chatgpt-discussions

[tacit vortex]

remind u ppl: my acc banned is a team acc. if someone got a large team with YEARLY subscription, you could loss EVERYTHING- money + data without you admiting a thing!

the whole team is wiped out.

[keen badge]

Nope

"Error creating image
Error creating image

It seems we're still facing technical difficulties with generating the image. While we're unable to create it at this moment, I'd recommend trying again later or using the design description provided to work with a graphic designer who can manually create the logo with the specified elements and color palette.

In the meantime, if there's anything else you need help with, please let me know!"

[gentle hull]

when is sora available, any news?

[cobalt idol]

no

[cobalt idol]

just this
https://openai.com/blog/sora-first-impressions

[gentle hull]

ty

[wispy jolt]

How about now, is Sora out now?

[thin fox]

I can't get into chatgpt on my home computer but I can get onto it on my laptop..anyone else having these issues? how do i get support

[glass fulcrum]

Not out yet

[plucky dagger]

man, chatgpt is sooo slow. it takes 4-5 minutes to give me a an answer. Any workarounds? could it be browser problem?

[gentle bronze]

4 to 5 minutes? that's long

[plucky dagger]

and its not even complex or long answers.

[maiden fractal]

Try a fresh new browser to compare

Like firefox if you are using chromium based

[gentle bronze]

or same browser in safe or private mode

[maiden fractal]

Do you have a vpn ?

[plucky dagger]

will try! no im in no vpn or anything its just super super slow i even stand up for a coffee and come back and still writting lol

[thin fox]

Dys Topia helped me solve it, but to answer your question no i didnt' use a vpn

[maiden fractal]

❤️

[gentle bronze]

I didn't do a thing, it was all you

[raven portal]

is the reply feature new or did i just never notice it?

i can highlight parts of chatgpts response and hit "reply"

[glass fulcrum]

0-0 ? I didn't know that it exist

[raven portal]

anyone else? try highlighting parts of a chatgpt response text

a quote button will appear

hit it

[honest jay]

It's been around for a little while now, yes! Handy feature! I can't recall exactly when it was added, and I think it was enabled/disabled/re-enabled at least once while they were initially launching the feature.

[raven portal]

so what does it do?

[glass fulcrum]

What it do? Cus never used it

oh same question

[honest jay]

It will basically just tell ChatGPT "this is what I'm asking about specifically". Like if ChatGPT gives you a bunch of code and you want to ask a specific question about just one small part, you can use the feature and say "What exactly is going on here?"

[gentle bronze]

it's a neat feature

[glass fulcrum]

Smart, i tought it was something similar what gemini adv got

[raven portal]

neat

[gentle hull]

I just tried gemini, it's even worse than chatgpt at disobeying

I wish they would come up with a power user option that we could give red lines to that it can never disobey

[ocean escarp]

ChatGPT disobeyed generating dark humors. So lame

[gentle hull]

My requests are simple.

Hey chatgpt when working together dont keep apologizing and writing 16 pages of text while we're going through steps

Gemini cant seem to do that either

[ocean escarp]

Couldn't u do that one by one?

[maiden fractal]

thats normal

[ocean escarp]

Yeah lol it's just an annoying feature in the standardized AIs. So, people still have to manually come up with dark humors on their own for meme purposes.

[honest jay]

CoT is a kind of LLM prompting strategy right? Like ToT and MoE?

[winter rampart]

is it NOW possible to give instructions to the AI inside a py code?

[brisk path]

Is it just me or is ChatGPT 4 like super fast lately

[uncut basin]

Hey Guys, Anyone has issue using the code interpreter feature right now ?

[raven portal]

imagine not being able to outsource your "dark humors"

we live in a society

[raven portal]

because thats a horrible prompt, how would gpt even know what to do with that?

i swear to god 90% of complaints about gpt here are not a problem with gpt at all
the problem is usually horrible prompts with very broken english
a large language models input is language, so how could it possibly be effective if the input language is bad?
trash in = trash out

[maiden fractal]

Having ci also change chatgpt answer

[tight island]

@raven portal I wholeheartedly agree, prompting quality is important, and if you give a bad prompt and get a bad result the issue is on the prompters side

[tight island]

I agree as well

[faint merlin]

Wait till you ask it to help you write a better prompt! It is soo much more polite in its wording than I’d ever be.

[magic zephyr]

Hey guys! I just saw the Temporary Chat function in my iOS app. When you click on information, it tells everything about the memory feature. Checking the app, I didn't find anything that activated the feature, but the fact that it appeared in my app indicates that it must be very close to being released, right?

No option for this appeared in the PC browser, not even for the temporary chat.

Anyway, I got excited here lol 😁

This temporary chat option appeared today

[willow hinge]

What is “information”?

[magic zephyr]

Sorry, I meant that when you click on the temporary chat mode for the first time, the information about the memory function appears

However, when closing and opening the chatgpt app, I just noticed that the temporary mode simply disappeared 😭

Apparently it was some temporary bug 😕

I'll keep hoping it's close to release the feature...

