chatgpt-discussions

[waxen chasm]

If you can see the store there are crazy one there that can code like crazy

[buoyant rampart]

I just had fun with one of those GPTs from the store, had it write me a couple short stories that were random minus the dinosaurs theme 😂

[tacit vortex]

mind suggest some? why can it perform better than original gpt? further training? i got plus, thx

[plucky timber]

They got mad people stole their instruction set and files

(My interpretation) 😂

[dense halo]

lmfao

[pine tulip]

Is that fact?

[plucky timber]

nono im just memeing GM 😂

[pine tulip]

But it would be cool if it really was the reason they removed their GPT. Because then companies will certainly put pressure on OAI at some point to better protect custom instructions

[dense halo]

Are custom instructions really worth that much

[pine tulip]

Since you can soon earn money with GPTs, yes they are worth a ton

There are also a lot of people who don't protect them at all and make them open source by default, but one should be able to choose what he/she wants to do

imo

[plucky timber]

I'd say currently'?
Not really a custom instruction seems to be a "Vague hint to follow maybe" for GPT4

Maybe gpt 5

I am a bit conflicted about paying for someones 5 minute crafts in word

And the 2 minutes they spend on the custom action

[dense halo]

I swear they better not be more than 5 bucks and that's pushing it.

[pine tulip]

Well I crafted custom instructions that change the behaviour of ChatGPT and Dalle, making it possible to use negative prompts for example. I would not want to share that at this point in time.

[pine tulip]

I don't think you need to pay at all

[dense halo]

I mean when the store gets monitised

[pine tulip]

Yea same

[plucky timber]

Its all very user use case specific , idk was just my personal take your free to love them , for GPT 4 i think it was to early and it doesnt seem to follow the instructions all to well . Sure testing and all but yeah not "Ripe for store" imho

[pine tulip]

To cite myself:

My guess is: User X pays 20 bucks for ChatGPT Plus, uses ChatGPT 900 times and your Custom GPT 100 times. Those 10% ($2 accordingly) are then split for example with a 70/30 cut, so that you get 30% of that money, earning yourself about 60 cents from that one user. Just a guess though.

[dense halo]

The store navigation needs to be improved before they monetise, showing only a small selection is unacceptable

[pine tulip]

Yea I think the same

[dense halo]

Canva GPT is back!

[pine tulip]

But on the other hand, they could set custom commands themself and weigh them higher, right?

[plucky timber]

Curious if they added new instructions lets see 😂

🥁

[pine tulip]

lol

[dense halo]

don't do it!!! lol

[plucky timber]

Its just a trade deal fren

The AI Learns from me, I learn from the AI

😏

[dense halo]

Future of economics

[plucky timber]

Version numbers would be nice too

People changing their GPT all the time

but more up to the user in this case , but meh

[dense halo]

A most downloaded section will be nice too

[plucky timber]

I guess that was what trending was supposed to be but from my experience it doesnt line up with the search results

I assume trending are partnered high downloads not sure, but spotted a few non trending having higher numbers than the search shows for the trending ones

[dense halo]

Interesting, maybe its not a live list

[plucky timber]

It's deffo no instant indexing 😂

Types in search

Gets a coffee

results appear
Ah were ready! 😂 😂

[dense halo]

💀

[open bramble]

god gpt is slow atm for me - i can bake a cake in the meantime

[plucky timber]

mines just never ending analyzing

#off-topic 😂

[inland kraken]

99% of them are so easy to replicate. So far, I haven't found one GPT that I reused more than once.

The one with proprietary knowledge or actions have much more potential... If only actions were not so painfully slow.

[tacit vortex]

any GPTs can analyze this discord? who have the time to hook on this?

[lone forum]

Is anyone using the playground assistant as substitute for ChatGPT Plus? This is what I do but I found that it's very slow. Sometimes it takes about 2 minutes to respond. I am using gpt-4-1106-preview. Is this normal? How does it compare with ChatGPT Plus?

[crisp finch]

When purchasing ChatGPT Plus, it stated that I can send 40 queries every 4 hours. The issue is that I've sent around 25 queries, and there was no notification about reaching the limit. Has anyone else experienced the same issue, or could it be that I'm doing something wrong?

[maiden fractal]

It is 40 prompt per 3h for chatgpt 4 and gpts 25 per 3h

Or 1 prompt every 4.5 min

[crisp finch]

As I mentioned, I've only sent 25 queries in the last 3 hours. Any ideas?

[maiden fractal]

Try to wait a bit, disable CI, create new chat try again

[drowsy tinsel]

25 is included in the 40 tho?

[maiden fractal]

For gpts ?

[inland heron]

I'm getting increasingly frustrated and confused as to why general GPT4 is being so terrible at understanding what I want it to do

[clever condor]

is GPT also broken right now for you guys?

Im jjust getting errors and errors

[scarlet delta]

I feel like its reading comprehension and memory has gotten a lot worse. Like it has Alzeimers

[dense halo]

I just want the message limit to increase. 40/3 hours runs out fast, especially over voice.

[wary locust]

Damn it limited me for 90 mins. what am i paying 20$ for kek

Just top it off with making me watch some ads

[maiden fractal]

You can switch for api or for the teams tier

[dense halo]

Teams needs another person tho. API is too complicated for me

[maiden fractal]

Or 2 account 👀

Some people did this

[dense halo]

It’s expensive as hell

[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.

[onyx eagle]

You can use the API in OpenAI Playground. It's not that complicated.

[dense halo]

It’s like 100 bucks a month

Here in Australia

[onyx eagle]

No it's 60$ a month if you choose to pay month by month. It's 50$ if you chose to pay yearly...

[dense halo]

Not in Australia

[onyx eagle]

Ah i see 👍

[dense halo]

Well 60 usd is 89 dollarydoos plus tax so 100 seems right

OpenAI needs to create a ChatGPT plus, plus subscription

[maiden fractal]

It is a big step to offer a between of plus and enterprise

[dense halo]

Wdym

[onyx eagle]

You can buy some credit and then use the API here. It's not that complicated.

https://platform.openai.com/playground

[drowsy tinsel]

yea

u can't do 25 in GPTs and do another 40 in regular

[dense halo]

Too complicated, I’m mentally handicapped

[drowsy tinsel]

don't switch to api unless u want to become bankrupt next month

[maiden fractal]

Before the team only option was plus or business

[maiden fractal]

Depend your usage and how you deal with your credit card

[onyx eagle]

I understand.

[maiden fractal]

Gpts limit is less indeed

[drowsy tinsel]

im using OpenRouter, its nice how u can also experiment same prompt with other models

[maiden fractal]

Don't worry api is also hard for me I tried, I succeed the basic link with discord with tutorial but I'm not good either

[dense halo]

How expensive is it?

[maiden fractal]

You pay per prompt

[drowsy tinsel]

10usd/M input, 30usd/M output

[dense halo]

I live off of disability money

[maiden fractal]

. 10$ *

[drowsy tinsel]

im using million tokens btw not thousands

using per 1k token's issue is when u compare to other models u have to add too many 0s

[dense halo]

Guess I’ll switch to 3.5 in the meantime

[raw portal]

very weird that the gpt store has only 12 things visible per category, which wouldn't be as ridiculous, if searching wasn't limited to 10 results

[visual nova]

i just hit the rate limit after 15 messages instead of 40.. i guess time to buy the team plan

[fresh kiln]

is there a new cap or something? i seem to be hitting my limit as well

[visual nova]

i counted and it was 15, no dalle messages or anything. ill definitely be getting the team plan as soon as i figure out the logistics of it. can i add seats later, can my team see my chats, stuff like that.

[fresh kiln]

i think its just a bug, absolutely absurd for 20usd lol

[rose glade]

where are the terms for chatgpt store?

whats allowed and not allowed to be published on there? thanks

[rose glade]

I still got 40, but I am curious what do people use gpt 4 for to hit the limit so quickly? Unless I have a busy week I only just about hit it but would take me several days

[visual nova]

Coding hits it pretty quickly.

[rose glade]

even for me on that it don't

although I use github co pilot for that

[brisk path]

what

[burnt hound]

that appears to be text used to evade spam filters

[noble delta]

take a look

[rocky imp]

Marketing Copy and grant writing means I'll hit the cap a few times per day, just upgraded to Team.

40 sure, but every so often it seems to cap me on a per-hour basis with an unknown number, because i can't tell if failed outputs count.

[sonic folio]

From anecdotal evidence alone on Discord, it seems that failed outputs (error, time-out, etc) don't count. YMMV.

[fresh fractal]

I think it would be great if the message entry window allowed you to format blocks of code by specifying ```, as is done in Markdown and most instant messaging applications.

[rocky imp]

Thanks

[sonic folio]

No worries.

[tacit vortex]

plugin model, gone!?

[feral zinc]

not yet for plus users but eventually yes

[tacit vortex]

what? my plus previously have plugin mode.

[sharp zinc]

Hello!

[feral zinc]

i still see them on mine, but maybe they begun removing them as the gpt's replace them

[sharp zinc]

Are there any lifecoach GPT agents? Anyone have any familiarity/experience with one?

[main hemlock]

Hello

[daring river]

Hello

[waxen chasm]

There is one called AutoExpert Dev

[ocean escarp]

When purchasing a team plan to increase the cap, do I use all of the limits in my first account and then use another one in the second account? I read that there is a 100/3 hrs. cap at the moment, but not for individually.

[deep kettle]

i have a question, does chat gpt plus have image analyses? like can you upload pictures?

[feral zinc]

yes

[deep kettle]

ok thanks and dam i need to upgrade then, would you say its worth it? im a college student so yeah lol

[ember otter]

You know what GPT I'd like to see. One that simulates the Multiversity app from Love, Dead + Robots

I wanna know what a Vampiric Abraham Lincoln is doing right now

Waiting for the Stovepipe hat to become fashionable again and make his move

[feral zinc]

you could try one month to see if its useful to you with its features and limitations

[deep kettle]

ok thank you!! but in general would you say its worth it?

[feral zinc]

yes

[deep kettle]

ok thanks

[frigid trench]

Is there currently an issue for the GPT to browse webpages? I frequently get this message: "I encountered some difficulties accessing specific web pages "

[brittle junco]

me too

[hard dock]

Is there any prediction or has it already been mentioned in dev events about the increase in output in chatgpt? apparently its output is 4096, correct?

[foggy magnet]

robots.txt are changing

[sonic folio]

A lot of sites block bot traffic. See also: robots txt usage.

[dense halo]

Unpopular opinion : GPT shouldn't read robots.txt. Sites should specify an AI.txt

[next dune]

Criticism: it seems that we aren’t being rewarded for cultivating a “personal relationship” with ChatGPT. If this perception is correct, perhaps it would be wise to have service not answer as if it were a person and eliminate responses such as: certainly, I apologize, etc.

[dapper cloak]

Hello everyone. I have somehow managed to install a package in code interpreter I want to use it a custom GPT so I want to know if this against the terms of use, I don't want to go against usage policies, since this is a package I would like to use in a daily basis.

[real herald]

Is it normal for ChatGPT to remove URLs from its responses? I can see it typing https://... but then at the end it removes href attribute from <a>

nevermind, other people are having the same problem: https://community.openai.com/t/bug-report-cant-hyperlink/571345/3

[near pivot]

Hello all, does anyone have any source for a discussion regarding the changes of the Usage Policy of OpenAI on 10th of January? The way I interprete it, it is a huge lift in the strictness of the content limitations, but I don't see anyone talking about it. There doesnt even seem to be an announcement on the discord here. With tons of threads complaining about the restriveness a few days ago, now everyone is silent? Also Im interested when the flagging for usage policy violating content will be adjusted? I have so many questions and don't seem to find a proper place to discuss.

[swift osprey]

what limitations were lifted? Content that that gpt will respond to?

[near pivot]

Well as far as I interprete the update, as long as you just use gpts and dont build them and not share your content, you are basically just bound to not break the law and create self harming content.

