

However, I think there might be an issue with the use of commas in the legal text.

In article 5(1)(c), [they prohibit] "[T]he placing on the market, the putting into service or the use of AI systems [...]"

• The placing on the market,
• The putting into
  • service
  • or the use of
• AI systems

That's my interpretation of the text, and I think the same would apply to art.5(1)(f)

The prohibition covers such systems in workplace and educational institutions, individual use outside these contexts is not addressed in this provision.

[sick pecan]

so if the individual then turned around and used it at a workplace or educational institution... who is at fault?

[young obsidian]

Noone

[sick pecan]

my assumption was potentially the individual and the corp. but if you say no one, then that means the lawyers win

[young obsidian]

It's not a crime for an individual to use the service.

[sick pecan]

not all lawyers are criminal lawyers ?

[young obsidian]

These are regulatory frameworks and it only applies to the corporations who provide the service

Natural persons are not sanctioned

[sick pecan]

sure. but seems unrelated. starting to feel like talking to a bot

[young obsidian]

This is what it's like talking to a law student

[young obsidian]

Basically, if you use it at work it's not a problem, but if your boss makes you use it at work it's his problem.

[sick pecan]

its not a legal problem. but its potentially still a problem

[young obsidian]

Why?

[sick pecan]

why have regulation in the first place if not addressing a problem?

hence why i say the laywers win. they get paid to figure this out

[young obsidian]

It is addressing a problem though. What if your boss used AI to measure your emotions during work to see if you aren't motivated enough for the job?

[solid bronze]

Open AI site is down ?

[sick pecan]

your question is posed like you assume this is a bad thing?

sounds biased. your lawyer training hopefully beats that out of you

[young obsidian]

What?

[young obsidian]

This is the intention of the EU act on AI

[sick pecan]

it has the potential to be abused. but lets be honest, most ppl are working a job they hate, and if the machine could better allocate us to jobs we enjoy, that would be an overall net positive

[young obsidian]

It's not bias, it's understanding of the intention of the legislators

[young obsidian]

That is not what the EU legislators think

If you disagree you can vote in EU elections

[sick pecan]

maybe. or maybe they just want some consideration

i dont think measuring emotion should be outlawed. its what they then do what that information that needs to be regulated

[narrow viper]

Hehe, I go AFK and miss all the fun. 🙂
Thanks for the notes @young obsidian I think all of this is philosophical anyway since OpenAI reached an agreement with EU yesterday. We'll see what happens.

[native otter]

So uh, I think the limits may be calculated unfairly..

Yesterday, I had 15 minutes left of Advanced Voice Mode chat time, but didn't use it, now over 24 hours based and it still says 15 minutes left, so seems like it didn't reset my limit...

[sick pecan]

did you refresh the page?

[native otter]

I completely killed the app then reopened it multiple times

[sick pecan]

damn

[native otter]

Like this must be a bug right, otherwise this is super unfair..

I don't say they have to accumulate left over time, but just reset as normal.

[sick pecan]

maybe its just a ui glitch. give it a go for 15+ minutes?

[native otter]

Yeah I will see if it will still let me use it after 15 minutes 👍

[native otter]

After finishing I got this, so it says 7 minutes..

Still weird tho as over 24 hours passed anyways...

[unkempt gyro]

iOS app got slightly redesigned it seems?

[dense halo]

What is the current limit?

[sick pecan]

some say 60 minutes a day. some say 90

[dense halo]

90 minutes??👀🤔

[hidden peak]

hi, i found an issue today, when i use the voice feature to talk in any gpt in gpt store, i can't hear the sound. but when i switch to the advanced voice feature by starting a new conversation with chatgpt, the sound is no problem to hear. does anybody know what the issue is?

[sick pecan]

that is standard voice mode

[hidden peak]

yes,

but i couldn't hear any sound

[sick pecan]

advanced voice mode doesnt work with any gpt or tools

is perhaps the biggest disappointment

but i dont wanna trigger that dude before who got upset about day 1 critisim

[hidden peak]

does this mean now only advanced voice mode can be heard?

[sick pecan]

you can only use it in a new chatgpt4o vanillla chat, no gpt

[hidden peak]

but all of the gpt in gpt store used the standard voice model, does this mean we can't use the voice feature in gpt anymore?

we can only use the advanced voice mode to talk and heard?

i tested it yesterday, it can work, but today, it won't work anymore. when i use the voice feature in gpt, i can't heard the sound

[sick pecan]

you shoudlnt even get that far

[hidden peak]

what does this mean?

i need help, i want to use the voice feature in my custom gpt😢

[sick pecan]

i also want advanced voice in custom gpt. that would be 10/10. but they crippled it

as soon as you @gpt the advanced voice mode feature becomes disabled, at least for me

[hidden peak]

😭

[eternal rock]

`` Server Tip: ``

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

[sick pecan]

advanced voice mode + custom gpt = world changing
advanced voice mode by itself = novelty

[hidden peak]

just want to be clear, so we can't use the standard voice mode after the advanced voice mode enrolled?

[sick pecan]

means we got 3-6 months to build a DIY solution that outperforms

[hoary laurel]

I am using an Android phone, and now there is a bug in the mobile app. When I turn on the voice mode, the app crashes.

[hidden peak]

does this also work for the free plan users? i saw the advanced voice mode is only available to plus users, can the free plan users use the standard voice mode?

[hoary laurel]

This problem occurred after I installed the latest beta version

[sick pecan]

free users dont get advanced voice mode

[hidden peak]

yes, can they use the voice feature in the custom gpt?

[sick pecan]

no idea. i always be paying

[hidden peak]

i just built my custom gpt yesterday, it's about the ai audio tutor, but today i just found out the voice feature can't work in gpt anymore. 😭 😭

[sick pecan]

i have 20+ custom gpts accessing all my apps. and i can only use them with standard voice mode

[hidden peak]

but with the standard voice mode, can you hear the voice sound when you tap the voice feature?

i couldn't hear the sound. and if i switch to the advanced voice model, it works no problem

[sick pecan]

maybe you have audio routing on your phone?

on my samsung you gotta swipe down and go to Media Output

in any case, sounds like your phone is broke

[hidden peak]

no, because when i switch the advanced voice mode, i can hear it with no problem.