[raven portal]

am i the only one who will be disabling the memory feature immediately on day 1?
or at least leave it enabled to see what it memorizes and disable it later down the road?
i just see it more as a downside than an upside. i like that every new chat is a completely fresh start.
in fact, one of my worries when i first started out using llms was that they remember stuff and i was thrilled to find out that they dont

for me personally i dont see the point, maybe you guys can explain how it will be more than a gimmick

[willow hinge]

Gotcha. Thanks

[willow hinge]

I write short stories and can’t stand when they use the same character names over and over. That is something that would be useful. It’s not tracking your conversations. It’s “remembering” what you tell it. For example “never use the name Timmy in stories you tell”. They already track all your conversations. Unless you have teams or enterprise. Then they just claim not to.

[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.

[magic zephyr]

To start a completely new chat, just activate the temporary chat function. What doesn't worry me is the fact that it has the temporary chat option. Maybe I would worry if we were forced to use the memory function.

[raven portal]

not necessary, i dont need memory so i will just disable it, which is an option

[magic zephyr]

The idea is to make things more practical. I have many doubts regarding the automotive mechanics of my car, as one of the examples. As mechanics is a very extensive topic, sometimes I want to talk about suspension, sometimes about electrics, sometimes about engines.

Currently I have two options: 1. I centralize everything in a chat about automotive doubts, resulting in a highly extensive chat, or 2. I have to start a new chat repeating the entire context about my car in each chat I start, the repairs made, the model, my driving style, etc.

[magic zephyr]

Yes, of course, that is, there are two ways of not using the memories resource. That's exactly what I want to say: For that reason, it's not that bad. In reality, anything that gives you a choice is always better.

Would you rather never be able to use a specific feature or have the option to use it or not use it whenever you want? That's having a bit of control, which I particularly appreciate.

[raven portal]

memory is probably not was would be helping you here, use the global custom instructions

[raven portal]

never said its bad

[magic zephyr]

Specifically talking about the memories feature, I think that in short I would say that it will make it easier and more practical in some cases.

[raven portal]

that was never a question

[magic zephyr]

"i just see it more as a downside than an upside."

If that's not saying it's something bad, then I don't know the concept of good or bad anymore lol

[magic zephyr]

I never said it was a question. Just a comment I made, in the form of a question.

[raven portal]

i personally see it that way, yes, thats why ill be disabling it

[magic zephyr]

I've been using this feature since day 0, and in my use I can say that it's more limited than a memory feature.

[magic zephyr]

And the fact that you can deactivate it, I see it as more of an advantage even in your context xD

I just want to say that if there was no option to disable it, I would agree with your statement that it is more of a disadvantage than an advantage.

[raven portal]

for my personal usage i see enabling it as more of a bad thing than a good thing, that was my theory and why i was asking for people to explain to me how it can be useful

[magic zephyr]

But I understood that. In my case, I gave you an example, and there was another comment giving an example of a story too.

[raven portal]

overall i think the option to have memory is a good new feature, my point was that its just not for me

[magic zephyr]

I understand you perfectly

But over time, if you have the chance to test it, you might not find a good use for it?

[raven portal]

great, thanks for that

as i said i will give it a chance and see what its remembering

but rn in theory i cant really imagine a scenario where id have it enabled

[magic zephyr]

At first glance, it really shouldn't be of any use to you. But who knows, who knows, over time some idea of use will appear 😁

[raven portal]

ill probably end up disabling it

[magic zephyr]

I think it would be quite fair if the feature could be transferred to whoever wanted hahaha

[raven portal]

memory makes a lot of sense in the grand scheme of things, people will want an ai that remembers what they said and adapts to them
if i was in charge of an llm i'd add it too, even if i personally wouldnt use it
its especially great for casuals/normies and less tech savy people

it adds a hint of magic for them when chatgpt suddenly remembers stuff they mentioned a while back, it makes it more human

[magic zephyr]

"it makes it more human"

This is one of the things that makes me want this feature lol

anything that humanizes an AI is welcome 😁

[hasty sonnet]

Q. Regarding CustomGPT & FIle.

If file was uploaded & csv/xsls - it runs pandas to read it. I add a big file, so it reads a small chunk.

Can the whole file be added into the context?

What files will be readed as a whole file?

[honest jay]

#community-help message

[tight island]

assuming it does not go over the context limit then it can read it all.

also Uploaded how?, Knowlegde? or user upload.

[honest jay]

When Evan from the OpenAI team gave a demo for GPTs (link to recording: #1201923397349548093 message), he mentioned that "short documents" can have their entire contents retrieved by the model, but that all other documents are accessed via a semantic search (jump to roughly 41min 15sec). He didn't specify what constitutes a "short document", but to me it sounds like he didn't mean "as long as it's shorter than the context limit", as filling the whole context limit seems pretty large to me. But I am not positive on this myself! Do you have any more info on this?

[raw pilot]

Hi everyone

Is gpt 5 gonna be better than gemini 1.5 pro ? What do you think 😄 ?