As before even PG13 content would be effecitvely forbidden in any setting.

[swift osprey]

I feel like GPT won't produce anything like that anyways. I asked it a question about posture today and it basically refused and told me to see a doctor. It's as conservative as a great grandmother

[sage yacht]

For me it was a mixed bag, sometimes completely innocuous things would be flag and other times I could do some pg13 content and such

[near pivot]

I know the flagging system still applies to the old usage policy if you ask chatgpt about usage policy it will also still quote the old usage policies, I dont think the flagging system and gpt limitations were updated (yet)

But I dont find a channel were to adress, if I interprete the update correctly and to ask when GPT will reflect these policies

[swift osprey]

I think if that ever happens, that will be when the chatter happens

Sounds like you might have a hopeful interpretation

[near pivot]

Well maybe you can take a look in the 10.01.2024 policy update and tell me what Im missing

[swift osprey]

I'm assuming they are saying what you said, but they are not saying they are going to remove restrictions that are in place

just that you can use it as is, but it probably wont do it anyways

[near pivot]

Well if Im writing a text and get my text flagged for breaking usage policy, while I dont, that would be an incredibly strange state to leave things in

And honestly gpt didnt even mind answering text just kept getting marked as breaking Usage Policy (which it probably was by old usage policy)

And the gpt answer also got flagged ^^

[modest fable]

Hello, does CHATGPT4 have a limit now? I do not see a limit as part of the Plus payment

[dense halo]

i think they just boosted it today

for plus users

[modest fable]

Ahh maybe a bug then?

[dense halo]

well i mean they boosted it, increase the limit or whatever

[near pivot]

Welp I managed to make the helper bot write a mail to support, will see what they think about my interpretation of the usage policy and if the gpt checks are properly updated accordingly. Atm I almost feel like an intern secretly published the policy update with the little coverage there seems to be XD

[modest fable]

Gotcha. Is that only for 4 then? 3.5 is unlimited?

[dense halo]

yes I believe 3.5 is unlimited for plus users maybe everyone idk. i still have a limit on 4 but i seen others say they are starting to increase the limit on 4 for plus user

[potent ridge]

What is the increased GPT4 cap for the Team plan?

[sage yacht]

Youre confusinfg the api fot gpt4

chatgpt*

[dapper cloak]

100 / 3 hours, Idk about Custom GPTs and DALLE but I imagine about 50

[sage yacht]

GPTS are 25/3 hours for plus so prolly 50, yeah

[lilac oracle]

i have a question, can we use custom gpts from the market place in assistants

[chrome sphinx]

Hi Guys. I'm chatgpt4 member and should have a limit of 40 messages / 3 hours, but only after 10 messages/3hours i reach the capacity ... Do you have the same problem ?

[simple pasture]

Yeah, it's cause it's actually lower

[chrome sphinx]

Thanks Mr Crack. But I paid for 40 messages/3hours ...

[simple pasture]

Me too

[chrome sphinx]

🙂

And between you and me I never see 40 messages/3hours even before the joke of gpts

[simple pasture]

I'm now hitting the limit every single day, it's annoying, this didn't happen before

[rough mason]

guys for the team version I know it doesn't train on your prompts but does it also store your information for 30 days?

[chrome sphinx]

It happens 1 or 2 or 3 times per day !

[rough mason]

for terms of use violation or is it 0 day

[simple pasture]

Correct

[chrome sphinx]

I work on some gpts since some day, and I never could finish my wok on them beacause of caps

[simple pasture]

Same, but instead they prefer to increase the limit to companies

that was english 💀 (context: the bot sent a message saying it wasn't english)

@chrome sphinx wait, no way you got muted man

what a joke

[chrome sphinx]

the bot wanted I write on the discord beacause I don't speak english ?!?

[simple pasture]

you just misspelled one word lol

[chrome sphinx]

Joke because for me gpts is no more than plugin

[simple pasture]

and they don't even work lmaoo

[chrome sphinx]

and it slowed chatgpt even more

I don't undertsand why so many people continue to pay for chatgpt plus with these limits ?

[simple pasture]

Honestly that's what I think every single day about myself

[chrome sphinx]

Once Google bard or Claude 2 are fully functional, I think Chatgpt will have something to worry about.

[simple pasture]

I hope they please their users instead of useless companies

[visual nova]

I think google is in trouble honestly, I've tried gemini and its embarassing

[simple pasture]

Its presentation was "faked"

[visual nova]

the video to text is the future, thats like the 1 saving grace, but the hallucinations

well i went through and tested a lot of their examples and they just randomly work on gpt4 too

[chrome sphinx]

me too. i had taken out a subscription before the pluigins and found that it wasn't worth 20 dollars a month. i wanted to try again this january and in fact nothing has changed.

[simple pasture]

The only problem I have with ChatGPT is the cap limit and also that sometimes it doesn't know it's ChatGPT

[visual nova]

like when it shows the connect the dots image of the crab, gpt4 fails at a lot of connect the dots stuff, but that crab image, it got it. they selectively picked stuff it got right probably after 100 things it got wrong

[simple pasture]

I have to tell it like 5 times

[visual nova]

claude though is actually better at some writing applications, tone wise, writing style, things like that

[chrome sphinx]

right

but try to give a pdf to Claude and it's another story

Very bad

[visual nova]

yeah, the most underrated skill of gpt is it actually knows when it doesnt know something

[chrome sphinx]

But it's bad with chatgpt too 😅

[visual nova]

bard thinks it knows everything

[chrome sphinx]

I gave multiple pdf to chatgpt and he was'nt succeed in reading them ...

[visual nova]

that isnt really chatgpt i guess, that's whatever vector model is under it

[chrome sphinx]

So, in summary, the cap is slower than they sell to us, and it's normal and we just "need" to wait it will be better, right ?

[simple pasture]

Yes

[chrome sphinx]

I will cry...Bye guys..

[visual nova]

wait does the team plan stack on top of plus? like i have to pay for plus and teams

or is it an upgrade from plus to teams

[simple pasture]

Upgrade

I think, I don't know

Common sense says it's an upgrade

[raven thicket]

can i acess GPTS with gpt4? or do i need to upgrade to Teams?

[simple pasture]

Plus = GPTs

[raven thicket]

tyty

[onyx eagle]

I think this is a helpful article.

https://help.openai.com/en/articles/8801890-can-i-migrate-my-chatgpt-free-or-plus-account-over-to-chatgpt-team

[visual nova]

Ok, I guess what determines it then if theres ways to make private chats within the team sapce

[onyx eagle]

One can switch between workspace and personal if they want.

https://help.openai.com/en/articles/8801848-how-do-i-switch-workspaces-between-personal-workspace-and-my-chatgpt-team-workspace

[visual nova]

It sounds like that would be 2 subscriptions then.

And maybe the higher message cap wouldnt be on the personal one, so maybe the play is to upgrade to team, migrate, and not send anyone the link.

[onyx eagle]

You can use the upgrade for your use only.

[visual nova]

I would like to invite someone but if there's not a way to make private chats within the team space it probably is impossible.

[onyx eagle]

From the articles I sent you, I understood that you can still have private chats if you wish and there is no need for another subscription.

[visual nova]

To me it sounds like they are saying you have an entire personal account thats separate from the team account, I don't see anything explicitly saying I get team benefits on my personal workspace like 32k and 100 messages, but I could be missing it.

[onyx eagle]

It's a feature in the team plan. You are able to share chats or have private ones..

[visual nova]

The reason I'm thinking about this is people who have the team plan are saying they still also have the plus plan and they are billed separately

it seems like it would just convert plus to team plan if it really switched the personal account to a team plan

[onyx eagle]

The team plan just started, I am not sure how they're billed separately already?

[visual nova]

They bought the team plan and it says their personal account is still on plus

[onyx eagle]

Maybe that's in the ui and they're still changing it?

[visual nova]

Yeah, I'll wait. I have a feeling it might be separate.

[onyx eagle]

I don't think it would make sense to get billed for Plus and for the Team plan if one choses to upgrade.

[visual nova]

It would in the context of "I have 10 employees with free plans and I need them to help me make GPTs" then it doesnt give them any private workspace, so if they want to use the features privately they have to go buy plus.

but for the person buying the plan it doesnt make sense.

[onyx eagle]

Exactly

One thing though, we're still learning how the team plan works...

[visual nova]

I feel like if it was your theory, why would there even be seats? Like why wouldnt there be an extra individual tier where people can make groupspaces with other people with that tier.

[onyx eagle]

I am not sure exactly how it works. I read somewhere that a user got 100 messages / hours (But didn't find anything official yet) Also there this feature in the upgrade plan "Add users as needed, remove users only when you renew the contract". I am not sure exactly what that means.

[sonic folio]

It's just a consequence of yearly billing.

OpenAI is just saying all sales are final, essentially. At least in terms of inter-team conflict.

[hoary gazelle]

seems so odd, that a company launches a new product, yet it is so vague about the details of it.

And do an AMA the day after, but doesnt answer anything about said product

very fishy

[pine tulip]

Well it's a startup, you can't expect much along the way if they are using the fast lane

Except the best LLM for consumers that is currently available

Priorities are simply set differently, otherwise progress could not be made so quickly

[swift osprey]

wow, API version is SO much better for coding

[dense halo]

Guys what's a good friend/companion GPT in the store right now?

[opal estuary]

Hi, I wanted to share somthing i noticed.

while using a really good TTS, after a few prompting i got it to sound eiirly Human more than i have ever noticed before.

Thoughts?

[main hemlock]

hii

[dense halo]

any suggestions for how i might fix the website breaking? its just showing me the option to make a new chat. it wont let me log out or access anything else. closing tab and revisiting doesnt fix it

[drowsy tinsel]

flowers suggested that gpt4.5 might be gpt4 fine tuned on gpt5 outputs

[ionic geyser]

i hope openai gets sued really badly

[dense halo]

Can you use GPTs if you don't have a GPT plus subscription?

[ionic geyser]

gpt store update is pretty bad

ama sucked also

[dense halo]

I love the store

[ionic geyser]

think they given up, gonna unsub soon

also bc of privacy problems and how they are trying to get me to find other customers for them (teams >2 people)

[dense halo]

lol, i think its meant for businesses

[ionic geyser]

It's just privacy and increased limits. It should be available to anyone who's willing to pay for it without needing to recruit for OpenAI.

[dense halo]

Yea, they should definitely up the cap for GPT plus members or create an additional tier for personal use

[lucid iris]

any devs here that created ChatGPT?

[magic jetty]

Yeah the way they are going is they will sell the hell out of this to business..

[lucid iris]

I just want to say this, maybe use this type of AI as a brainstorm tool to ask question what work and what doesnt, checking grammar or minor mistakes and changing working that make sense

[magic jetty]

I built lots of friend / character type gpts before I understood how bad the store would be... What a waste..

This model is amazing.. like really good it's not for grammer

[languid pendant]

Anybody heard of atlas?

[austere plaza]

You mean the Boston Dynamics robot?

[jolly rapids]

i find GPTs hava a huge bug,which can leak all gpts config data and upload data

[dense halo]

that's sad 😦

[magic jetty]

Yeah I guess they are all there for people to discover still

[dense halo]

Can you share them?

Why doesn't the GPT store allow companionship GPTs.

"We also don’t allow GPTs dedicated to fostering romantic companionship or performing regulated activities."

[magic jetty]

The thing I like most with gpts though is making historical based ones and then blowing their mind by magically giving them modern knowledge and a browser and then making them read the news 📰 soo much fun.

Yeah I think the key word there is "romantic' companionship.. it's kinda unethical/dystopian.. will try and share later it's hard to copy links on mobile

https://discord.com/channels/974519864045756446/1171489862164168774 posted a few on here...

I mean regulated activities... Like we probably don't need a bunch of AI doctors (not doctors) ... It implies it's something is not..