[sick pecan]

thats so weird

[hidden peak]

i could use it yesterday after the advanced model enrolled on my phone, and i couldn't hear the sound by using the standard voice mode

[hidden peak]

can you use it with no problem? the standard voice feature?

[sick pecan]

yeah advanced falls back to standard no issue

and with standard i can use custom gpt. so tbh its kinda better right now

[onyx roost]

Ummm... I updated my app and lost access to advanced voice mode.

[hidden peak]

strange.

[onyx roost]

I still have plus

[hidden peak]

why can't i heard the sound if i use the standard voice mode in my custom gpt?

i can only use the advanced model to talk to chatgpt4o

when does the advanced voice model can be integrated into the custom gpt in the gpt store?

[onyx roost]

Now I can't even use regular voice mode

[hidden peak]

To ChatGPT support team: i can use the voice feature to talk with any custom GPT in the GPT store, but the sound can't be heard. there is an issue with the sound output. and on the UI interface, it shows that I start a new conversation with the advanced voice model, but if i switch to the advanced voice model, it will transfer me to talk with the ChatGPT 4o, not the custom GPT in the GPT store. the version of my app is Chatgpt for iOS, the newest version. @languid valley

[viral gazelle]

When will ChatGPT-o1-preview be able to upload documents and such

[hidden peak]

i need to use the voice feature in my custom gpt, please😥

[onyx roost]

Has no one else lost access?

[tawny imp]

I’m plus but never seem to get early access

[onyx roost]

Voice mode is now telling me there are no limits to advanced voice

In Advanced Voice Mode, you're not limited by time! Feel free to enjoy our chat for as long as you'd like.

[mighty dirge]

Here is what I posted in the bug reports about this voice issue in custom GPT's : Description:

I've encountered an issue with the standard voice function while using Custom GPTs in the ChatGPT app on iOS 18 (on both phone and iPad). The standard voice feature was working fine in Custom GPT conversations earlier today. However, now, after I provide a voice input, the system hangs and does not respond with voice output. When I close the voice interface, I can see that my input has been transcribed and the GPT has responded in text, but there is no audio playback.

This issue appears to be isolated to Custom GPTs, as the standard voice feature works perfectly in regular GPT conversations. The problem seems to have occurred suddenly today, despite everything functioning properly earlier.

Another user I’m working with has also reported the same issue.

Here’s a detailed description of the issue using the suggested format:

ChatGPT shared link (optional): N/A

Steps to reproduce:

Open a Custom GPT conversation in the ChatGPT app on iOS 18 (on either phone or iPad).
Use voice input to interact with the Custom GPT using the standard voice function.
Wait for the system to respond.
Expected result:

The GPT should respond with voice output after hearing my input using the standard voice feature.

Actual result:

The system hangs after hearing my voice input. When I close the voice interface, I can see the response in the text window, but no audio playback is heard.

Additional information:

Platform: ChatGPT app
OS: iOS 18 (both phone and iPad)
The issue is specific to Custom GPTs. The standard voice function was working earlier today, and it continues to work in regular GPT conversations.

[sick pecan]

tldr: wall of text

[viral gazelle]

Custom GPTs still use GPT-4... not 4o, or o1.

[sick pecan]

you sure they dont use 4o?

[viral gazelle]

Positive

[sick pecan]

(and if so, source?)

[viral gazelle]

Custom GPTs havn't been updated with new versions since GPT-4. Create a customGPT and ask it directly if it's using the 4o version. it will even tell you it has no idea of 4o

[sick pecan]

my experience is that gpt4 cannot do more than 200 SLOC, gpt4o can do 400+ SLOC. and gpts can do 400+ SLOC ergo gpts use gpt4o

[brisk path]

they use 4o bro. you can switch it to 4 turbo if you mention it in a 4 turbo chat, but otherwise it's 4o

[viral gazelle]

nope... tried it for months... cannot switch.

Your thinking of playground or API... this is the chatGPT disscussion, not the API or playground environment.

I don't like 1o-preview as it won't let you upload documents

[honest jay]

LLMs are not reliable self-reporters - they don't know their own version when asked.

[honest jay]

Right - that's the kind of question they're not reliable for. There aren't exceptions to the knowledge cutoff date - that cutoff date applies to everything, it doesn't "know itself" implicitly beyond that

[viral gazelle]

Custom GPTs also don't have access to the memory feature, which can only be used by 4o... even 1o doesn't use the memory feature yet.

[honest jay]

They just haven't rolled it out for GPTs yet, not due to a model difference https://help.openai.com/en/articles/8983148-does-memory-function-with-gpts

Memory will eventually work per-GPT, according to OpenAI, and that implementation isn't yet ready

[viral gazelle]

Why don't you prove that custom GPT in chatgpt (not playground or API) uses 4o

[honest jay]

https://openai.com/chatgpt/pricing/
Free accounts have limited access to 4o and GPTs. No access to 4. GPTs don't run on 4o for Free, and 4 for Plus

Also various mentions of 4o here: https://help.openai.com/en/articles/8554407-gpts-faq

[eternal rock]

Get notified for server events!

Click the button below to pick up the <@&1261377106890199132> role.

[viral gazelle]

That link doesn't mention 4o in regards to custom GPTs... just as the ability to use it as a plus member.

You cannot create a GPT in chatGPT that uses the GPT-4o model or 1o-preview model. You can use GPT-4o, but cannot create a new GPT using anything but "GPT-4" model.

I am talking about CUSTOM GPTs, not regular use as a plus member.

[sick pecan]

gimme advanced voice mode custom gpt !@!!!

[frank wolf]

Why is gpt down ?!?