[glass fulcrum]

yea it will

but won't release soon

[raw pilot]

😦

[glass fulcrum]

gpt-5 probably will be a competitor to gemini 2.0

well you better wait for gpt4.5

[raw pilot]

And do you better use chat gpt 4 or gpt 4-turbo playground ?

[glass fulcrum]

i only use chat gpt-4

don't use playgronud or api

[raw pilot]

Ok 😄

[cobalt idol]

chatGPT is supposed to have the same model - gpt-4-0125-preview (but only 32k context)
but you generally have more control in playground

[hasty sonnet]

User upload. For Knowledge - i believe it's just RAG

[tacit vortex]

pls tell openai cs that in chatbot, one can choose chatgpt plus, but only "other" for teams. my msg on plus was served quick but for "team", looks like never reached thx.

[ocean escarp]

Yeah, that's unfortunate

[untold gulch]

I'm rather confused. ChatGPT seems to know about Dirty Pipe, despite it being discovered after January 2022.

https://chat.openai.com/share/1bbb9f08-590f-458d-b530-b0dd8e50f40a

User: Tell me about a computer vulnerability named "Dirty Pipe"


ChatGPT: "Dirty Pipe" refers to a Linux kernel vulnerability discovered in March 2022. The vulnerability, officially designated as CVE-2022-0847, was named "Dirty Pipe" due to its similarity to the previously patched "Dirty COW" vulnerability and its exploitation of the Linux kernel's copy-on-write mechanism.

The Dirty Pipe vulnerability affected Linux kernels versions 5.8 through 5.16.11, and it allowed unprivileged users to gain root privileges on affected systems by exploiting a race condition in the Linux kernel's implementation of the copy-on-write (COW) mechanism.

In essence, the vulnerability could be exploited by an attacker to modify read-only files, leading to potential privilege escalation and arbitrary code execution. This posed a significant security risk for Linux-based systems, including servers and personal computers running affected kernel versions.

As with any serious security vulnerability, the Linux kernel developers promptly released patches to address the issue. System administrators and users were advised to apply these patches as soon as possible to mitigate the risk of exploitation. Additionally, security advisories were issued by various Linux distributions to inform users about the vulnerability and the availability of patches.


User: When was your last knowledge update?


ChatGPT: My last knowledge update was in January 2022.

(Note: "Dirty Pipe" seems to falsely flag OpenAI's AI moderation quite a bit, so be aware of that)

[marsh anchor]

Wasn’t there an update that gave it information up to April 2023?

[vague lynx]

another downage?

[prime swallow]

Depends on if he's using 3.5 or 4.

Additionally, the models may be spot-taught various things at any time, they do seem to improve the models whenever, and that requires some new updates and new knowledge, even if it's not everything. So it's possible a model might know 'a thing' that's more recent than the typical timer for its training cutoff.

[lunar lichen]

Hey everyone, I've been wondering about this for a while and couldn't find reliable sources elsewhere: Is the GPT-4 model used in the web and mobile apps the same as the GPT-4 turbo model?

[marsh anchor]

Gotcha gotcha. Thank you for the info.

[elfin birch]

From the man himself—Sam Altman—19:15 from the OpenAI Devday opening keynote:

A small one, ChatGPT now uses GPT-4 Turbo with all the latest improvements, including the latest knowledge cutoff, which will continue to update. That's all live today.

[lunar lichen]

Thanks for the reference. So it is all recent version then. Thank you so much!

[hybrid totem]

can someone that has access to gpt 4 please input this to it
I know that the answer is c but I don't know the steps.
https://cdn.discordapp.com/attachments/1223059000963371114/1223061867715629166/063AD738-73F7-4D87-A505-B9C1D7C3CECA.png?ex=66187c20&is=66060720&hm=3820287617fbcfb2523b3e182acceb6320cebe2da472642bd9ec06f4cdddf5e6&

[jolly scarab]

Did chat gpt4 degraded or what? Barely can give any useful code 🤔

tho, respect for refund policy, very fast and fair

[peak snow]

I haven't been having any noticeable issues, have your prompt styles all remained the same?

[jolly scarab]

Somewhat, it's more like human communication 😄 Gave him code snippet to tweak, told what i need. But gpt 4 became very lazy to what I remember few months ago, it skips chunk of codes. That other new gpt on opposite, gave very good and detailed code that made a very complex task working

[visual nova]

is the "hm.. something went wrong" on when using gpt4 a common thing?

i can have 3.5 chats right now but i cant do 4, tried 2 different devices and 3 different browsers, relogging, clearing cache

[wet shell]

It's very slow and throwing errors for me as well

[violet yew]

same

[tender topaz]

Anyone have any advice on organizing thoughts amist the plethora of AI to focus singularly on on task or project?

I can write "Create a custom chat box."
Then that's not going to keep me focused. Plenty of ways to do it

[visual nova]

yeah, gpt4 been down for a few hours completely, hopefully it resolves itself.