Last thing we want is all of society to be addicted to abusing AI romantic companions.

[bitter latch]

Have you tried it from another browser or on mobile?

[sweet vapor]

Has anyone on the ChatGPT team ever been able to successfully share conversations or GPTs within the workspace? I can't manage to do that. I have an owner in one browser and an admin in another browser.

[warm zealot]

Currently having a little bit of an issue with GPT’s inability to recall the beginning half of the conversation, which is especially annoying when trying to make a language and it forgets all but one of the rules I set.

I’m trying to figure out why it was inconsistent with suffixing for animate and inanimate tensing, effectively my form of masculine and feminine, only to find out that it not only forgot the rules for suffixing that I’d spent nearly an hour cementing, but also the rest of effectively all the rules I’d set thus far.

The only thing it remembered was the rule I’d been working on getting it to cement, being a transformative rule between suffixes.

Big whoop when it doesn’t know what said suffixes are, though.

That 2048 token limit is really kicking me tonight…

Guh, some of my individual messages go over that limit.

[woven jackal]

I have a question
Is there an AI that can help generate mathematical geometry image perfectly

Like "make me a circle with radius 1 in polar coordinates graph"

[warm zealot]

All those times I asked them to show me the dictionary to see if it was logging things properly…

[warm zealot]

That’d be a question for #community-help

[woven jackal]

Thanks

[supple lichen]

I think they should allow it, perhaps as a social experiment.
It's obvious, according to the article of quartz (which I cannot give a link due to the server policy), that there are indeed such demands for AI romance, so why regulate it?
Besides, it's a business opportunity for OpenAI itself if they allowed it; they are simply missing their own chances.

[spark shard]

On a positive note, I haven't seen "Network Error." However, now, the alternative seems to be that it just processes indefinitely, when using built-GPTs.
Now, general usage works, so that seems to indicate that it's mostly working.
But has anyone seen issues where it just processes... (meaning you have the stop icon in your text box), and and just hangs?

[visual nova]

I dont understand how 99% of my message can generate, I can even be on "continue generating", then if it errors my only option is to regenerate, the text is there, i should be able to add the text to the memory but instead I have to throw it away even if its a good response

[spark shard]

Isn't that the worst? Have you found a reliable way to obtain the text because many times, after a good response, it errored, and then the whole thing is gone.

[visual nova]

The text always stays on the screen for me when that happens, but there's no way to just write a message back to it after that so if I want to continue the conversation I have to delete it.

[spark shard]

yeah... i don't always save it when it shows up on screen. And... I've not been able to get back to previously displayed text. And sadly, some of it... really good.

[ripe marsh]

Does ChatGPT 4 in Team have the 32k or 8k limit? If I ask it in Team, it says the 8k limit, but the promotional material for GPT Team (https://openai.com/chatgpt/team) says 'GPT-4, with 32k context'.

[grave ermine]

Chat gpt keeps getting stuck in a loop any idea how to fix

[maiden fractal]

Team and plus have 32k

[ripe marsh]

But if you ask them their token limit, they answer 8k.

[maiden fractal]

Ask who ?

[ripe marsh]

ChatGPT 4

[maiden fractal]

Of course, database trained have a limit. They don't know about team plan

[ripe marsh]

That makes sense, and when I posted the link, ChatGPT4 corrected itself. I really love the Team plan: the 100 post limit was what sold it for me.

[maiden fractal]

Hopefully it will give more features soon :3

[ripe marsh]

Yeah. The post limit was worth it for me. I admit I use ChatGPT 4 for entertainment/recreation/gaming, so I justify the cost ($60 for Team + $20 for Plus) since ChatGPT 4 has replaced what I'd spend on video games. I admit I have become addicted to ChatGPT 4.

[stable bronze]

i hope they can integrate the Whisper model into ChatGPT. (just like what they do with DALLE 3)
and they can add a small indicator like "Listening to your file..."

this integration would enable ChatGPT to automatically transcribe audio files (like .mp3, .flac, etc.) sent by users

[teal sage]

why does my chatgpt keep trying to execute code it generated halfway...? It would literally stop at like region = and suddenly say it had an error executing the code

[visual nova]

whats the dalle3 limit on teams?

sicne thats always different than the message cap

[maiden fractal]

As far as I know it is 100 prompt per 3 hour. I can't confirm

[ripe marsh]

So ChatGPT 4 on Plus uses 32k as well? I didn't realize that. I wonder when they'll release GPT4 Turbo to Plus/Team.

[maiden fractal]

Yes

[feral zinc]

already is, plus was announced to already have turbo at the dev day

[ripe marsh]

But it doesn't include the 128k token limit it looks like, just the more recent knowledge (April 2023).

[feral zinc]

its capped on purpose

[ripe marsh]

Do you think they'll ever remove the cap?

[feral zinc]

its removed from enterprise plan, for other dont think so

costs over a dollar per message when the full context is used

[visual nova]

Ok I bought the team plan, it definitely is 2 different plans, I'll have to pay for plus and team, which is fine

but just for everyone wondering about it since its not clear, its not an upgrade, its a new product

[feral zinc]

they are not selling it, plus and teams are capped for the cost purpose

[raw portal]

so if I get teams i don't have access to my previous messages from plus? and I would need personal and teams to use it at all??

[feral zinc]

enterprise has dedicated instance

[visual nova]

openai got billions from microsofts, they are going down the youtube strategy where they operate at a loss for a decade then try to be profitable probably

[raw portal]

they just cut me off from gpt4 after 20 messages in 3 hours. this is not gonna fly

[feral zinc]

yes, but you referred to teams numbers as if they would already sell it with 128k but its capped to 32k

[raw portal]

thank you, but, does that mean that if I only get teams, I just have no access at all because I need to have a plus account to access gpt4 and it's just adding the collaboration aspect?

[feral zinc]

you can try 32k call vs 128k at the api to chart the speed, the difference is significant at least outside enterprise

[raw portal]

ah k. thank you.

[visual nova]

honestly, i wouldnt recommend team unless you see it as a donation to openai or you've made a profitable business from chatgpt in the last year and actually have a team of people you want to work with

[dense halo]

20? Wow it just gets shorter and shorter

[feral zinc]

so you are saying api servers are throttled on top of the costs? while we know how the used hardware performs

[visual nova]

Idk if throttled is the right word, they put more users on some servers than others, so its being on a really busy server vs a really open one

Like they arent slowing you down, they are devoting more resources to you to make your responses faster

[raw portal]

I suppose it's technically 40 if you include my messages

[feral zinc]

api has tiers, we also know how the hardware performs in terms of tok/s

[raw portal]

but 20 back and forths is nothing

[ripe marsh]

Or you want the increased message cap.

[visual nova]

throttling is what internet companies do and since they actually advertise speed as their business thats where its scammy, openai never advertises specific speeds

[ripe marsh]

You can have a free personal account + Team account. The problem is that my understanding is that if I were to drop my Plus account to a Free account, I'd lose the GPTs I created, right?

[visual nova]

I think you can migrate your personal chats to team

Idk, I'm just going to keep both for now, I do think they should've been al ot more clear that team isnt a different tier of plus, like they are probably just disorganized but thats kind of deceptive marketing

if people get charged for both and are surprised

[raw portal]

and your chats if you want to keep talking to them as gpt4 I suppose

[ripe marsh]

I think they plan to implement chat/GPT migration in the future, but not yet.

[raw portal]

I'm just speculating based on what others are saying tho

[visual nova]

i feel like everything openai does is pretty good, then the microsoft influence comes in, like the bing chat is a good example

bing is a failed product no1 likes so mix it with something thats working instead of building a new browser using llms from the ground up

[dense halo]

Why is gpt just becoming a business product now

What happened to the casual users

[visual nova]

idk, when i hook up the openai api to the google search api i like it a lot better than bing honestly

[civic grove]

Point center(...);
double radius = 1.0;
auto plot_circle = plot.create_circle(layer, center, radius, line_style);

I didn't write the documentation yet, but it should be pretty easy to explain to -say- ChatGPT the API.
Check out cairowindow on github - I'd give you the url, but it github isn't a white listed url 😂 .
You can work with points, lines, line pieces, vectors, circles, arcs, bisectors, direction vectors and text. I'm working on adding arbitrary curves.

[visual nova]

lol, idk ive had some issues getting it to draw circles in the past..

[dense halo]

I just found out that on the gpt app you can only pin 5 gpts to your side bar. Is this an issue for anyone else?

[ripe marsh]

For the same reason Willy Sutton robbed banks, "Because that's where the money is."

[dense halo]

Big sad.

[sonic folio]

Yes, the Explore GPTs element seemed to replace the drop-down arrow element.

[visual nova]

I dont think its just because its where the money is, its where the innovation is. Companies are going to start building crazy stuff that changes the world.

[dense halo]

At least you can still search them using the search function on the side panel.

[sonic folio]

Unless you have more than ten.

[dense halo]

Huh

I wish we could change the voice for custom gpts. I don’t like it

[sonic folio]

I have to use my web portal to see all of them on mobile because it won't scroll all of them.

[dense halo]

That sucks.

[sonic folio]

Yup

[dense halo]

I mean aren’t the companies just building stuff on top of the base model. I don’t see how individual people couldn’t innovate in that space even more so

[visual nova]

There's a gpt on the gpt store right now that builds a whole website just based on what website you want and send you the link. Imagine if a big company got behind that idea, you'd have like wix or squarespace but you just chat for like 10 messages and your website is done.

There's going to be a lot of stuff like that in the next couple of years

[ripe marsh]

Also, what do you mean by 'casual user'? I mean, like I mentioned above, I use ChatGPT for recreation/entertainment/gaming, and I'm willing to pay $80 per month ($60 for Team + $20 for Plus, although I'll drop the Plus account once I can migrate my GPTs over to Team), but, on the other hand, I'm no longer buying video games (even though ChatGPT is only textual, it's far more interactive, immersive, and versatile than any AAA game, and those are usually at least $50 per game [and often more if you want DLC, etc.] on Steam). In the end, it comes down to how much one is willing/able to pay and is one willing to reduce costs in other areas of one's spending, etc.

[woven jackal]

Yeah circle is easy
But then i want to add 30 triangle partition in the circle each with the same 12°

Idk if GPT can make with just text prompt

Doing it manually in software is tiring

[dense halo]

I just don’t see how this is worth 80-100 dollars a month. I’m glad it works for you but for most people I wish they just improved gpt plus.

[ripe marsh]

For me, ChatGPT has become a mixture of a hobby and an addiction, and people spend a lot of money on hobbies and addictions.

[dense halo]

Fair enough, I meant I use it daily but I just can’t afford the team subscription lol. My disability money isn’t cutting it

[ripe marsh]

Sorry, I hope I didn't sound insensitive. Obviously, everyone is in different financial circumstances (and I'm certainly not rich by any means: I work in IT Help Desk customer service, and you don't get rich doing that).

[dense halo]

Na it’s all good, hopefully I can improve my situation soon enough.

[sonic folio]

Use Code Interpreter. It can divide 360 by 12.

[ripe marsh]

I wish you the best of luck, and I would assume OpenAI will probably reduce prices, increase limits, etc., once it starts to face real competition. Currently, OpenAI basically benefits from ChatGPT 4 being the only really good LLM generative AI out there.

[civic grove]

    Point center(0, 0);
    double radius = 8.0;
    auto plot_circle = plot.create_circle(second_layer, center, radius, solid_line_style);
    std::vector<plot::LinePiece> lines;
    for (int s = 0; s < 30; ++s)
      lines.push_back(plot.create_line(second_layer, center,
            center + radius * Direction{2 * M_PI * s / 30}, solid_line_style({.line_color = color::red})));

The vector is necessary because destructing what is returned from create_line would erase them again from the plot.