[frank wolf]

@deep flax

@steady holly

@bitter latch

Mods what’s going on?!

[steady holly]

https://tenor.com/view/you-rang-gif-20538562

[bitter latch]

https://status.openai.com/

[frank wolf]

Why is chatgpt down

[steady holly]

We don't have anything to do with ChatGPT, the mods oversee the discord.

Check the link Toror sent

[bitter latch]

ChatGPT seems up and available, try using a different browser or device

[frank wolf]

Im getting authentication token has been invalidated

[bitter latch]

Is this from the API or website?

[frank wolf]

iPhone app

[bitter latch]

uninstall and reinstall

[frank wolf]

I had to reset password…

Stupid thing could’ve just said that

[bitter latch]

👍

[dense halo]

Is it just me or does the sol voice sound different (more robotic) in the actual voice conversation than in the menu where you can choose the voices via listening to a sample

[tight island]

Don’t ping mandalorian, He’s absolutely merciless

I’m sorry Toror don’t ban me

My record is clean

[bitter latch]

For now 👀

[tight island]

Ish

[echo copper]

I got a message in the app about the new voice capabilities. But I can't interrupt it for example. I'm in the EU, do I have the new voice chat now or not?

[sick pecan]

if it's a white circle is standard. if it's blue then it's advanced

[tight island]

My circle is the color of stars

[bitter latch]

I don't believe advanced voice works in the EU due to regulations

[tight island]

Yup GDPR regulations

Typical of EU

[dense halo]

Based eu

[tight island]

They coddle us like we can’t read things

HEY LOOK A CONTRACT “ I’m just gonna meaninglessly sign it just to sign it “

GDPR kicks in

Suddenly I have rights

It’s like bit defender

Contracts that tell me I can’t sue companies I sign with are nullified

[inland violet]

The EU and its reguatory are so bad. I cant understand it

[tight island]

I completely understand it

[tight island]

You from the US?

[inland violet]

No from germany.

[sick pecan]

the corportism world is so bad. i can't understand why ppl aren't more like the eu.

[inland violet]

And i have to get an US VPN to try advanced voice mode so bad

[twilit lintel]

hello 👋

[tight island]

GDPR is good because when you sign any kind of Disney contract you can’t sue.

Well those contracts are illegal in the EU, therefore regardless of the contract I can still sue

[sonic comet]

They just being reckless with any and all decisions that they are making.

[tight island]

And if an EU person doesn’t have the funds to sue, they can file a direct rights complaint to the EU where they will sue for them, and then the person gets compensation

It’s actually neat

[inland violet]

yes, and then you wonder why people are so angry

[tight island]

I personally exploit GDPR to my benefit

Of course within the framework of the law

[inland violet]

I don't want to open a political discussion here, but it makes you wonder why people vote AFD in Germany

[tight island]

Sure it ain’t due to your insane crime rate

[echo copper]

What is regulated here though? What does the new voice do that the old one didn't?

[tight island]

It temporarily stores received information

And because it has vision due to data privacy laws in EU countries if you’re caught in the background

[sick pecan]

the new voice can read emotion

[tight island]

Pseudo mind reading

[sick pecan]

which feeds into social manipulation

[tight island]

Yup

People are quite easy to manipulate

Show them a million dollars and see what happens

[raven sand]

Figure I'd leave this here.

[twilit lintel]

😮

[raven sand]

Figure I'd leave this here

[sick pecan]

you are using standard mode

or maybe it's so good at tone detection it saw through your acting

mattvidpro got it to elicit the correct tone multiple times

[raven sand]

I'm literally not using standard mode

Interesting. Guess I must be an idiot. I'm sure mattvidpro's tone was just so much more human. How was I acting? I gave it an angry tone. it didnt pick up on it. it couldnt describe my voice at all. it doesnt hear your voice. its text being sent to it. I'm very much using advanced mode. My webcam didnt pick up on the cloud but its there...its advanced mode....they're straight up lying to all of you but okay bud

[eternal rock]

This is a Safe for Work Discord Server

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

Posting NSFW content will result in a permanent ban.

[open edge]

So people from EU can’t use the advanced voice mode?

[normal quarry]

nope

[tardy totem]

It's amazing that openai don't choose to just follow gdpr, instead of needing a separate version for EU.

[sick pecan]

Here are some key points that highlight why OpenAI cannot simply follow GDPR without needing to adapt or create a separate version for the EU:

1. Data Collection and Processing...
2. Data Subject Rights...
3. Automated Decision-Making...
4. International Data Transfers...
5. Regulatory Scrutiny...

[sacred aspen]

Is the teams advanced voice quota larger than personal? Personal id reckon is about 45min - 1 hour

[scarlet bloom]

am i able to use voice mode on pc?

[lusty jay]

did i just get a response from o1?? (the right response says it thought for a few seconds but im a free user)

[opaque parrot]

Soooo we’ve been paying for months hoping to get early access to it, then looking forward to access when “all plus subscribers get it” as they promise in their announcements only to then get slapped in the face. Nice

Surely they should have known that EU subscribers won’t be getting access to advanced voice mode before they announced that “all Plus users will be getting access to it this week”

[sick pecan]

we demand the CTO resign!!! (ohwait)

member when the openai cto promised advanced voice mode "in the coming weeks" (lol)

[half locust]

So if I buy chatgpt plus today (and i'm in an eligible region), I'll get the advanced voice mode or am I going to be duped the same way as EU users

[sick pecan]

good question

[young obsidian]

Which country are you in?

[half locust]

india

[young obsidian]

There is no restriction as far as I know

So you'll get it, assuming it's been rolled out to all users.