[timid bramble]

pretty sure they are updating dalle 3. Got an email from them earlier

could be wrong tho

[merry wagon]

Its still down, nothing on the status page

[timid wraith]

I'm getting a token every few seconds, seems perfectly fine

[onyx ore]

Is chatgpt down?

oh yep yall are saying it is

[eager fjord]

hi I seems to be getting a "Unexpected server error"

[cobalt idol]

https://status.openai.com/incidents/r5vx858m9pxt

[buoyant berry]

bro

[eager hull]

same

[shadow osprey]

what dumb update was pushed to GPT 4 again, it was fine for 2 weeks, now it's like I am arguing with my girlfriend as well.
"Use 'IF' statements instead of switch"
"I understand your request, but I'm unable to fulfill it."

jesus, why am I paying for this if it's gonna act like this

[cobalt idol]

out of interest, what do you get if you ask for cutoff date?

[maiden fractal]

i suggest #prompt-engineering

[prime swallow]

Oh yeah, send all the disgruntled folks to us there again 😄

Seriously though, Shunrai. Sorry you're having trouble. Sometimes a conversation 'goes off the rails', is this still happening if you start a new conversation?

Sometimes you can save the current conversation by going up above the problem and editing your previous input (even if you just save it as-is, it sends it again from that point in the conversation) which forces the model to generate a new response from 'up there, back in time'. You can probably get an answer that doesn't have the problem, and then continue from there without trouble.

Depending on what the trouble is!

If you're having even new conversations messing up, definitely come talk to us over in #prompt-engineering where we can take the space and time to help dig through what the problem might be and how to work around it by altering how we talk to the model.

[prime swallow]

If you're interested in trying to get help, that is 🙂

[dense halo]

no

[keen mesa]

just go claude 3 opus bro gpt 4 is over atm

[molten nebula]

bro

wth, they bringing out these voice things now, hello I AM A HUMAN. Existential crisis

[sick junco]

uh oh

[tight island]

ayo

[ember otter]

GPT said my question might violate the content policy or TOS. But it makes no sense

Is GPT Stupid?

[cobalt idol]

newest turbo model loves to (more than the previous version)

[ember otter]

I asked about being able to consciously manipulate the fabric of space in order to move faster than light

[feral zinc]

then you can share it here?

[ember otter]

Superpowers are against the TOS

[tight island]

entirely depends on what u base it on.

if u base it on an already existing story it can trigger that or a brand name.

[ember otter]

But now, when asking about destroying Mars with an FTL Tungsten Rod, nothing

[dense halo]

Wen memory for germany ?

[tight island]

i mean this is a fair question, but even a needle would do the trick

[wet shell]

How does one combat one's disillusionment with GPT-4? It seems incapable of completing a structured task, following directions, or doing anything but providing listicle summaries of a topic. I wonder if the GPT-4 that I used to interact with back in April of last year was real, or if it was just my own hallucination at this point, LOL -- that version that was seen in the famous "Sparks of AGI" video, where did it go?

[ember otter]

Huh, didn't think about that.

That'll make my Tabletop character's life easier

[tight island]

anything faster than light does major damage

[ember otter]

I could carry around ball bearings, like a pouch

[elfin birch]

We do not know about further roll-outs right now

[old walrus]

is chatgpt down?

[glass fulcrum]

not for me

[elfin birch]

There were issues as reported by https://status.openai.com but it seems to have subsided; no issues on my end.

[old walrus]

nvm it works

[narrow viper]

I have similar laments. IMO, ChatGPT has been significantly nerfed since release. But I'm inspired when I have a long discussion with it that just involves conversation, and correlation of concepts. LLMs process text. That's all they do. They are not engines for processing structured tasks. They are not Oracles of information. The quality of discussion in v4 is extremely high - compare to v3.5. As soon as I ask ChatGPT to "do something" as a part of a Custom GPT or through the API, I'm somewhat disappointed. I look forward to any v4+ update that improves on some of this ... but I'm still not expecting AGI out of this. Setting proper expectations is in the recipe for success and happiness with these tools.

[keen meadow]

It’s not just about expectations. The difference in GPT-4 when it launched vs. now is bonkers and heartbreaking. I have been a ChatGPT Plus user since the day it rolled out internationally but finally had to cancel my subscription a few weeks back. Imo, GPT-4 is a distant third place to Gemini Advanced and Claude Pro right now. It’s madness. Maybe it’s just devastating memory/attention limitations thanks to all the new ChatGPT features added over time. I really hope GPT-4.5 or at least GPT-5 brings OpenAI back to form.

[narrow viper]

I think we're saying the same thing. Early GPT4 was much better than now in terms of its ability to follow prompts consistently and incorporate detail from context into subsequent exchanges. But when comparing its ability to find and discuss related topics - the heart of multidimensional vectorization and embedding - it's really freakin good. I'm just trying to separate expectations. The more we expect AGI the more disappointed we get. The more we distinguish between LLMs in terms of what they are designed to do, I think the GPT line continues to be superior. That said, it's all rolled into a single package - GPT is an aggregation of the LLM, its training on data, and other things that it does in terms of being intuitive and linking topics. We can't separate that stuff out to pick and choose features between platforms. So it's really frustrating when the great one we have messes up so frequently - when the prodigy child misbehaves so badly.