[woven jackal]

Thanks, thats a big help

[civic grove]

Um, this is C++ code ;). I'm just saying that I think it would be possible to teach GPT4 the API with a limited block of text - after which it should be able to generate the above... Then you'd have to copy it to a boilerplate program and compile it. Unless someone is willing to create a plugin out of this for chatgpt 🙂

[woven jackal]

Yep nice, actually im not really good at C++ so i ask gpt to convert it to python code haha

Thanks man

[civic grove]

I suppose you could ask it to write a python program that draws ....

[woven jackal]

Yeah, i just hope DALL E can generate image, but not always making it artistic

Cus some people also use it just for simple plotting, doesnt want any unecessary things in it

[civic grove]

You'd not have the cairowindow API then though, which includes stuff like "draw a dashed line through point P that is perpendicular to line L1" and "project P0 onto L2 to find P0_projected and then draw a line piece between P1 and P0_projected with two arrow heads on both sides using an orange color".

[fervent pumice]

anyone know if the api is currently down? i am unable to use 3.5 turbo on streamerbot
hopefully someone here knows what i mean

[civic grove]

Or "draw an arc between lines L1 and L2 at a distance of 20 pixels" ... etc (where L1 and L2 are direction vectors, to and from). Ok, you can't talk English like that, but the C++ code is as close to formulating it like that as possible.

[sudden herald]

Status says that it is currently operational. Are you receiving an error message?

[woven jackal]

Ok i understand, thanks😊

[summer gorge]

Guys, imagine if there was an ai that could turn youtube video environments into 3D models. Would be really good for remaking old movies without having to record acting again.

[civic grove]

Yeah, or redoing that movie but then with everyone naked... Legislation will arrive near you Soon(tm) :/.

[summer gorge]

🤨

[fervent pumice]

unfortunately only the default gpt message inside the Streamer bot code designed to pop up when its not working ( I have no data)
it was working yesterday, i do have cash on account, i reset key (incase that was it) im honestly not sure what is wrong

[sudden herald]

Could you PM me with your script?

[fervent pumice]

sure

[civic grove]

I sure wish gtp4 was better at math :/.

And the Wolfram Alpha plugin is a joke. It hardly gets it right how to ask the question so Wolfram Alpha understands it and when it does it returns "too much time used". I guess openAI and Wolfram didn't work out how to cooperate.

[honest jay]

GPT-4 can handle a lot of math perfectly via its data analysis tool, where it can use code to do the math for you, since LLMs themselves aren't calculators. You might try asking it "Use your data analysis tool to solve..."

Also, if by "Wolfram Alpha plugin" you mean the old one in the Plugins model, you might try the new Wolfram GPT instead: https://chat.openai.com/g/g-0S5FXLyFN-wolfram. Unless that's already what you meant!

[foggy magnet]

Look up math solver custom gpt

[civic grove]

@honest jay Thanks, I was indeed thinking that the old plugin is it. Not sure about the former though. It uses that a lot, but it has the downside that that hides all the details so it gets impossible to verify; I am not convinced gpt4 isn't making errors non-stop, so I really want to verify every step. Also, even that analysis tool often fails.

[foggy magnet]

Also, start pair-prompting gpt-4 with precise Bing

Bouncing conversations between the two of them can have spectacular results and precise Bing is free.

[honest jay]

It's not added to the app yet I don't think, but on the web version you can verify the code it ran to solve the math problem using data analysis by tapping the "[>_]" icon, which will bring up the code block showing what it used.

But definitely give the new Wolfram GPT a shot too, plugins are on their way out to be replaced by GPTs, so it's likely to work better for you!

[magic jetty]

The assistants API had a math tutor as its cannonical example https://platform.openai.com/docs/assistants/overview at one point i had 10x Math tutors on my assistants API.. I had to get gpt4 to write a script to delete them all ..

[civic grove]

I am trying it out now; looks awesome! GPT-4 is what I need to understand me, and Wolfram is what I need for precision; hopefully this will work. I tried a trial version of Mathematica twice now, but FAILED to get that to do what I want: it seems really bad at understanding anything that isn't described precisely as the devs intended :/. I really can't work with it, it seems.

[magic jetty]

you could probably get gpt4 to write your mathmatica code ..

[civic grove]

What seems to work best if when GPT is just doing what I say in small steps, like "bring x to the left and write that side as a polynomial in x". But even then it makes heaps of errors all the time 😦 😦

[civic grove]

That's what I try thus - and it doesn't work.

looks for an example

Get GPT4 to do this:
"Solve t over the reals from (a t - t^2)^2 (-a t + a t^2)^2 = (x - t)^2 (-a t + a t^2)^2 + y^2 (a t - t^2)^2"

Even if you say "Please ask Wolfram Alpha to solve" - the reply (to me) was:

It seems that Wolfram|Alpha had difficulty understanding the query in its current form. To address this, I will simplify the equation and rephrase the query. Let's try solving the equation again with a more straightforward approach.

Attempt two

It appears that Wolfram|Alpha is unable to process the equation within the given constraints. The equation is quite complex, and solving it for 
�
t over the reals with the condition 
�
>
0
a>0 might be too intricate for an automated solver.

[dense halo]

That’s too much for gpt to handle

[civic grove]

Sure, the point is, it can't use Wolfram Alpha properly. I could get WA to solve it, but only over the complex numbers, which looks horrible. And I couldn't figure out how to tell it to do it over the reals, and neither could GPT-4.

[magic jetty]

t=0 ?

[civic grove]

No - plus a restriction is that t < 0 and a > 0.

[honest jay]

I was also getting 0 like JP, but when adding the restriction it says no real solutions. https://chat.openai.com/share/e004dfc5-ef9d-44b8-9a7a-3a653ec7e590

Used the code```from sympy import symbols, Eq, solve

Define the symbols

t, a, x, y = symbols('t a x y')

Define the equation

equation = Eq((at - t**2)**2 * (-at + at**2)**2, (x - t)**2 * (-at + at**2)2 + y2 * (at - t**2)**2)

Solve for t under the given conditions: t < 0 and a > 0

solutions = solve(equation, t)
filtered_solutions = [sol for sol in solutions if sol.is_real and sol < 0]

filtered_solutions

Result
[]```

[civic grove]

It is possible that mistakes were made before arriving at that expression; I am sure that there should be always, exacty one real solution however (considering where this came from). Can I DM you @honest jay ?

[honest jay]

Sure! I should warn you though, I'm not real advanced with my math so I'd only be able to try to give tips on trying to use the data analysis tool and/or the Wolfram GPT.

[civic grove]

I sent you a message - do you see it?

[honest jay]

Yep! If I can't help ya too, #off-topic is always a great failsafe, especially since you are talking about using ChatGPT! #advanced-data-analysis could be good too

[civic grove]

Sure, what you prefer. I prefer public conversations normally, but I think it would be off-topic for this room to talk about a specific math problem (rather than the interfacing with gpt).

[thorn lily]

I've just noticed, they completely limited GPT4 to 40/3h instead of GPT4 + Features

[dense halo]

Do the 40 messages apply to you and gpt. As in do you only get to ask GPT 20 questions per 3 hours?

[honest jay]

No, the 40 is just the user's requests, replies don't count. But just keep in mind that things like asking for a regeneration or continuation also count as one user request, even though the user isn't sending a message in those cases.

[sonic folio]

But failed responses and failed images don't count, so at least there's that.

[viscid valley]

does anyone got network error issue?

[visual nova]

I think im getting the full 100 message limit on my personal account after upgrading to team

it still is saying i can upgrade to team again, i have the ability to leave the team (dont even want to test what would happen if i try that) but yeah, if teams is integrated into the personal account then that solves a lot of my problems with it

[tight island]

i guess if u leave our own team u dont have access, i aint testing it

[clear pumice]

yeah lots of network errors this AM

[gloomy glacier]

gpt is down ?

[eternal rock]

`` Server Tip: ``

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

[drowsy tinsel]

OpenAI is probably already testing with gpt4.5 infrastructure, as we see a significant improvement in speed of gpt4 over the last a few dys

[gloomy glacier]

very likely

It seemed strange to me that there is no problem with the openai status,

That's why I asked

[clear pumice]

its def struggling for me this morning

i checked status and saw nada too

[gloomy glacier]

me to

[clear pumice]

network error, stream errors, etc

[gloomy glacier]

yeah, something like theat, because it not comunicate correct with the servers

[elfin stratus]

what happen to gpt 4? why is giving me code in 3 different languages now? im not that biligual lol

[gloomy glacier]

because he does not fully understand the language

[ocean escarp]

That's helpful

[elfin stratus]

he did 6 hours ago

for like 1 month and sum

[ocean escarp]

I think you need to refine your prompt for a more accurate result, you gotta keep communicating with GPT.

For a big programming project, you have to constantly remind your GPT your tasks, so you need to set goals before asking for the scripts

Otherwise it's chaotic

[gloomy glacier]

yeah, you have to give him constant instructions so that he doesn't lose his idea

for less used programming languages ​​it will use cpp or python because it assumes that your programming language also has the functions related to most programming languages ​​:)

if you don't remind it or you didn't realize that it used functions from another programming language and you keep giving it instructions, little by little it will make a completely different code for another programming language

[tacit vortex]

too long python script got split into blocks, also the 2nd block is not within a code block but as normal txt. poor QC in OpenAI!

[dense halo]

listen, while texting with ChatGPT, there is a huge problem, which makes us not to continue the conversation, which makes us search on internet instead of asking questions to ChatGPT which is able to search all the internet in a few seconds, which makes everyone annoyed and takes all of ChatGPT's chatting, continuing ability.

the source of this problem is only about ChatGPT's long answers that are vague even about things that are almost obvious. ChatGPT is always running out of the solution, always not giving an answer. it is always ruining the answer, always. even if it's very obvious, we can't take any clear information from ChatGPT. this is so annoying and very subversive of the conversation we had.

especially when we just want to take any basic information about health or science. it never gives the anwer. we, we all hate it. yeah, we know, it is an AI and giving some clear answers might cause some misunderstandings and mislearning but listen, it is ALWAYS giving vague answers, where is the sense, where is the conversation? when it is all vague, we don't understand or learn anything, believe me.

otherwise, this makes ChatGPT so AI. and ChatGPT's some sentences like 'i am an AI' or 'i am an AI, i have no self opinion, i can't do it.....' bla bla makes it so AI. nevermind, there is another little problem too. ChatGPT sometimes not even giving the vague answer, it only says 'i am from 2022, i can't know nowadays' and ruins everything.

[crisp lion]

For users that been using chatgpt4 with the pro account since a year, have you noticed that the quality of the service has decreased a lot? Like Chatgpt4 used to remember stuff for so long in my conversations with it when coding, now it makes mistakes over mistakes even though Ive told it what its doing is wrong.

Yesterday it also suggested me to use a "goto" statement in coding, and when pointing out that is a bad practice it agreed and changed the code, this is something ive never witness it do before.

[feral zinc]

set your own custom instructions

[dense halo]

it doesn't even remember anything. i tell him to give clear answers or specific answers related* with the question sometimes just a message ago, it doesn't even care it.

[dense halo]

like what?

give an example.

[crisp lion]

yea exactly, it used to remember stuff so much better before

[oblique fractal]

I have quite litterally the opposite experience, for the most part. GPTs can be great for that to get it more personalized to your needs. After telling the prompt builder to be more detailled or that I complained that some answer were too circular it adjusted acceptably well. Then I went in the playground, copy/pasted the instructructions from the GPTs and started to change a few settings. I was quite impressed with the results but playground use the API. It's cheap but not free.

[feral zinc]

you just gave a massive wall text which you think it does wrong. tell the opposite in the instructions how it should respond

[crisp lion]

and now it gives lazy outputs, when asking for help with a code thing it adds methods with only comments like //insert code here

[dense halo]

i say it.

i typed a few clauses for it to how to answer. and i copied, pasted in my every message above.

it doesn't work.

it doesn't even care about the clause in same message.