[half locust]

yeah that assumption is what I'm trying to validate

cause 1 month of chatgpt plus is more expensive than 6 months of internet cost here

and I have openai API access so advanced voice mode is the only reason I'm interested in plus

[young obsidian]

You will get it this week

I recieved it within the first hour btw

[half locust]

oh

[young obsidian]

So it's likely not a late rollout

[half locust]

Got it. thanks

[young obsidian]

Here it's approx half of the monthly price for internet

No about ⅙th actually

Damn internet is expensive

[half locust]

I paid 40 dollars here for fully unlimited (and I mean truly unlimited) 5g data for the whole year

[young obsidian]

I pay 50 per month

For personal

But wifi costs 150 per month

[half locust]

yeah internet is dirt cheap in india. and speeds are catching up to europe / us if not already better

[polar jungle]

Guys, do you ever tried to use 1 account premium chatgpt using 2 device and using voice mode at the same time?

[zenith lichen]

Can o1 do image gen or browse internet?

[feral zinc]

no

[rose spire]

Why does advanced voice mode have some distorted sounds?

Like Bssskk

[weak prairie]

She really gets crazy 💀

[rugged igloo]

lol wow thats crazy

[weak prairie]

Yeah…

do you know the limit? Is it like 60 mins per day?

[solemn eagle]

Its 60 min

[rugged igloo]

yea it is

[solemn eagle]

Then wait 20 hours

[weak prairie]

Okay gotcha

[hidden charm]

Hi there guys, is there any "experts" on NLP here?

[solemn eagle]

Having too much fun with AVM.. doing Arabian, Jamaican, African and all sorts of accents.. makes me laugh.

I got it to pretend to be an Indian scammer 😆

[rugged igloo]

LOL"

[solemn eagle]

I got like 10 mins or more of just random footage .. thinking to just make a YouTube chanel for it but cant be bothered

[weak prairie]

That’d be fun to try lmaoo

[solemn eagle]

I'll clip this one

Indian Scammer

Man theres so much more lol.

[drowsy wagon]

Finally

[solemn eagle]

Enjoy your 60 mins lol... it goes by real fast.

[drowsy wagon]

60min/day?

[solemn eagle]

Yep

[drowsy wagon]

Damn

[solemn eagle]

It flies so be careful... gotta wait like 20 hours for reset.

[drowsy wagon]

I mean it just got launched this week, I expect the limits to be relaxed soon

[solemn eagle]

Yep, for sure. I think they are just testing to see how it handles everyone using it.

[weak prairie]

Lmaooo

[solemn eagle]

Man why do I get other peoples messages in my account lol

Happens so often.

[feral zinc]

account compromised

[solemn eagle]

African woman scared of a Lion

[young obsidian]

😭

[solemn eagle]

I hate the daily limit so much .. was creating a cool gangster scene lol

[young obsidian]

I hate how they blocked VPNs

Us EU users won't have access for at least another 6 months

My guess is it won't ever get released here

I think there’s about a 50% chance that OpenAI will modify the feature to comply, but it would likely take at least 6 months.

[sour wasp]

The new voice mode is alright, but for slow talkers it's pretty tough to use. I tend to pause a lot when talking, and ChatGPT would start talking thinking I've finished when I haven't.

[sick pecan]

hold to talk in standard mode is underrated

[ocean rampart]

8 Hours with this guy today

[solemn eagle]

Mine is always the same 20 hours

Maybe they are changing it now slowly.

[ocean rampart]

openai plays with us!

[ocean rampart]

or maybe this new and next time you'll get 8 Hours

[solemn eagle]

Yeah probably.

Im trying to upload a video where I got it to playout gangster scenes lol

[quasi ice]

Hey, is there someone here who knows a good Video Upscaler? I made a 3D animation but its only in 720p

[brisk tapir]

say... Topaz Video AI

[cobalt idol]

is the advanced voice not supposed to be available in EU?
just read article about it, but got it like 2 days ago in EU 🤷‍♂️

ah right its gone again 🤣

was still there yesterday

[feral zinc]

the app shows the small notification regardless where you are and if you have a paid sub

[cobalt idol]

had no time to try it dammit 🤣

[brisk tapir]

weird, seems certain voice types like Maple cannot access Memory

[quasi ice]

Is there a free version?

[brisk tapir]

no...

[young basin]

I was just sitting here with voice mode running as I was typing and it suddenly just asked me how my day was going. I checked the log and it said 'Transcript unavailable' for my own side of the chat, which must have prompted the voice to respond. For a moment though, it really felt like having an autonomous ai with me hahahaha. I just put a screenshot in #off-topic

[sick turtle]

anyone else advanced voice mode made the app laggy?

[winter elbow]

Niceee, circle???

[solemn eagle]

?

[hardy current]

Does anyone have any examples of custom gpt for a sort of “work journal”? Thinking while at work dictating thoughts, issues, progress and having it sort them in some way.

[near breach]

The o1 model stops reasoning, even worse, I find that all of the models have reduced capacities to gpt 4o-mini. Why should it happen ,and how to solve the problem

[feral zinc]

that's what its like currently. nothing to solve

[near breach]

Oops

sounds like no one can use o1 now

i met the problem three days ago，and changed a browser to solve. but nothing is useful now

[solemn eagle]

Cant wait for ChatGPT on Windows ... then I can just talk to it with my headphones and then when vision/share screen releases, gold.

In about a year everyones gonna be talking to their phones and pc's all day everyday lol

[hard dew]

Can this bot stop writing into memory EVERY SINGLE damn thing I say???? Has anyone else noticed this for the past few days? It doesn't matter what you say or how you say it, the bot puts it into memory as "Mentioned that..." and whatever else.

[near breach]

You may want to clear its memory

[hard dew]

That is obviously exactly what I have been doing every single time it's writing useless garbage in there, and not stuff that I actually and explicitly tell it to remember.

I noticed also that it seems to occur / become worse as the chat you have gets longer... for whatever reason.

[near breach]

That may because you met the same problem just like me. The model capacity is set to GPT 4o-mini which cant satisfy your requirement

[hard dew]

I didn't mean it as it doesn't memorize stuff when I tell it to do that, it does, but it also started spamming complete trash in there I swear literally every single exchange, regardless of if I do or don't tell it to. And it seems to occur / get worse as the chat you have with it gets longer, for whatever reason.