[wet shell]

I think that even if we give up the idea that there were 'sparks of AGI' (those MIT and microsoft guys just hallucinated all that too, LOL), even if we just focus on conversation, the 'depth' of the responses has been greatly reduced -- I can't drill down to details with it anymore, I get glittering generalities, like a 1000ft overview, and when I try to 'drill down' I can't get to the level where anything is concrete. But that's something that it used to be able to do -- 10 months ago, I could get down to specific examples and discuss them, now it'll just start repeating itself, or telling me how hard the problem is

[tight island]

i mean the only thing claude and the others have going for them is contextm Sure OAI could release a 1 million token model, but what good would that be, u cant afford it anyway. RN they're working on bringing cost of tokens down. OAI in terms of management of their LLM is vastly superior.

[wet shell]

I've looked into it and those huge context windows are mostly smoke and mirrors, it surely can pull out some specific bit from that context window some specific (seemingly high) percentage of the time, but that doesn't mean it's keeping track of everything there or that everything in the context window is having significant influence on the current response -- granted, I haven't tried Claude Opus yet, and I am curious about it, but there seems reason to doubt the usefulness

[tight island]

agreed, this is why i stick to oai, they're way more reliable.

[wet shell]

Here's what I really want -- I want to be able to go to it with a topic and get a giant info dump of everything it knows about that topic down to the tiniest detail -- I want it to write me a 5000 page textbook on dinner plates, their variations and history (or whatever, LOL) -- it's not possible with the current gen -- one can take high level outline and recursively try to fill it in, but you'll get duplicate sub-trees all over the place, wasted tokens that you still have to pay for, diminishing returns that seem to come sooner and sooner with each iteration of GPT-4

[narrow viper]

Completely agree with this.

[wet shell]

One thing I've been trying to do for the last couple of days and having no luck at, is the problem of finding things that are missing from a category or set of things. Like "here is the criteria for inclusion in this list and here is the current list: ... Can you suggest to me some things that are missing from this list" Maybe this is just a limitation of the current transformer architecture, but it absolutely cannot follow the given criteria when suggesting new items for the list, It's the same as if I gave no criteria at all, it ignores that every time

[narrow viper]

I think you're confusing the abilities of GPT with the ChatGPT application. I just noted the opposite situation in #gpt-models . But your points are still valid. There are ongoing technical discussions about how all models tend to do well in retrieving data from the beginning or end of context, but not from the middle. More context won't help with that, only better algorithms for processing large context sets without ongoing nanny prompts where we insist that it continues to review existing context.
So, you're talking about the technology, not really limitations or abilities of ChatGPT.

[wet shell]

I flip-flopped on that a couple times but didn't clearly indicate, sorry for the confusion 🙂

[narrow viper]

I do take a lot of care in ChatGPT to get the GPT-4 model to consider things that are Not in the current context. That is : tell me something I don't know, that I might not be aware of, and don't tell me to go look for information because that's why I'm using a chatbot. But yes, the current GPT model, and perhaps the code involved in processing the chat context, is not adept at filling in gaps like that. But sometimes ... it's still amazing.

[wet shell]

But I'm unclear myself, ex, in the last example, as to whether the inability to choose new items that fit a set of criteria is a limitation of transformers in general (like inability to count accurately), or if it's an endumbenning of GPT-4 specifically 🤷

I feel like it used to be able to do that, but I can't prove it, I think there's a lot of that going around LOL

[narrow viper]

I hear ya. Consider, if the API queried the GPT4 model and it got a similar result, then the issue is with the LLM, not the ChatGPT application. There are limits imposed on the ChatGPT context number of tokens and on the API - I don't recall ATM if they are the same. But that's another distinction between model and app.

I feel like it used to be able to do that, but I can't prove it, I think there's a lot of that going around LOL
Yeah, we're going by a lot of warm-fuzzy intuitive perceptions here. "ChatGPT __ seems__ dumber than it was several months ago, and on a day to day basis I'd swear I could tell that "something" was tweaked back or forward, but without notifications or our ability to run actual traces or other diagnostics, we simply can't measure this with anything but our senses.

[wet shell]

I've worked as a software dev on large production systems for a big company that I won't name -- even if they're not releasing tweaked models silently from week to week, I feel like there are devops-level configuration changes happening silently for operational reasons 'whole thing is overloaded, tweak this parameter and do a rolling restart' and then everyone jumps on and complains that it just got dumb 😂 There's not much transparency in what's happening for the end users

[random sand]

im confuse what to use completion api or assisstant api for my usecase. My usecase is that i have to give the website bussiness information to the gpt and ask for summary. Then i want the gpt to create 5 titles of quiz that that are engaging, reflective of the company's core services, and appealing to the target audience. then the user selects the title and the gpt is asked to create efficient quizes from those titles.