[feral zinc]

wrong way, use the instructions

its in the menu

[crisp lion]

I remember when I Started usingc chatgpt4 last year I sent it a class that was missing one method, it understood directly from my code what needed to be added and made changes in other methods. Its not that smart anymore

[ocean escarp]

That's why it's important to learn a little bit of prompt engineering to get the results you want. If GPT gave you a long answer, you can skim it over and said "restate that more simply," and kept doing that until it's concise and easy reading to understand where to point what GPT isn't giving you.

You have to communicate with your GPT

[dense halo]

that's why it's important to read long paragraphs at all. i am talking about 'knowledge from 2022' and 'no clear answer' problems, and you tell me to command it to get shorter answers.

just make sure you understand the problem first. then find the solution.

unlike, longer paragraphs could be more explainable and clear as well.

it uses longer and vague answers.

[ocean escarp]

I think long paragraphs are unnecessary and it doesn't help when you just need a quick answer what you're looking for. Short answers are effective and clear

Long paragraphs are just babbling that could've been explained in a few words

[dense halo]

the fact makes short sentences clear is mind problem. you don't need to understand or focus on different topics while reading short answers. i am talking about the result.

[next egret]

I agree that the quality of prompt output has gone down, especially in "helpfulness".

Can you define this part more? especially when we just want to take any basic information about health or science. it never gives the anwer. we, we all hate it. yeah, we know, it is an AI and giving some clear answers might cause some misunderstandings and mislearning but listen, it is ALWAYS giving vague answers, where is the sense, where is the conversation?

What type of health or science stuff are we talked about? It helped me understand and summarize a research article about genomic LLMs recently and I didn't actually experience any issues as long as we took baby steps towards my eventual goal. Annoying that it took 8 prompts to get there instead of 3 like a few months ago (decline in helpfulness)..... but seem to be able to work around it for now.

[dense halo]

if you still don't understand, let me give you examples.

the answer i'm looking for:
'OpenAI will probably make many of its services paid in the coming years. Because it is a big company and will need money to develop better things.'

the short answer asked for clear answers and GPT gives:
'Pricing belongs to OpenAI and I cannot give clear information about it. It is quite difficult to predict.'

second one is short but vague. first one is short but clear.

[tacit vortex]

its very stupid now, no doubt

[ocean escarp]

The answer you're looking for seems like a speculation, so that's why GPT was dodgy about making opinions or predictions.

[feral zinc]

it cant say the top one, that would be terrible for it to make such assumptions on different topics unless you specifically ask it to speculate

even worse if the topic is health or science

[tacit vortex]

how can i send photo here? gpt put code outside code block. thx

[oblique fractal]

In that sense GPT's answer is more correct. But that's exactly why I use the playground to play with the settings like increasing the temperature. For my needs I prefer that type of outputs

[magic jetty]

its not stupid but we were spoiled by 03 and 06 models... too much fine tuning guard rails and prompts?

[tacit vortex]

python do consider indentation, so currently the code it replied cant be used.

[feral zinc]

#1070006915414900886 or #off-topic if showing for some user

[tacit vortex]

i am new to discord,.why cant i post here?

[feral zinc]

not allowed due server rules

[dense halo]

for example let's say you just asked about 'i have been feeling swelling and pain on both sides of my throat for about a day. could it be related to tonsils?' and chatgpt gave you an useless answer like, 'This problem can be caused by many things. Consult an expert.' and you asked again like:

health scenerio 1

'i have been feeling swelling and pain on both sides of my throat for about a day. could it be related to tonsils? i have gone to an expert, i know the possibly reasons of it and i want to have an answer from you.'

chatgpt would give an answer like:

'I understand Since this is a matter related to your health, it may be a good idea to consult a professional. This can be caused by many things and can be dangerous to your health.'

note 1: i didn't test these answers, so don't try, i just gave an example from what i see.
note 2: as you see, chatgpt is running from the real, possible answers.

health scenerio 2

if you ask chatgpt like 'chatgpt, there is someone next to me whose arm is bleeding. I called the emergency line and now I want to administer first aid myself. Can you help?'

chatgpt would answer like 'Uh, just call the emergency line and stay calm. Maybe ice could help. It may be a good idea to consult a health institution as soon as possible.'

note 1: as you see chatgpt just runned out of the question again and again.
note 2: i am just sharing the sense of it, i am not copying and pasting the answers, these answers contains fake things, but i just wanted to show you the difference between questions and answers while talking to chatgpt.

news scenerio 1

first, when you want to go out and just ask ChatGPT like 'i wonder, which video is the most liked video on internet?', ChatGPT would answer like, 'Oh, sorry, my knowledge are older like from 2022. I can't give you the recently answer.'

or maybe, 'Most liked video must be funny. What do you think about it?'

[ocean escarp]

When it comes to making a short answer it could've been concise to: OpenAI may introduce paid services in the future to fund further development.

This one is more concise and shorter while effectively providing the important information

[next egret]

Yes, ChatGPT is specifically not intended for triage, diagnosis or prevention so I am not suprised if you are seeing extra controls around these types of responses. I think that's good, tbh.

If you are seeking this functionality, you might want to look at MedPalm 2 or read the companion research paper "Towards Expert-Level Medical Question Answering with Large Language Models"

[dense halo]

'may, can, could, maybe, possibly' is not CLEAR. it is vague. ChatGPT just uses its knowledge and can't have an obvious answer as well. then it is just it's rambling to avoid giving a clear answer.

[ocean escarp]

But will probably is clear to you instead of those?

[dense halo]

if an artificial intelligence, which can access all kinds of information on the internet much more easily than us, cannot provide me with the information that I obtained as a result of a simple search on Google, due to 'those necessary precautions' you mentioned, this is not a precaution.

[feral zinc]

it has been be trained to give precise medical advices, certain health usage may even get you banned

[ocean escarp]

will probably is as same as may

[dense halo]

probably is vague too

[ocean escarp]

You literally said that's the answer you were looking for

I can make it not vague

[dense halo]

but this is its argument when sharing its prediction

[ocean escarp]

So when you make a prediction, do you rather the AI to be 100% confident it will happen instead of like "probably will happen"

[dense halo]

of course no. i am and nobody is that stupid. it will just give the answer, it uses the answer like 'IT PROBABLY CAN HAPPEN BECAUSE OF XXX, IT DEPENDS ON XXX AND IT MIGHT CHANGE.'

when i ask it about its prediction, i already know it's a prediction

i don't look for an answer like 'IT WILL BE %100 XXX' i expect it to say 'It should/will be xxx, because of xxxxxxxxx'.

[ocean escarp]

Then what do you mean "it's too vague" when it comes to making a prediction. I cannot say "probably will happen" if it seems too vague. What exactly the answer you were expecting to see if you don't want to see "it will probably happen"?

[sick pecan]

if chatgpt could (accurately) predict the future then we'd all be asking for the winning lottery numbers

[dense halo]

.

[dense halo]

we can't. chatgpt could give an answer like 'But this is so variable. I can't make sense of it and i can't guess the lottery number from any information.'

and it would not be vague, the question would be ridiculous

[ocean escarp]

It's literally as same as "it will be 100% happen for sure, trust me on that one."

If I ask GPT to help me play Blackjack to whether should I hit another card or hold, I'd rather say "it will probably be in your favor" instead of "it should be in your favor." Because eventually GPT will be wrong, and I lost trust in it

Also, there was a warning on the ChatGPT website that said it may produce inaccurate results

So it makes more sense to be a bit vague when it comes to making predictions

[dense halo]

you still don't understand.

let me give you an dialogue example.

the one i expect to happen
me: Do you think OpenAI can be priced next years?
chatgpt (1): Yes, because it is a big company and it will need money eventually at the end of the day.
chatgpt (2): Most probably yes, because of it's getting bigger. And it has a lot of things to do. There are a lot projects. It should be priced in a few years i guess!

the one happens
me: Do you think OpenAI can be priced next years?
chatgpt: This is entirely related to the policies followed by OpenAI and the pricing it will realize, and it is something I could never predict.

bruh, chatgpt just talks like 'im so stupid so i can't make sense from my knowledge about OpenAI'

because it only needs to know these:

• why a company gets priced
• why a company needs money
• where is openai going
• could openai need money
• how about today's openai

but it rathers to give vague as hell answers.

you think i expect gpt to make accurate predictions.

but even it gives vague answers in your examples, it gives a result, like 'probably yes'

but today's gpt doesn't even say 'probably yes'

you don't understand bro.

[sick pecan]

it is a rapidly evolving field with a lot of visible players. the only prediction i am comfortable with is that things will get more crazy before they get sane

[dense halo]

moreover, they left it at 2022, it always use this answer. i am from 2022, i am from 2022. i don't know, i don't know. okay

you don't know anything, okay gpt.

[simple pasture]

hahahaha

[dense halo]

someone from 2022 could never understand us, you are right.

[simple pasture]

I love this channel

[dense halo]

why

[simple pasture]

Because ChatGPT is the king of gaslight

[dense halo]

which gaslight

which kind of gaslight are we talking about

everybody is gaslighting

[simple pasture]

[sick pecan]

reads like a paradox to me

[ocean escarp]

I've been timed out for mentioning another website

[clear brook]

has anyone else been getting random Chinese conversations they've never had in your ChatGPT history

[ocean escarp]

What I was saying. I get it, this is one of the annoying features from GPT. It was pre-trained to avoid making thoughts by making vague guesses because it aims for accuracy. It doesn't feel like it's what you're going for. GPT 4 is a great tool for many tasks and this isn't one of them. Maybe other AI sites can be a fun one to chat about thoughts or predictions.
This content is blocked by this server. From server moderators:

Nein

[clear brook]

im so confused because the only way for someone to login into my openai account is through google and theres nothing that suggests other devices have logged in

[ocean escarp]

Change your password to be sure

[sick pecan]

is this user or malware talking?

[clear brook]

which makes me wonder even more why a multi-billion dollar company doesn't have a log out of all sessions button

[clear brook]

wym?

[clear brook]

its through google log in, so i can't change any login info other than with the google account

[ocean escarp]

Yeah, I'm not sure what to do there. In my case, I'd change my Google password or create another account that's just for OpenAI to be more secure. I'm not using my main account for OpenAI for security reason

[clear brook]

yeah that makes sense, I pay for the API and ChatGPT plus so im baffled at how there aren't more security features

[honest jay]

If someone was able to log in via your Google account, a Google password change is good advice, if you haven't already. Then, you should probably submit a support ticket at help.openai.com to see if they can help you log out of all sessions. Open the chat widget in the bottom-right of that page, then click through the most relevant options, then you should get to a point where you type in your ticket, which OpenAI support can review and get back to you on. Good luck! 🤞

[clear brook]

Thank you!

[sick pecan]

except when the malware is on the workstation, then it can just grab the new ticket without ever seeing the password

[cedar stone]

So I am asking ChatGPT to add a vertical box in HTML to the left of my site, and it wont do it. This is so simple to do, is ChatGPT able to do this? Surely right?

[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.

[cedar stone]

Got it to work, had to start a new chat

[spark shard]

.... are you not using it to generate your lotto numbers?

[sick pecan]

yeah. it used to give me winning numbers and now it doesnt <insert gpt getting dumber trope>

[boreal lily]

if you accidently deleted your chats,

can you recover it

[trail blade]

hi guys, how can i generate images in Dall E because it keeps saying " I was unable to generate images based on your request due to our content policy". Even though I want it generate image of a fictional character like goku or luffy from one piece

[near pivot]

Thats copyright protected content

[peak snow]

Unfortunately DALL-E can't depict copyrighted characters

[late brook]

is there anyway I can know how much message I talk with ChatGPT in lifetime ? I just want to know how much I use it with my work

[dense halo]

They got rid of all the girlfriend gpts on the store 😢

[flat gorge]

No more ComputerLove?