[swift zinc]

ARE NON-PAYING USERS NOT WORTHY?

[dense halo]

Honestly, the advanced voice mode just feels so different, I kinda prefer the old one better. I feel a bit scared cuz it just sounded and feel too real. >.<

[torpid kelp]

Any news on when the mac app will get advanced voice?

I got it on phone but doesnt work on mac app

[normal quarry]

no idea, I am waiting for it also

[torpid kelp]

rip

[celest topaz]

Are you guys able to login?

[near breach]

not fixed yet？

[feral zinc]

?

the model just cant count properly, if you try 10 times it may or may not get it correctly

[olive lynx]

AutoGPT is so slow

the previous one is better, why did they even merge the model with a refresh button, its so slow. and i always get so many errors

[olive lynx]

Just turn it off at the settings, The memory could be turned off, a switch is there.

i used to have that problem too, but i found the memory switch

[hard dew]

It's still going to make use of the actual memories already stored within though, right?

[crystal tapir]

Is it just me? Using AVM turns my phone super laggy.

[uncut vale]

Hi

[feral zinc]

https://discord.com/channels/974519864045756446/1288220468373098496

[crystal tapir]

Nice. Thanks!

[sacred aspen]

Does team have higher limits for advances than personal?

[azure sentinel]

anyone else have chatgpt advanced mode create these hallucinated memories? only the beginning part about learning the piano is true, everything else is hallucinated and have no idea where it came from

[vocal cipher]

Chatgpt 4o is thinking now? Im the only one getting this responses?

[burnt shadow]

Anyone else missing the “redo” or “retry” button in generations?

It used to be next to the thumbs up and down button but I am not seeing it prompts generated anymore

[feral zinc]

it got removed from custom gpt's

[worthy lynx]

Agreed, I've noticed it also does not do well with children's voices. My kids really want to talk to ChatGPT, but because kids are slower than the average adult at speech, it tends to misunderstand or interrupt them.

[pulsar carbon]

hi guys i tried advanced voice it's pretty amazing

[crisp cairn]

interesting

[burnt shadow]

hmm weird, I dont have it even on basic ChatGPT. When I ask it to generate an image, there is no longer the redo options there either

[burnt shadow]

Also why would they remove it? I guess to reduce the # of generations at once time? or as often?

[sonic comet]

No GPT-4o cannot use Chain-of-Thought. You are seeing probably because you regenerated the response with o1-mini or o1-preview.

@deep flax Can o1-mini be used in the chats where the attachment exists?

Or is OpenAI testing the attachments feature for the o1 models and it led to the glitch?

[feral zinc]

free user? i have it in other modes with plus

[burnt shadow]

Nope paid user. I am a plus user

[burnt shadow]

Example I just typed in, you can see there is not redo or generate again option under the image

I also put in a bug report, but have heard nothing back yet

[blissful yew]

Virtual Pretend Singing

[burnt shadow]

may I ask how to enable this? I dont see a way to enable the new voice feature

[blissful yew]

You need gpt plus, also u can't be in the eu

[burnt shadow]

I qualify for both of those lol

I am plus, and live in USA

[blissful yew]

Well you should have it

Did you update your app?

[burnt shadow]

Yeah no updates on the App Store

[floral echo]

what if we live in earth

and have subscription

[vocal cipher]

well im positive of that I was using chatgpt 4o and it display the same reasoning af 01-preview. it stated, even, the time it took to process the request of the excel formula

[dense halo]

i dont understand why dont they let it sing?

like what is that going to effect? talk about censorship overload

[blissful yew]

Everyone should have it who has a subscription except people in the eu

[blissful yew]

Cuz if it sings a copyrighted song is their point of view

[onyx roost]

Gemini live is now available for everyone regardless of if you have their subscription. Open ai needs to up their game and increase advance voice usage. 1 hour limit is not enough time to make it useful

[feral zinc]

https://help.openai.com/en/articles/8400625-voice-mode-faq?q=file#h_94f126e132

[sick pecan]

or.. making it limited is the oldest marketing trick in the book. Google really should be doing the same

it even creates engagement in terms of ppl publicly wanting more. it's genius

[azure sentinel]

advanced voice mode feels so limited compared to what they showed

the videos showed it having personality, the released version is like regular chatgpt with no personality, basically the same as the standard voice mode with less latency

[feral zinc]

it is limited and even the faq mentions that several filters have been implemented

[tough bramble]

That's not how much you've used it, it's when it'll reset the limit. You can use it like 1 hour every day. It resets next day the time you first used it the day before.

[tough bramble]

Yes, most of what they showcased is gone.

[azure sentinel]

its more than filtered i dont care about singing all of the personality is gone

[sick pecan]

including the cto (badoom tish)

[azure sentinel]

its just standard voice mode now with no latency

[sick pecan]

it is more than that

[azure sentinel]

every sentence still ends with "anything else you want to talk about?" sentences like that lmao

[tough bramble]

It will refuse to generate sound effects. (My guidelines won't let me talk about that.)
It will refuse to sing. (Easy to get it to, though)
It cannot differentiate between speakers. It literally can't detect your voice nor tonation.
It cannot detect emotion. (I'm sorry, but I can't help you with that).

The last point makes the whole "EU makes it illegal because it detects emotion" nonsense even more stupid.

Because it doesn't.

If you sound sad, it'll immediately cut you off and say it can't help with that, completely ignoring what you said.

[sick pecan]

mattvidpro YouTuber tested it and it did read emotion many times

[azure sentinel]

whats wrong with detecting emotion anyway literally what is the point of all these stupid restrictions

[tough bramble]

(I guess it technically does detect emotion then, but it'll refuse to do anything with it)

[tough bramble]

Not after the final release... more was available during the closed phase.

[sick pecan]

i am talking two days ago

[tough bramble]

Yeah, no. If it responds to emotion, it's pretending.