[tight island]

actually...

it is...

🗿

if u utilize 50 GPT 4 Previews with an intricate workflow system that allows them to collaborate which is possible aka a Swarm, it can do it.

i dont even want to think about the cost.

i think GPT 5 will be a Swarm in of itself.

[wet shell]

I'll add "for a reasonable cost" to it then 😂 I worked on an automated software dev shop scripted in python at one point, and it mostly worked pretty well, it started with a requirements document, then produced a technical requirements doc, implementation plan, etc, each using multiple API calls to build, but the cost of testing got to the point where I couldn't afford to continue, I'm poor, LOL

[narrow viper]

@random sand stick to #ai-discussions for that.

[tight island]

GPT 4 is a touchy subject in regards to "dumbness" and ill tell you guys why.

[wet shell]

"Alright, that's bug's fixed, It'll only cost me $15 to re-run it again and find the next one"

[tight island]

if you ban a certain area of discussion, u ban and or completely limit subsequent sub areas, and those subareas as well

thats the issue with moderating a LLM

even slightly connected but not relevant to it areas, get heavily moderated as well, thats the problem.

imma just go out on a limb here and say lets say they ban "Creative Writing" which they wont

imagine the amount of affected areas.

issue is when moderating an area specificity is extremely important.

u gotta specify the limit and etc.

[tight island]

token consuming essentially.

[narrow viper]

That's a lot of nebulous commentary. Can you be more specific? I think the more specific someone gets in an explanation of the "slippery slope" concern, the more we find that the "subsequent sub areas" are areas that most people would prefer to avoid anyway. For example, are we talking about stories involving crime or inhumane acts? I don't think moderation of these topics is bad if it shelters casual discussion from accidental exposure.
I actually do think we should have an option to turn off moderation for our own purposes. You and I should be allowed to create really gory horror stories and other content that we might find entertaining. The line between these concepts is that such things don't intersect with normal conversation. Discussion over coffee or with a bot don't normally, casually include notes of blood and guts, mass destruction, or other topics that are unacceptable in public. So I appreciate the default moderation, and I question the motives of anyone who advocates default removal or reduction of their efficacy.

[tight island]

i think higher tiered subscriptions should allow the turnoff of filter.

but things will be monitored and stored.

rather have that

rather have my convos monitored constantly when theres no moderation on.

[wet shell]

that might open the door for it to accidentally produce content that is illegal though... and once it has it's not like you can take it back if it's logged and stored

[tight island]

thats a fair and valid concern.

or just a less restricted model for higher tiered members.

[wet shell]

I'd rather it has some filtering, but I have gotten flagged on inocuous things, or even had refusals of things that had nothing bad or malicious or gross because it's trying so hard not to violate copyright, etc

[tight island]

i understand the copyright thing, but instead of flat out denying it, why not put a twist on it. like..

pikatchu but with horns

🤷‍♂️

i can see the issues arising already but u know where i am going at.

[wet shell]

haha -- if you feed it a ML research paper PDF and ask it to summarize the algorithm from the paper in the form of pseudocode, it won't do that, or it wouldn't last time I tried, because it can't facilitate my evil attempt at copyright infringement LOL

[narrow viper]

but I have gotten flagged on inocuous things
This is where the tooling needs to be improved, not 100% removed. Everything in, um, moderation.

[tight island]

i have gotten flagged on some things as well but i ignore the flag since i got nothing to worry about

more often than not its a misunderstanding on GPT's part.

[wet shell]

but if you take the same paper and chop off the attribution, author's names, etc, it'll do it no problem -- I tried to argue with it that becuase the paper was open-access and published that the authors were inviting people to attempt reproducing the paper as part of the scientific process, and it agreed on all points "nevertheless, I am unable to assist you with this"

[tight island]

that is unless, u make a customgpt for that purpose.

so ive noticed.

[narrow viper]

I need to break. Thanks for the chat y'all!

[chilly mist]

Guys I have problem. I'm trying to log in to my account. The plus membership and my chats have disappeared from my account. It's as if a new account is showing up. On mobile, with the same email, everything works as before. The plus membership is valid, and the old chats can be found.

[maiden fractal]

#community-help message

[shut dome]

Hello, today's announcement about the voice engine is amazing, I want to ask if this service is available if I purchase the ChatGPT plus, or if this is a demo for future releases?
thank you

[steady path]

I just noticed that now GPT 4 says "Usage limits may apply"... do we finally have more than 40msg every 3 hours limit?

[tight island]

i have now reached usage limit on personal and both seats on teams, i have a problem i go sleep now.

[ivory elm]

Ain't no way people are still using ChatGPT in April 2024 🤦‍♂️

[tight island]

ofc claude 2 is bad why use that when u can use gpt

[maiden fractal]

still using chatgpt working great

[dark jungle]

Its a Chrome Tab navigator. Really useful too!

[feral steeple]

Yo guys I have a question... Is Claude AI 3 is greater than ChatGPT right now?