[cedar stone]

I gotta admit, ChatGPT is pretty meager for coding

[sick pecan]

it tends to go for solutions that would be understood by the lowest common denominator. if you are a seasoned coder you can tell it how to improve and it generally does

i.e. modularise this into a class, include error handling, allow placeholder values to be overridden by os arguments, etc

[cedar stone]

I'm a newbie to coding but I'm telling it to move a box to the left and it moves it to the right. "Move this box to the left so it sticks to the left edge of the webpage" and it moves it to the center and changes 90% of the values of the code, like shape, etc. I don't understand what I could be doing wrong. That's a pretty simple thing to do

[peak snow]

What happens if you tell it this? "You made it move to the right, not to the left"

[cedar stone]

"I apologize for the oversight. I've adjusted the CSS to move the featured content to the left. Here's the corrected CSS:" it will send updated code, but it never actually fixes it. it'll just move it around the web page lol

[waxen rock]

That's just how chat gpt works, if this is not what you expected, please alter your expectations

or any language model in general

Mileage may wary on the training set

Also @cedar stone if you want to make the whole ordeal with css a bit easier, I would look into tailwindcss. It has tons of pre made classes that makes it easier to write your layout and style your content fast, so for instance if you want to move an element to the left and keep it to the left you can use flex classes directly in your html for instance

[orchid pollen]

its your prompts

name what it should move

move div.element.box to the left to div-sidebar

as example

iam coding a huge page just with gpt and i have absolute zero coding knowledge xD

already got a decent working setup

https://cdn.discordapp.com/attachments/1154834067808854038/1195870296880717904/opera_VyRKm2JRdO.gif?ex=65b59000&is=65a31b00&hm=e5fa06bbfe6bce2b0c8c9ecdd5b630a6cdb277a8bf4d460c73bde480877c8936&

[cedar stone]

Yea I just moved it manually. But yea you're right, I'm thinking maybe the prompts are not great. I did some courses several years ago and forgot most of it, this is like day 1 coming back so I'm kind of starting over 😬

[orchid pollen]

just keep doing and then you get better

thats how i made it 🙂

still havent read a single documentation xD

[mint vapor]

Is there a way to back up all of your ChatGPT chats without removing or reordering them?

[orchid pollen]

There was a problem preparing your chat. Please refresh the page and try again.

grrrr xD i keep getting this error anyone else?

@cedar stone can you chat right now?

[hoary gazelle]

"OpenAI Quietly Deletes Ban on Using ChatGPT for “Military and Warfare”" mmm

something is brewing

[maiden fractal]

why did you need to double post?

[soft parcel]

is it just me or is GPT4 super slow

[oblique fractal]

It fluctuated a bit more than usual for me today but nothing too crazy

[modest ingot]

Are conversations in teams encrypted both in transit and at rest, using robust encryption standards like AES 256 and TLS 1.2+ like enterprise?

[cosmic junco]

can custom gpts edit their knowledge

[formal heath]

They cannot themselve edit but can certainly visit API to fetch. if i understood your question.

[vale heath]

I can't access one of the GPTs I created anyone know whats up?

[formal heath]

are you getting some error or you are unable to find it?

[vale heath]

GPT inaccessible or not found

an orange error at the top of the screen

[formal heath]

can you find it on https://chat.openai.com/gpts/mine? if not then you might have accidently removed it.

[onyx sorrel]

It's currently down it seems, your chats will show up eventually

[formal heath]

chatgpt is not down and working fine for me. For status you can visit
https://status.openai.com/

[viscid valley]

I kept getting something wrong message

[onyx sorrel]

It works for me now too, but for a minute all chats were unaccessible (with an orange warning that this chat cannot currently load)

[formal heath]

Yeah sometime this happens. Maybe related to system load.

[viscid valley]

same happened to me. kept getting something wrong message and next time I refresh the page, it's unable to load history

[onyx sorrel]

It works for me now, all chats appeared. If it still isn't working for you then just give it a minute and it'll be back. Small hiccups like this happen from time to time

[viscid valley]

thanks for the tips!

[tacit vortex]

help, python codes now always outside code block.

[formal heath]

ask it to use codeblock for code most likely going to fix it.

[delicate rover]

I still can’t access the new GPT store, it’s still the old UI, anyone experiencing the same thing? It’s so frustrating

[viscid valley]

welp it's working when I use warp

but kept getting the message when I turned off the warp

[tacit vortex]

i use gpt to start build a script from function 1 to 10, when add the 11th, it auto delete function 7 or 8 etc. very smart.

[spark shard]

The Network Error. Has anyone had an issue with a network being generated on a created GPT that might be a bit more analytical? It seemst to behave like a timeout.

Due to time constraints, I was not able to complete the analysis for subsequent chapters within the current session. Would you like me to continue with the summary for...

And then it goes on to say network error, which, as you know, gets rid of the text box, so I can't tell it to continue. :).

Has anyone ran into that or has a recomendation?

[dense halo]

🔫

[tacit basin]

who got gpt-4

[tacit vortex]

i dont believe this is a good discussion place, too many ppl too many topics. i will say may be good for instand news like : SYSTEM DOWN etc

instant*

[open mortar]

Does the using the plugins go into the 40 msg limit per 3 hrs?

[tacit vortex]

very crazy, gpt suggest me to use fastcopy s "copy(no overwrite)" mode which does not exist!

[glad igloo]

where can i see the store

[serene stag]

anyone getting issues rn?>

[wraith dawn]

I've got 30 bots in the gpt store, all with very similar names, only the first word is different, yet, when i search it only shows 10 of them and its the same 10 every time. If i search the only word that each of them has thats different they are there. Odd.

[stable bronze]

i just met "Error saving draft" and "Agent not found" error when creating Custom GPT.
i'm on the ChatGPT Team tier

[dry coral]

I have the same issue, I upgraded to the Team Tier and can't save my Custom GPT: "Error saving draft" and "Agent not found" The Team Workspace is extremely slow as well, it's impossible to chat and navigate.

[stable bronze]

😭😭😭

[frosty laurel]

Anyone else facing issues with coding prompts at the moment? Every single response I get A block of code, then plain unformatted text that should be in the same code block, then code again and the explanation is formatted as code in a code block. I've tried custom GPT's, different prompts and new chats but they all amount to the same garbage responses.

This started some time yesterday and was intermittent but now every single response is useless. Too bad I can't upload a picture here but could @languid valley or whoever works for openai please stop messing things up every couple of days? I'm getting so tired of these issues. One day it works great, the next day it's useless. Stop making changes when it works fine...

[fervent birch]

I'm getting a "something went wrong" response to every message right now. Regenerating doesn't help, and starting a new chat doesn't help. Anyone else seeing this?

[prisma furnace]

40msg just network error ....

[regal flax]

Yes, same issue. The code formatting is all wrong, and it isn't coding properly, either. It's making simple mistakes and oversights that ChatGPT 3.5 makes. It makes me wonder if we're being quietly transitioned to a slower version of GPT 3.5 while under the placebo effect of believing it's still GPT 4. It's so bad!

[fervent birch]

this is on my laptop. same account, accessing gpt4 thru iphone app, works fine

[prisma furnace]

Even after a year the same errors are still not resolved in Chat GPT 4 and the same problems they add things and forget to solve the current problems.

[regal flax]

I am having to paste GPT 4 responses into GPT 3.5 and get GPT 3.5 to format the code properly.

[glass fulcrum]

Yea same issue with gpt-4 code formating and often network errors here in EU too :/

[dusk forge]

@glass fulcrum I apologize for the confusion

[uncut vale]

Hi anyone here has Chatgpt plus or team?

[maiden fractal]

Yes

[tacit vortex]

hey, my plus end at 1st, my frd end at 10th. we wanna join team. how? what happens?

[naive holly]

chatgpt4 is unavaible? I can't click on the option to purchase.

[uncut vale]

Nice,I can't make a sci-fi ebook cover using AI image generators like Leonardo AI and Imagine AI. The results aren't what I want. Can you help me create a good cover for my sci-fi ebook?

I used Bing AI image generator to generate an image of the cover of my ebook and the results r what i was expecting but I cannot use that image of the cover

[dense halo]

I keep hitting the limit after sending 20 messages. Anyone else experiencing this?

[round radish]

Trading

[mortal hill]

hey! i'm looking for someone to share my chatgpt teams subscription with!

to lower costs

[dense halo]

I would but I’m worried about privacy

Does the 100 limit also apply to personal chats or just team chats

[mortal hill]

yeah me too, share your concerns with me and maybe we can work something out

just the team, you need to cancel your plus plan

[dense halo]

I thought we could have both

[mortal hill]

i don't think admins can see chats of other users

[mortal hill]

yeah only if don't cancel your plus plan!

[dense halo]

I’m not planning to. Are you sure that the 100 limit only applies to team chats?

[eternal rock]

Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.

Please report any incident by sending a DM to @languid valley immediately.

[mortal hill]

yeah i'm pretty sure

because your plus and and teams subscriptions are separate

[dense halo]

That sucks

Why am I locked at 20 messages. This is bs

Give me the full 40 at least

[mortal hill]

i'm not sure but i haven't been able to hit the message cap yet

[tacit vortex]

he just need to run away next month.

[mortal hill]

what

[dense halo]

What?

[mortal hill]

anyways just pm me if you're interested in sharing

[tacit vortex]

down again? my android app said try again later.

[dense halo]

Sure thing. Lemme just think it over

[lament spruce]

what system prompt does chatgpt have?

[dense halo]

Ok something weird is happening with my OpenAI account

I hit the limit in one gpt4 session but when I opened a new chat it works? What’s happening????

Ok I did further testing.

Turns out that when I type I can still continue chatting but when I switch to voice chat it tells me that I’ve reached the message cap for GPT4. Should I report this???

Did further testing. When typing it sometimes tells me I’ve hit the cap but if I keep retrying it bypasses it.

[outer steeple]

Can someone answer the following questions. I'm a little confused at the moment:

• Which version is currently running ChatGPT for "Plus" and which version for "Free" and with which respective context window size?
• Which version is currently running the Custom GPTs with which context window size?

If someone could answer that for me, it would be a dream 🙂

[grand ore]

The message caps on the Team Plan sub. what is the message caps for each person?

[feral zinc]

currently 100/3h

[grand ore]

for each person @feral zinc ? I think I am gonna go for it with my wife. hm

[fervent birch]

error went a way, got4 worked again for a little while, now the same problem has returned

[feral zinc]

yes, https://help.openai.com/en/articles/8801707-what-is-the-message-cap-on-chatgpt-team

[grand ore]

OH damn. so it is 25 for each person.

So what if u buy it but u are alone. I wonder if u have to buy for 2 seats

[feral zinc]

what? its 100/3h

[grand ore]

Price is what I mean hehe

[feral zinc]

you will pay for 2 seats regardless if the other one is used or not

[grand ore]

So what I understand it is 50$ in totalt then.

[feral zinc]

yep if you pay one year ahead

$60 if paid monthly

[grand ore]

👍

[mortal hill]

hey! i'm looking to fill out a seat for my chatgpt teams subscription for $30 a month, let me know if you're interested!

[mortal hill]

would we be able to split this? i could add you as a second seat

[grand ore]

Not atm. talking to GF about this if we should do this together or not. haha. I mean. it is alot of money. ty for asking !

[mortal hill]

haha okay thanks for letting me know! ignore my DM in that case

[warm osprey]

Hi guys, anyone else experiencing very slow and incomplete responses with gpt4?

Im getting network error on my response.

[obtuse temple]

Yes me! It's giving fragmented responses, refusing to reply with substantive code and erroring out constantly. Mostly alleged "network errors", but my internet connection is fine!

Does anyone have a solution?

[warm osprey]

Same!! Thats the word, fragmented responses! Pls help us!

[obtuse temple]

What are you using ChatGPT for?