Try it. Sound really sad, and it'll ignore what you said. You can sound really sad and say "What is the definition of a proton?" and it'll go "I'm sorry, I can't actually help you with that".

[sick pecan]

arguably everything it does is pretend (llm in general)

[kind kettle]

How will I know that I have access to Advanced voice on my plus account , yet? It sounds the same right now...

[feral zinc]

well... even today people have posted videos of it doing racist voice lines, threattening to cut someone to pieces and cook them, singing copyrighted songs, replicating movie scenes, doing bedroom moaning etc.

[tough bramble]

Except it can't actually detect anything you do. You can speak like a robot, high pitch voice, low pitch voice, and it's unable to tell what you sound like.

[feral zinc]

it can still be tricked to do a lot

[tough bramble]

"Yes, you can actually do that." works every time; for things that is the LLM itself, and not the moderator model.

[azure sentinel]

and what changes then even if it does those things? lol, i can understand placing common sense restrictions to make your product less embarassing but there's no reason to cut out the majority of functionality

no ones life is gonna be ruined by explicitely forcing an ai to say something racist or forcing it to do a moaning sound

nothing changes in the real world

[feral zinc]

copyright law cases, media headlines of chatgpt voice being used to threaten to assassinate a senator, etc.?

[azure sentinel]

so you can already paste that into literally any text to speech software?

[feral zinc]

yeah but do those software sit in the official us government ai board?

[azure sentinel]

regardless it probably wouldnt be a copyright violation to have it sing a song anyway, no more so than it is to actually train it on the songs(how else did it learn it) or have a text based llm spit out lyrics

[azure sentinel]

regardless i didnt say there shouldnt be any restrictions

you shouldnt be able to easily feed it stuff to threaten people i dont think anyone would defend that but i dont think it changes anything at the end of the day even if it did considering you can already plug the same text into pretty much any text to speech software

[feral zinc]

ai is not a person singing in a shower, its a paid product serving people and subject to licensing fees even if its karaoke quality performance

[azure sentinel]

whats the difference between that and a text based llm spitting out copyrighted lyrics or having a text to speech read copyright texts or books or anything like that lol

[tough bramble]

After saying it couldn't rap and do beat boxing at the same time, I just said "Yea, you can actually do that.", and it just does it. Every time.

[azure sentinel]

theres not really any difference chatgpt isnt as restrictive as something like claude which takes it to the extreme and you cant even tell it to edit any lyrics

[feral zinc]

try to make it say copyrighted lyrics and you get a warning, the normal text version

[narrow viper]

The world is still shaking because AI can write high quality stories and prose. Then the images. Then there are the videos. Singing is going to happen but the world needs to catch up to the tech, get a grip on what it all means and how to use it.

[tough bramble]

I have a feeling they autotune your voice input to a single tone, preventing it from detecting tonation. Because it definitely can't do that.

And it can't repeat what you said back to you in the same way you did.

Ask it to sing a song you just sang, and it uses a completely different melody.

[azure sentinel]

at the end of the day its the users decision and whatever users make they should be the ones responsible for whatever the consequences are, no one thought of suing adobe for the users ability to make fake/illegal pictures using photoshop

ai regulation to me feels so forced for something that is not generally intelligent yet

[tough bramble]

ChatGPT singing the song I just sang to it, but using a completely different melody.

[azure sentinel]

theres nothing you can currently do with ai that you couldn't do previously before llm's came out

[feral zinc]

openai has made official commitments to ai safety to different governments around the world

[azure sentinel]

imagine if every piece of human innovation was this regulated and looked over and overthought to this degree, probably movies faults for convincing people that ai's are gonna rise up and take over the world as if an LLM that has no thought or memory outside of its current context or ability outside of a chatbox/voice system is gonna be able to do anything dangerous, unless you consider being possibly forced to say mean/sexual things "dangerous"

[sick pecan]

they proved it can be done. someone else will make it better

the beauty of a free market is that this is practically a certainty

[azure sentinel]

i'm just glad china is a thing honestly give it a few years after ai is completely restricted and locked down in western countries you'll see a lot of development out of china, even right now there are models that are almost as good as SORA that open ai refuses to release for ridiculous safety reasons

[worthy lynx]

How can I get the mathematical notation from ChatGPT to paste into a Word document and maintain the notational formatting?

[feral zinc]

yeah openai dall-e 3 is super nerfed but there are so many different services today that are far less censored, free and paid

[azure sentinel]

who would think the super authoritarian country would have less restrictions than supposed "free" western countries

[sick pecan]

member sora? now we have three viable alternatives

[feral zinc]

sora will probably never happen or will never allow image to video due "safety" but there are already half a dozen of services that do it

[azure sentinel]

and none of them seem very dangerous so what really are these safety concerns

lol

even gpt4o cant have its image generation tools for some reason why is it still using dalle-3

what happened to gpt4o being an omni model theres nothing omni about it except for advanced voice mode which has had most of its features cut

[feral zinc]

yep 🤷‍♂️

[azure sentinel]

are they not releasing these to make gpt5 more enticing?

perhaps a backup plan in case the larger compute and training data in gpt5 doesnt have the big effect they're expecting

still being able to release it as the first fully released omni model

[sterile socket]

o1-preview is unreasonably good at Python code refactoring! Not perfect by any stretch.

I learned new things from it, but the really good suggestions came from colleagues

[azure sentinel]

tried using it for C++ refactoring and it is unable to do it especially for large files

hallucinates heavily still

[sterile socket]

Please add a feature to scroll fast to previous/next user prompt! It’s too easy to lose the big picture when the responses are so long.

[azure sentinel]

what i hate most about advanced mode is every single response always has to end with a stupid question

it breaks any immersion that youre not talking to an ai which is probably why they added it

[spice star]

there won't be a GPT-5 afaik, o1 (non-preview) is intended to be the next-gen model

and if I were to guess they're not releasing the native image output because of moderation concerns

a lot easier to put guardrails on image generation if you can see what's happening in plain text

[narrow viper]

I believe Custom Instructions are used with AVM. Try adding into CI that you do not want responses to end with any kind of open follow-up, like "I'm here when you need me" or "Let me know if I can help with anything else". Be explicit.