Because last time I check on the statistics based on the Claude AI official website it said Claude AI 3 is way better than ChatGPT? What do you guys think?

I mean I'm still using ChatGPT but I'm still wondering if what Claude AI said is true or not

[rancid otter]

yes everyone i know just uses opus

im stuck with chatgpt for now

[feral steeple]

But besides of greater response of opus , what does opus offer more than ChatGPT?

[rancid otter]

from what i read: more likely to answer with full code. better code. better context retrieval.

I don't like how chatgpt just forget segments of code so when you tyr to follow up it creates variations, incomplete code annd so

althou no dall-e, no custom GPTs etc..

also i think they are experimenting with very huge context

[feral steeple]

I see , well this is very insightful for you to share it. Thanks. Oh but I have last question, do you think that Opus is suitable for university students like me? Because right now I'm working as Exco academics for my organizationfor my university and I want to integrate AI into our learning system. So, I'm wondering if Opus Ai can handle problems more about facts questions or mathematical questions ( correctly ofc) for instance like biology and maths.... What do you think?

[rancid otter]

I'm also undergrad but from what i know no LLM is good at math by itself. I would integrate wolfram alpha to check procedure and answers

or another platform that provides accurate answers

also given that all LLMs are prone to hallucinations i wouldn't trust it to always provide 100% correct facts

[feral steeple]

Wolfram Alpha eh? Interesting, I need to do some research regarding this . Thanks

[wet shell]

I just signed up for it (I'm not ditching OpenAI, just experimenting, I imagine that both will have uses for me and it'll give me something to do when I'm waiting for my message limit to reset) -- API access is in closed beta in the FAQ, so it's web-interface only, if that matters to your use

[rancid otter]

yeah for now all the facts and math that you generate with LLM need to be checked by humans/experts

[feral steeple]

I see even Claude AI cannot escape from blunders and mistakes... Very well. Thanks for the information, this can help me regarding on my decision making as Exco academics. Thanks armored and I hope you have a nice day

[rancid otter]

have u tried code

[rancid otter]

np

[wet shell]

not yet -- but this might not be the best venue to discuss, might be more appropriate over in gpt-discussions

[rancid otter]

ok

[ocean escarp]

Is it against the guideline to have GPT recite the exact phrasings from the provided excerpts? It kept shutting me down, saying they're not allowed to do that

[hybrid violet]

I'm model starved

pls gib new

[rancid otter]

im model annoyed. chatgpt seems dumber and makes me waste time

[elfin birch]

I'm

[ocean escarp]

User
Then where's a part that explains how to read the monitor

ChatGPT
To read the monitor, simply look at the display for your glucose level after applying the blood to the test strip.

To read, I gotta look at the monitor lol

[ocean escarp]

I tried to use a plugin, but it doesn't work

[rancid otter]

You need to specify in your prompt to use Wolfram to verify the math you are attempting

it sometimes create a malformed request to the Wolfram API tho, so if an error occur you better expand the request tag

[ocean escarp]

Like this? Go Go Wolfram Alpha for this geometric problem

[rancid otter]

I remember adding something like "Use wolfram to verify your procedure and answers"

remember to activate the wolfram plugin so it can see it

[dreamy barn]

GPT overwrote my convo again and I lost 20 gens with no <x/x> to get them back

[deft mountain]

thats wild

[dreamy barn]

It has happened about a handful of times it's so sad. I lost lots of art.

[ocean escarp]

I'll try that after my cap restore

[sick pecan]

sharing the best art in discord creates not only a backup, but a curated history (=high quality)

my favourite discord filter: from:drinko has:image (on other servers)

[rocky vine]

when did gpt knows to correct itself without me pointing out mistake

[rancid otter]

why dont you just download the imgs

[calm sleet]

feels like most sessions are regressing back to being a bad google clone from 1998... ie, direct user to link with brief description of what to do.. anyone else having this experience on this version? which begs the question, why not use another tool if we have to practically beg the model to do the work

also this is from sessions with enough clear prompts defining goals for the session etc.. not a fresh session

[tight island]

Thought the same

[cloud pewter]

it keeps popping errors for me rn

[raven portal]

does the wolfram gpt actually give chatgpt math abilities?

btw is chatgpt down for anyone else?

[warped bolt]

Hi can someone taught me about perfect Prompts

[raven portal]

oh its back. it was just down for me

[warped bolt]

Lol

[raven portal]

sure
learn english

[warped bolt]

Nerdcatto

[warped bolt]

🙃 Solly 💓

[raven portal]

its better to use your native language than extremely broken english like this

[raven portal]

that was my actual advice to you for "perfect prompts"

[warped bolt]

Oo lol thanks 💓

Whenever someone talk with me face to face - at that time my English becomes good idk why this happens to me@raven portal

[marsh stirrup]

Then imagine ChatGPT's face in front of you and go prompt it with a perfect prompt

[warped bolt]

Xd ya i guess

[raven portal]

use voice input with the chatgpt phone app then

[warped bolt]

That's a cool feature

Actually it's a breeze for me

[marsh stirrup]

It's such a breeze you might as well start going by BREEZO

[warped bolt]

🙃 hm hm I will

[tight island]

For good outputs do better prompting.