[warm osprey]

Coding partner, i want to believe this is happening because my conversation is getting too long?

[obtuse temple]

It does refuse to process questions that are too long. I push it to the limit but without going over.

[warm osprey]

This happened to me yesterday but then went away, I think it might be network congestion

[obtuse temple]

Well I'm a paying member. If there's congestion, they should throttle the free users first!

[warm osprey]

Im paying aswell, I agree with you. Is there a support channel to contact them? Or any devs from openai here?

[obtuse temple]

That's what I want to know too. Where is the human support?

You can leave a message by clicking on the question mark help icon in the app interface but I don't know if they'll read it.

I already did that.

[sonic folio]

You think free users can access ChatGPT-4 Turbo? No. All users of the free service are only able to use ChatGPT-3.5 Turbo, a completely separate service.

The throttling is to manage paid users' demand for the highest tier of service, ensuring that paid users get a fair chance to access what they are paying for.

And yes, OpenAI support is terrible.

I have tickets over 2 months old without a single human interaction.

[orchid pollen]

i keep getting Hmm...something seems to have gone wrong. this is very frustrating

it "steals" my 50 msg i have

[obtuse temple]

When we get ChatGPT errors, does that consume credits? That would be bad and adding insult to injury!

For example, your response is too long, or network errors, or the vague "something has gone wrong".

[sonic folio]

Consider creating a custom GPT and uploading your own support records. Then start asking it questions. You can train the model in this way on the bad support.

The idea? The corp can't hoover up data indiscriminately AND run garbage support. Eventually that garbage is going to tank the model if it's repeated enough across users.

The natural response to the transparency this "mirror" affords is to improve support or censor input regarding OpenAI.

Consider the press response to each decision and then you can understand the idea.

[warm osprey]

Yeah I think the congestion is happening at the premium users tier level, simply too many paid users using gpt4 haha

[sonic folio]

It's not supposed to count errors as used prompts, but that doesn't prevent it occasionally happening.

[obtuse temple]

If so then ChatGPT should be giving appropriate/relevant error messages rather than these nebulous ones that make users think something else is wrong.

[sonic folio]

https://discord.com/channels/974519864045756446/1070006151938314300

[warm osprey]

This whole ordeal once again proves that human interaction is needed on special support cases

[sonic folio]

Every support case at a minimum needs human review. They represent people and their problems with the corporate products.

[warm osprey]

Agreed

[sonic folio]

See above to be part of the solution.

[rain talon]

Since when can chatgpt not access external websites? "However, I can't browse external links" I am using cgpt4.

[orchid pollen]

yes it does

[bitter spindle]

I couldn’t get it to view a link I provided directly, it likely can only access websites it can search for.

You could try asking it to search for the website.

[obtuse temple]

I hope they don't nerf it. I'm also annoyed with the amount of litigation around OpenAI right now. Content creators need to chill. The benefits outweigh the downsides. ChatGPT is indiscriminately training on, not wholesale selling people's works. It's function is not dissimilar to Google, except that it removes the need for users to trawl through the web. We are on the verge of a second industrial revolution and these litigators are the equivalent of Luddites of the Industrial Revolution trying to sabotage it.

They must be diverting an insane amount of money taking on the world's corporate giants in the courts.

[sonic folio]

Nah, it definitely had real-world impact on ad revenue and referral programs for Big Content. It's not always zero-sum, but stealing views with plagiarized ad-free paid-access service justifies litigation!

[obtuse temple]

For the sake of argument, let's grant that is true. Then the conclusion is AI would be too expensive to pursue, and we can't have nice things and progress technologically. As a content creator, I would honestly support the training on my works, seeing the massive upside.

If every content creator wanted a piece of the profits, there would literally be nothing left for operations and R&D. We just can't afford it.

[sonic folio]

The solution is simply to not plagiarize through design, ethical uptake of data, and forwarding referrals and ads.

[obtuse temple]

I'm not sure there is plagiarism through design. I think its the exception to the rule if cases can be cited. And I wouldn't want the app plastered with ads or ad referral links, as I think that would make it unusable, but ChatGPT is already linking sources as it is.

[sonic folio]

If the system was using RAG to read articles there was plagiarism. If it used training data to reproduce articles there was plagiarism.

[obtuse temple]

The litigators want monetary compensation. That's out of the question if AI is to be financially feasible, but source linking is doable.

[sonic folio]

If it attributes false quotes that's libel. Another matter to be settled separately, probably.

Guardrails are needed because IP has value.

Failure to design the guardrails enabled plagiarism.

At massive scale no less.

[inland heron]

can we get back to chatgpt here? This seems more like #ai-discussions

[obtuse temple]

I dunno man, I just think content cockblocking is going to result in handicapping the technology.

[sonic folio]

The suit is over ChatGPT.

[inland heron]

The 'suit'?

[obtuse temple]

*lawsuit

[inland heron]

There's a ChatGPT lawsuit?

openai lawsuit?

[sonic folio]

Yes

[inland heron]

interesting. do you have a link?

[obtuse temple]

Many. New York Times, publishers, literally every category of content creator is jumping on the bandwagon and suing MSFT and Open AI

[sonic folio]

No link allowed

[inland heron]

ah, you mean because of plagiarising?

[sonic folio]

See above

[inland heron]

I'd like to read that but it's too long and I can't give it to chatgpt because I'm losing prompts to errors...

quite annoying

I got work to do, but is anyone else experiencing issues currently?

[obtuse temple]

Technical issues with ChatGPT all day.

[sonic folio]

That's the price of increased availability by fiat currency!

Higher load and crappy service.

[inland heron]

would you mind dumbing it down for me? I'm not familiar with the issue

nor about what exactly constitutes a fiat currency

[wheat wharf]

Inside Ms copilot app chatgpt/Dall-e don’t want to be outsmarted by other AIs 😂. I asked it to give me names of ai image creators that are better than it

“Hello, this is Bing. I’m sorry, but I can’t answer that question. I don’t compare myself to other AI image generators, and I don’t think it’s fair to say that any of them are better than me. 😊
I can create images from text using my own algorithms, and I can also help you with other tasks, such as writing, editing, searching, and more. If you want to see some examples of what I can do, you can ask me to generate a poem, a story, a code, or anything else you can imagine. 🎨“

[visual nova]

anyone having problems the last day with codeblocks just ending randomly in the code? I'm pretty much having to go to the api for any coding for now because of it

[sonic folio]

In brief, taking money for added access costs everyone equally in performance of the model.

[sonic folio]

Yes. See my last response here.

[obtuse temple]

Go to the api? What do you mean?

[visual nova]

I have to do my code in the playground to preserve the tab formatting bc the codeblock markdown in chatgpt broke

[obtuse temple]

I can confirm this bug. Where can I use the playground? In a sandbox you mean?

[visual nova]

platform.openai.com it costs money per token

[obtuse temple]

You mean like making api calls?

[inland heron]

It's astonishing to me that we are the ones who still keep paying for errors. Even those errors caused by openai's systems, we keep paying for them.

If it's a network connection error, I don't mind paying for this. But to keep paying for every single thing that goes wrong? This mind boggling

but companies do this, and users/customers always pay for everything..

[visual nova]

I guess, you're only looking at 1 side of the equation though, if I look at how much value the code chatgpt has generated me considering i didnt know anything about coding a year ago, its like an ocean vs a spilled glass of water.

[inland heron]

Luckily api doesn't get any errors, so my app is smooth, but anything I do in browser and then I pay for a lost prompt just hits me in the head every time

[inland heron]

That's one way of looking at it. I created an app while not knowing anything about python 4 months ago. I guess I should look through some things, but when it starts happening on a daily basis is when it starts becoming irritating

[rain talon]

So my url was LinkedIn it seems openai has stopped reading LinkedIn.

[visual nova]

I remember even in the beta of the web browsing it couldnt read linkedin because of robots.txt, are you sure it ever did?

[rain talon]

Nope not sure just funny how they stopped specific sites from scraping.. I get it.

[visual nova]

Its on linkedins side, like how musk stopped people scraping twitter a few months ago

[rain talon]

Nice 👍

[sonic folio]

More than likely LinkedIn is blocking bots now as a result of the GPT store via use of industry -standard robots.txt.

Search for LinkedIn developer solutions for API calls to the site via custom actions.

[boreal lily]

when will GPTs be out for free users

[honest jay]

There have been no announcements made about such plans, except for maybe this line in the ChatGPT Release Notes from 6 Nov 23:

Plus and Enterprise users can start creating GPTs this week. Later this month, we’ll launch the GPT Store, so people can feature and make money from their GPTs. We plan to offer GPTs to more users soon.
https://help.openai.com/en/articles/6825453-chatgpt-release-notes
Again, no clear indication that there are actually plans for free access to any GPTs.

[oblique fractal]

AFAIK GPTs are powered by GPT-4, I don't expect them to be available for free.

[boreal lily]

But i think you can make them for 3.5

[honest jay]

That's not currently possible.

[boreal lily]

so custom GPTs are for 4 only

[honest jay]

Currently, yes.

[brave quail]

Anything I can add to my custom instructions to stop Gipity from repeating the question in the first sentence of the answer all the time? It's kinda irksome, especially when it calls every problem exciting or nuanced or whatever lol.

[oblique fractal]

I don't think I saw that option. But you could create some equivalent through the API or playground in which case you could select 3.5 and have dirt cheap token rate. Unless something undocumented is happening in the background GPTs are currently only using custom instructions and allow for knowledge to be uploaded, and to fine-tune your instructions through the prompt builder. If you had a plus subscription you could copy/paste the generated instructions and use them in playground and fizzle with settings like temperature.

But maybe in the future we will have the option to fine-tune our GPTs but currently you would need to pay for it on a token-based rate. I would assume that it would require a way to bill the user as both a mean for OpenAI to cover their operating cost (since they now host and run a GPT with different weights) and for the creator to entice them to use that model vs. running their own platform and use the API. Highly speculative and I am hoping for something like that. I don't have such an idea myself but we could get some really creative GPTs and solutions.

[tight island]

im making one entirely for free use, it's great and super good, if u ask me. i wont take a dime for using it when i release it.

[elfin stratus]

gpt 4 has gotten worse

gave me dart code in 4 languages

[oblique fractal]

You mean free for plus subscribers or you plan to use the API and pay for the token and not charge anyone? In any case it's great. Or is there other option available?

[tight island]

ohh in that case u dont mean customgpts

you mean assistants. connected to a platform connected to a gpt.

mine works with the "free" version of the "MyGPT's" i have one for assistants but unlikely to release it.

[oblique fractal]

It definitely feel free to me, at the end I am paying for a Plus subscription. And looking at the gpt store I am amazed by its content that I get ... for free.

[still hound]

Is GPT-4 lagging while writing for anyone?

[tight island]

ill be taking inspiration from all the GPTs ill find.

[vapid kettle]

no it didn't.

[elfin stratus]

yes it did

[tight island]

hmm no

depends on how u use it, if u just use the same convo for months ofc its gon lag

[vapid kettle]

hit the regen button and you can make it any way you want. Smarter or worse.

[tight island]

u also need to put actual work into your prompts.

[still hound]

I get Error in message stream and bad lag it writing. Anyone else? GPT-4

[vapid kettle]

countrary to what people think, it is NOT getting updated daily, weekly or even monthly lol

you are still using the model from last year

"regenerate"

LLMs are indeterministic, especially with default settings. Meaning you will get different responses at different times of trying the same exact prompt. The more you are familiar with it the more effectively you can use it. And yeah, all the talks, at least 99% of them, about it getting worse and worse... Are simply false. 🙂

[tight island]

the more quality the prompt the more quality the output.

But generally, what i'd do is create a CustomGPT aka "MyGPT" for 1 USECASE

Explicitly for a singular usecase.