[brisk path]

Not true. The GPT series will continue alongside the o series

[narrow viper]

Yeah, o1 is significantly different and better, and costs much more. That's why I just tossed out this open request: #gpt-models message

[modern frigate]

Hello world! We've been getting this message for two weeks now and after the deadly message comes up, our prompts are deleted. Can anyone offer any thoughts/ideas/solutions? Thank you!

[narrow viper]

I've seen that a lot too. I posted a bug report recently for a similar issue. Do you find that the message comes up when you edit a prompt after a response?

[modern frigate]

Thanks for your message. Actually, no, just on original/new prompts.

[narrow viper]

Might it have to do with your content? The filters often don't come out and say "I'm sorry Dave, I can't do that". There is often simply an error.

[modern frigate]

I doubt it since it's straight forward grammar checks and such, nothing wild. Thanks for the insight. I will send bug report now. Appreciate it!!

[narrow viper]

Check this out: #1288534015179227270 message
You might want to add a note that your context is destroyed even without an edit. The issues might be related. HTH

[candid badger]

is there still no way to have a temporary chat with voice?

[elfin lava]

Why would u want to do temporary chat with voice?

[native otter]

I made AVM act like it was in a heated discussion about football vs basketball, where it responds to it self as the other person..

"Both" Would have slightly different voices and interrupt each other.

[winged willow]

AVM is not available with custom GPT's https://help.openai.com/en/articles/8400625-voice-mode-faq

[azure sentinel]

I doubt it since o1 is seemingly just a finetuned gpt4o with chain of thought reasoning, its not a next generation model by any stretch nor is it even a replacement for gpt4o

[spice star]

o1-preview isn't the full o1 model

[azure sentinel]

its still based off of gpt4o as far as i know @spice star

regardless even with more parameters quality wont improve much for things that really matter i dont think, like coding for example is already better with the smaller o1 mini model

[sonic comet]

Then maybe it's a glitch similar to what I experienced.

[solemn eagle]

ChatGPT has a breakdown

[tight island]

USE HER, NOTHING IS wrong with her!

Claude is NOTHING

[nova fable]

Lol

[tight island]

I think GPT is good for coding once we have agents, meaning a single model for each and every coding language

Not a huge model for everything but many smaller models working together

[tight island]

Yup that’s true, try the api it’s actually insane

[tight island]

@drifting ruin

It’s time

[drifting ruin]

well damn, someone jailbroke it fr fr

lol, my gpt page tells me to get an app on a mac or android, and shows me a link

[twilit lintel]

me

[tight island]

It’s time to prep, we must choose sides

I’m choosing the side of AI

[ruby kite]

welcome to the club...
i see this too and did not got advanced voicemode...
now i am some sort of sad :/

[drifting ruin]

wait, are you not a paid subscriber?

[ruby kite]

i am ...

[drifting ruin]

because if you are, you should have it, rollout ended on the 25th

[ruby kite]

thats the sad...

[drifting ruin]

I just saw it on sama twitter page

weird

[drifting ruin]

I choose the side that keeps me alive

[twilit lintel]

👍

[ruby kite]

i recheckt my app now...
it say gpt plus...
so i have a plus membership...
and in voicemode it says you dont have currently access to it but check againg later for updates..

[drifting ruin]

weird

have you tried updating the app?

that's what did it for me

[ruby kite]

i did i installed the app entire new about 10 h ago and installed the last update in 2 h ago

[drifting ruin]

speaking of app, when is the windows app showing up!?

hmmmm, where are you from?

[ruby kite]

germany...

[drifting ruin]

EU won't get the app for a while

ah, that's what it is

want a tip?

uninstall the app, and try through a vpn

[azure sentinel]

i dont think thats how agents work

wouldnt need their own model for each language

agents would work on larger projects refining their use and fixing any bugs etc without human input

whatever you would be able to do with agent based workflow the average user could already do themselves even if they have no knowledge of the programming language they're using as long as they understand how programming generally works

aside from refactoring probably if you dont know best practices of the framework or language you're using

large language models are useless for things like game development generally aside from being very minor helpful companions for things like math at least until we have a truly omni agent based model that can code something, play it, refine it, fix bugs, etc

even then it'll probably still fail and then say it accomplished it correctly a lot of the time due to the nature of these models lol

even gpto1 will spit out hallucinated code and say its correct even if it knows how to "think" beforehand

then you tell it to correct something and it hallucinates more

but at least its phd level whatever that means lol ive found more use for coding with claude 3.5 sonnet

if a phd graduate joined my team with that level of coding skill i'd be concerned about who's giving out these phd's

ai can code but it cant program and thats the biggest factor i see theres a large difference between coding and programming

the days of writing actual hand written code are nearly over and codemonkey jobs are gonna go extinct but programming wont go away since programming/computer science has nothing to do with the actual syntax or anything or actually writing the specific lines of code, but rather in building architectures

i cant see even 25% of programmers 10 years from now though actually writing hand written code on a regular basis

same with how punch cards died out for programming

so will text based coding

[glossy roost]

"new advanced" voice chat is garbage

[azure sentinel]

game dev's jobs are basically safe, high level software engineers and stuff are safe too, but website/app coders are really gonna die out as a job especially on the entry level

[azure sentinel]

true

they nerfed it

[glossy roost]

its the same as old voice chat, but less good

[azure sentinel]

its better than the old voice chat

[glossy roost]

i dont know if it is tbh

[azure sentinel]

better than gemini live too

its good in terms of accents, different languages, etc. its the same model as gpt4o so it will generally have the same responses as the old voice chat but with less latency due to there being no speech to text to text to speech layer and is able to speak very well