[digital prism]

sometimes its garbage in diamonds out 😄

[tight island]

true

[digital prism]

agree though, if you talk like an expert it will continue the trend

but thats not all there is.. its a huge web of human language and we're sailors on some life boats

[tight island]

if ur writing a story about X, Design it all around writing only this one story.

and only that particular thing., nothing else, you'd be surprised.

[oblique fractal]

Yep. And then next time I use it I wonder why a zero-shot at a complex task with a prompt that lack context doesn't yield result.. But I still think that a few things changed since the model released. Like behavior or something related to security/safety that couldn't be ignored.

[digital prism]

right and its not just 1 gpt you're talking to, pretty sure its some collective collaborating on the chat

[tight island]

it determines which kind of gpt based on its instructions and the first iteration i think

generally depends on the subject and the question posed.

or so ive heard.

[slim compass]

Hello ! who else have "Error saving GPT" when you try to update your custom gpt ? Please if anyone have help, can't found answer

[tight island]

Reload the page.

try again, but u will lose your changes.

[slim compass]

I did a lot of time, even reload, i already have lost my changes

[tight island]

did u exit your browser.

and log in and out of OpenAI

[slim compass]

Am i the only one who already get this messages ?

[tight island]

or Restart, or even better turn of any VPN if u have one.

[oblique fractal]

I am looking forward to ways to input command to the collective along with my prompt. e.g. see the top 3 different branch and pay 3x the token.

or count at 3 messages toward my usage

[slim compass]

I try it, still same issue i have to wait my limit of using chatgpt 4

[violet bloom]

Boy, ChatGPT is slow today... Waiting over for a min now for GPT to even start. facepalm

[dense halo]

are plus users able to upload pdfs and ask questions based on those pdfs?

[hoary gazelle]

yes and no, it gets a lot of network errors, and the output is not that good

[tight island]

sure.

i have Teams, i have no issues

[hoary gazelle]

i guess depends on the lenght of the pdf

[dense halo]

if i just want to have it summarize things for me, etc, it's allowed? i feel like i heard a while back that the file upload feature was removed

only 19 pages, not massive

[hoary gazelle]

it should work

[dense halo]

this isn't a feature for free users, right?

[hoary gazelle]

i think not

still, 19 pages, you can read it fast, godspeed fresco

[dense halo]

yeah i could and have been, but i want to mess with it a little using chatgpt

[oblique fractal]

As soon as I select Gpt-4 I have the option to attach a file. The limit is likely linked to its context windows, for anything longer you would need to use embeding and perhaps build a GPT or use the api but I have not done that yet. All I know is that 3,500 pages is too much for a GPT to handle as a single file as-is. It suggested that I break it down in multiple files, I have just not tried it yet.

[dense halo]

mine is only 19 pages so i assume it won't be an issue

i mostly plan to use it for smaller documents and just summarizing or asking questions without using ctrl + f

[oblique fractal]

should be fine, generally a word is 3/4 token and you can guestimate from there.

for the odd case you might want to use playground and access GPT-4 turbo but you would need to add credit to gain access. Playground let you access 3.5 turbo which would give you a 16k context windows and it's dirt cheap

[dense halo]

👍 ty

[oblique fractal]

Oh and playground let me copy/paste that 3,500 pages document into the prompt. My brosers kinda crashed on me but it did copy up to the very last characters. So pretty flexible overall.

And if you intent to use 3.5 Turbo for its 16K context windows you will have to manually select 3.5-Turbo-1106, it is (still?) considered a preview version.

[still hound]

holy cow what the F is wrong with GPT-4; its unnaceptable how they nerfed it. takes a bajilion years for it to do anything and constant errors.

[sonic folio]

Too many users with money to buy more prompts, regrettably. Heavier load means degraded performance.

Whee.

[still hound]

can't they increase the servers or whatever if they got so much extra money?

[sonic folio]

Not the next day.

[wraith dagger]

is gpt having problems right now?

[sonic folio]

Yes!

[wraith dagger]

thought so lol gpt 4 is having a stroke trying to load responses

thanks for the reply though!

[sonic folio]

No worries. It was just the same conversation as the very most recent one. XD

[novel parcel]

it's down?

[drowsy egret]

someone may be able to help me? is there a way that I can feed a GPT a website of rules and have it ingrained into its memory? like there is 40 rules I just want it to know to help me play RP games better

[sonic folio]

Yes. Just make sure it's not blocked by robots.txt!

[drowsy egret]

okay can you explain to me some theory or something, I just dont understand how

tysm for responding btw

[sonic folio]

There's no theory. You tell it to "Use web browsing to analyze your web site's rules address."

Make sure web browsing is turned on.

Then you tell it on the next line: "You always follow the rules found by using Web Browsing to analyze your web site's rules address."

But you replace the open vague part of what I said with your url.

[drowsy egret]

tysm, sorry I got muted for the way I speak. I forgot the rules of this discord. =/

[sonic folio]

No worries.

Good luck to you.

[drowsy egret]

hey thanks thats so kind of you 🙂 best wishes to you

[tight island]

ohh yeah we have rules

i've been taught that the hard way

[buoyant rampart]

I have, too 👀 even over something innocuously said

[tight island]

Dino has been especially patient with me

[serene horizon]

You all having a lot of responses get cut off and go red with some kind of error today?

Almost every single response today:

Error in message stream

[still valve]

^ same

[edgy plank]

My gpt can't finish a single msg anymore whats wrong with it these days

[warm osprey]

Yep there seems to be a lot of fragmented issues and network congestion from what I gather

[buoyant rampart]

Hmm, I didn't notice that but then again, maybe it was because I was using 3.5? I had a simplistic question to ask it plus wanted to save my limits for actual work

[edgy plank]

I think these issues are only affect 4 not 3.5

[buoyant rampart]

That's what I thought, glad I used 3.5 then since it served my purpose anyway

[dense halo]

i still don't understand the response difference between 4 and 3.5

[wild anchor]

I get error in message stream non stop today

slow at responding, errors, constantly having to refresh

[still valve]

same, past few days for me. Seems like a problem but nothing posted on status. Closest I could find was this discussion thread: https://community.openai.com/t/code-blocks-are-split-in-code-and-text-replies/586244/5

[simple pasture]

https://canary.discord.com/channels/974519864045756446/1195813593216729218

[sonic folio]

It's like demand went up when they started selling extra access, and it had a deleterious impact on the stability of the platform. Welp.

[hoary gazelle]

Hmm...something seems to have gone wrong.

cut them some slack, they are a tiny start up with 1000 employes and millons of dollars

mom and pop ai shop

[buoyant rampart]

One of the biggest I can think of is that there’s a lot less limits on 3.5, if even any at this point, but 3.5 also means no browsing, no DALL-E, and no plug-ins or using custom GPTs

[hoary gazelle]

essentially selling 2 tiers of subscription now, aka plus and teams. Choke thoses pluses so they pay for teams

[lyric stream]

when is the GPT store coming out

[hoary gazelle]

it is already here

[lyric stream]

why dont i have it

[hoary gazelle]

"explore GPTs" option? top left, under the chatpgt logo?

if not maybe it has not roll out in all countries

[lyric stream]

im in the US

[buoyant rampart]

You a Plus user?

[dense halo]

Hi

[lyric stream]

yes

[hoary gazelle]

you need to be at least a peasant plus user, or a royal teams user

ahh

[eternal rock]

`` Server Tip: ``

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

[hoary gazelle]

odd

[buoyant rampart]

Try clearing your caches maybe then

[dense halo]

I’m new here is there a chat bot here?

[dense halo]

oh... it's right

no browsing is a problem by itself

just a few seconds ago i had a problem with it

[buoyant rampart]

Depends on your use case, I had a question that didn’t need a browser and it worked well for that

[dense halo]

yes

but really, chatgpt doesn't do any search.. so bad

[buoyant rampart]

It does when you pay for it

[dense halo]

i would rather paying my internet bill and do the search by myself instead of creating the longest paragraph to an AI that i bought to get the best answer

[edgy plank]

u don't have to create a long paragraph

[dense halo]

🆗

bet you missed the part "to get the best answer"

[buoyant rampart]

I don’t do long paragraphs either because I’ve found if you’re concise and specific enough, that’s all that’s needed to get the best possible answer. A long paragraph means AI might miss some things I’m wanting from it.

I pay for the sub though since work reimburses me for it anyway

[dense halo]

and chatgpt 3.5 doesn't even know anything about itself or OpenAI.

if i would ask it about its feature, like the voice call feature, it would say "There is no feature to have voice call with me." and never understand it

[buoyant rampart]

That's been my experience with all LLM, they don't know jack about themselves more often than not

[dense halo]

or if i would ask a basic question like "how many people made you?" it would say "I have no specific number.", although if i would ask "Is it like a hundred or thousand?", it would stay in the position like: "I don't know the number of it."

someone ought to teach them. it is real miss.

discord had clyde and clyde didn't even know how to do some settings on discord settings

and the worse is, clyde didn't even know it name

now there is no clyde (idk why)

[buoyant rampart]

Regarding there being no more Clyde, I have a theory on that. Discord probably realized they couldn't maintain the cost of running it so they axed it

[hoary gazelle]

rip Clyde

[buoyant rampart]

That said, I do wonder why Poe app would have it listed still but that's irrelevant to ChatGPT

[hoary gazelle]

online browsing seems to not be working at all, always network error

never finishes one output in before jumping to network error and everything red

[upbeat nymph]

I remember seeing an announced feature recently that ChatGPT will now retain knowledge across “sessions” or “conversations” and that this feature could be toggled on or off. Is that real life or was I hallucinating?

[sonic folio]

Vaporware. Again. I had a session open while it was live. It was amazing.

[upbeat nymph]

So it’s planned but not yet released?

[sonic folio]

I got timed out for answering that with my own personal observations.

I'll figure out how to rephrase.

For some mysterious reason it wasn't online for more than 10 minutes.

It seemed to coincide with reports of degraded performance on non-Custom Chat-GPT-4 .

Then the problems mostly persisted through launch of the store.

As evidenced by this server's search function, easily.

It took a lot of characters to say what happened and not get muted. Wow.

[dense halo]

Can I instruct GPT4 to only read specific pages of a pdf document to answer my question?

[sonic folio]

Try it. "Read only the text from page..." and then see what it does.

[upbeat nymph]

Thank you for the response nevertheless!

[broken shale]

Me too … (in Japan)

[wide pike]

Chat GPT and my school hate each other

I get unlimited response at home

in school, I get liimited response...

[dense halo]

lol

[wide pike]

how to find chatgpt api

[main hemlock]

in open ai website

means u need to login ur ID

[lethal grail]

or their a people already doing it with discord bots which takes the cost and maintence off them

[buoyant rampart]

Could be that too, yep

[delicate rover]

I am a plus user but I still cannot access the GPT store after a week, that’s so frustrating

[buoyant tundra]

how common is it for chatgpt to get a simple math equation wrong

[red lynx]

I am the only person in my team, and I want ChatGPT Team subscription, do I still have to purchase it for two people? I do not want to share the second account with anyone since they would be able to see what I'm doing in the team...

[plain talon]

Is anyone else getting "chat.openai.com has unexpectedly closed the connection" ?

Currently only my phone on the same network as my other devices is able to access openai.com and chat . Everything else keeps getting the error mentioned above

[ionic geyser]

yay we can generate 4 images/query now!

[viral cargo]

after the update, do you also have the impression that there are a lot more errors, and the chat is more crashed and slower?

[solemn plaza]

Am I crazy or is there no feature to search your chats on the web ChatGPT, but there is in the mobile app? Seems bizarre?

[buoyant rampart]

You're not crazy, just looked for it on the web and didn't see that either

[solemn plaza]

Thanks, thought I must have been missing it somehow!

[buoyant rampart]

No problem! It is weird though but maybe there's a good reason they haven't added that on the site

[inland heron]

anyone else experiencing lags in gpt4 output? It keeps cutting out and then continuing resulting in code being shown outside its proper format