[glossy roost]

its just really really lame to be baited with "new multi mode AI! can actually hear you!" but then its still transcription based and still exactly the same. except you can interupt the thing now

[azure sentinel]

but they removed a lot of the cool features "for safety" whatever that means

its not transcription based tho

[glossy roost]

it absolutely is

[azure sentinel]

its been proven not to be

[glossy roost]

its still sending ur speach to a talk to txt api and then reading it

[brisk path]

That's literally not true

[glossy roost]

literally ASK THE THING

lol

[azure sentinel]

it doesnt know its own abilities

[tight island]

You don’t think that’s the point, smaller and more expert models on a select topic can help with accuracy a hundredfold

[azure sentinel]

its still stupid its not real ai yet

[glossy roost]

y'all are wrong, im sorry

[eternal rock]

`` Server Tip: ``

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

[brisk path]

Rule #1 of AI - never ask it for info on how it works or it's own capabilities

[azure sentinel]

not really needed

[tight island]

I have a swarm consisting of 17 individual fine tuned models 10 of them being specifically for individual code library, and one for the actual IDE. Where they can talk on the topic and correct mistakes before the final answer, the others I am not disclosing.

But I can tell you this 2-6 thousand lines of error free and fully functional code

[azure sentinel]

ai is already good at coding which is where the differences between languages generally lie

[tight island]

Although it’s downside is cost

[azure sentinel]

i rarely get issues in terms of syntax or anything

[eternal rock]

Get notified for server events!

Click the button below to pick up the <@&1261377106890199132> role.

[brisk path]

The only "transcription" done is so you can read the chat after the voice session is done. Otherwise, audio is directly sent to the AI and the model generated the audio sent to you directly

[azure sentinel]

its just bad at programming @tight island

not a good software engineer but a good codemonkey

[tight island]

That’s because you’re doing it wrong

🤷‍♀️

I don’t use the web interface

[azure sentinel]

try having it setup to do game engine development or low level unreal engine code/overriding functionality of the engine and i'll believe in the ai's ability as a software engineer

[brisk path]

can we just appreciate how OpenAI lets us use tools and vision with 4o mini inside of ChatGPT now 😭🙏

[tight island]

And I would assume you would have me disclose the project afterwards so you can monetize on it

Forget it

I don’t unlike the majority in here ask gpt for snippets, no offense to any snippet mains

[azure sentinel]

i'm not asking literally im saying its incapable of doing it

due to its inability to test its own code

[tight island]

Ha!

[azure sentinel]

you cant be a software engineer and not test your own code

[tight island]

Mine can test it I didn’t spend an entire year on it for nothing

[brisk path]

AI isn't good at complex, large scale coding right now. It won't be for like 2 years.

Minimum

[azure sentinel]

it can play its own game it developed using unreal engine?

join openai then or sell your product in that case but i doubt its that capable since that requires more than a large language model and goes beyond the scope of even video processing

[tight island]

It can test and run the code, install libraries, and write the code.

You know mate this ain’t a new thing

It’s been around for a year

[brisk path]

What code... 💀

This must be some pretty basic code if it can do that

I don't think LLMs can currently make an entire game in UE and test it, but that's just me

[tight island]

Why do you guys insist on UE

[brisk path]

idk the other person was saying UE so I just went with it

[azure sentinel]

its just an example of something ai cant use

[tight island]

That’s not what I created it for

[azure sentinel]

no one is saying it cant develop websites or console apps and stuff

it can obviously test things that you can test using unit testing

no one would doubt that

[brisk path]

barely

[tight island]

I ain’t using it for games

Such a silly thing

[brisk path]

what are you developing tho

kinda need to know so this discussion doesn't suck

[tight island]

GitHub Repo Generator

[brisk path]

oh

[drifting ruin]

games aren't that silly... imo, look at omniverse, also oai started as ai that was for dota

[tight island]

For the purpose of my swarm it’s silly

[brisk path]

Omni...

[drifting ruin]

that's fair

[tight island]

It’s created solely for mass coding

[drifting ruin]

omniverse is op

I have to say o1 is very good for coding

like insanely

[tight island]

It is

I have been thinking of introducing it into my swarm

But the cost is already high enough to run the damn thing

[glossy roost]

"Additionally, we run our existing moderation model over a text transcription of both audio input and audio output to detect if either contains potentially harmful language, and will block a generation if soL." so its still being run thru transcription before it gets to the user

[brisk path]

OpenAI should really release a "ChatGPT" API or something idk

[tight island]

It’s just parameters that are the issue

[drifting ruin]

chatgpt api? they already have that

[brisk path]

Like, you can just have API calls tied to your ChatGPT Plus usage limits

[tight island]

Ohh that

[brisk path]

So like it's easy for devs to begin using it without having to pay for credits that might not be all used

[tight island]

That would indeed be nice

[brisk path]

Since they scrapped the $5 free trial that comes with every account... Sadly

[tight island]

I think we should scrap custom GPTs

[drifting ruin]

nooooo

[tight island]

But wait

[drifting ruin]

I think we should expand them

[glossy roost]

some of them are really good

[tight island]

Getting to it

[drifting ruin]

just gimme control over ui in a custom gpt

I'm already salivating

also allow me to tell it which parts to edit, and which parts to keep

I'd so get artifacts in there!

[tight island]

I think we should scrap Custom GPTs and then.

Have set amount of Custom GPTs, where let’s say a Plus user can maximum make 3

So when you’re making a custom gpt you can select how many slots it should take. Each slot doubling the context etc.

Then have the option for parameter control. Etc and advanced settings.

[drifting ruin]

that seems like a waste of creativity

why would you limit someone to just 3?

[tight island]

Just an example

[drifting ruin]

nah, no limits on the amounts of gpts, just on usage

that makes much more sense

[glossy roost]

so they can squeeze more pennies?

[tight island]

Exactly

[twilit lintel]

😂

[glossy roost]

ew

i hate that its as true as it is