Work on the prompts ensuring clarity and consistency and getting the task properly formulated, ensure that you’re very explicit in what you want with no room for interpretation for other meanings or give it a persona

That’s literally all there is to it

[burnt wraith]

does gpt team and gpt plus have the same underlying models? in discord it says advanced models idk if they mean advanced vs free 3.5gpt or advanced over gpt4 from the plus subscription

i didn't read anything about advanced models on their page

[rich meadow]

I have a question regarding the gpt4 annual plan, is the 100 messages shared between the 2 accounts ?

[keen meadow]

There is no comparison between Claude 3 Opus and ChatGPT Plus with GPT-4. Claude smokes ChatGPT right now. I'd love to compare Opus to GPT-4 when GPT-4 first launched, as that might be neck-and-neck or even to the advantage of ChatGPT, but right now, I'd argue strongly that there's no comparison. Opus is a competitor for whatever comes next from OpenAI. The difference in code quality is especially obvious.

[tough peak]

Chaat Guptaran

[dreamy barn]

I do but when I'm using my I pad I cannot download all of them at all times. I also like to take my prompts it creates and document them too. Making it harder to achieve without my external. WHEN THE CODE GLITCHES and loses the images and CHAT HISTORY it will never come back tho.

Wish they all saved like the use to on DALLE’s site or how MJ does it

[raven portal]

why cant they just give claude browsing

[warped bolt]

Hi

[warped bolt]

Actually I guess now iam getting better output after using The ChatGpt's Voice feature thanks dude 👍🏻

[fleet fog]

so.. anyone know how lazy GPT5 is gonna be? I could always work pretty well with 4, even when people said it became lazy for them. it's absolutely useless to me now, what is this.

[prime swallow]

My advice is quite different from Steve's:

In general, good prompt engineering is

1. pick any language you know really well that the AI understands too.

2. understand exactly what you want the AI to provide.

3. explain this, focusing on what you want the AI to actually do. Using language as accurately as you can, avoid typos and grammar mistakes and communicate clearly as possible.

4. check the output carefully, verify you get what you intended. Remember to fact check, and be extra careful with any math, sources, code, or other details that the AI is known to be especially likely to hallucinate.

[prime swallow]

Care to share 1+ conversations where you're having trouble? Maybe we can see why and offer suggestions.

Me, I'm not having trouble as a rule. There's that rare 'magic chat window' where the model does seem to grab some lower quality training data, then stick with the concepts that are not very useful for my needs that time - but that's an easy fix - new conversation, as those are rare.

[prime swallow]

No, it's per account, and team can have more than just 2 accounts on it.

[crude slate]

I don’t get why refusal messages count as gpt 4 usage

It’s so annoying trying to stop the ai from wasting my usage

Like if I ask it to justify it just continues normally

[prime swallow]

The model doesn't even necessarily know why a message generates a refusal, and may not be able to answer coherently even if it does know - and hallucinations are hard to tell from fact.

Maybe better to discuss here, showing the prompt and output, and get help to find a way to get what you want, instead of debating with the model, using your limited messages there. Every input to the model, ideally, is probably tailored to getting precisely what you do want from the model to the best of your ability.

Why you may have 'some interest' in knowing why a message was refused, I'm betting that you really want what you asked for in the first place, so a focus on how to help you get that is probably going to be more productive.

[zealous delta]

Hey, I'm taking a class in ChatGPT, and I've encountered something very strange. I used it to format text for easy memorization, by turning all but the first letter of every word into dashes. So "To be or not to be" becomes "T- b- o- n-- t- b-" etc. I fed through the Gettysburg Address, and it consistently omits the "o--" for "our" in the opening line. Looking further, it goes off course in other spots, too, such as cycling several times through "so conceived and so dedicated" before continuing.

I'd love to hear from others who try this prompt. Just tell it to convert all but the first letter of each word into dashes, then feed a text through.

[crude slate]

What model?

Gpt 3.5 struggles more with instructions

[crude slate]

(Do not be repetitive, do not start with or include “as” much) Continue seamlessly off the last output with Emeric chatting with Arlise about her vampirism, handing her a small pastry he had pocketed from the kitchen that day, be very slow and long more detailed descriptive and ensure you do not focus on themes or summarize, please ensure you generate at least over 600 words filled with lots and lots of dialogue!!! Please ensure to always end with a line of dialogue, (Always show, never tell)(please ensure dialogue does not feel contrived or forced, but natural with characters mannerisms)(do not generate any forced contrived philosophical dialogue!!!!)

It’s possible it hallucinates because the complex prompt

But shortening it causes undesired style to be generated that contaminates further generations

[prime swallow]

This kind of thing is super hard for the model, because it doesn't actually 'see' language like we do. It actually has to guess what the first letters (or other letters) of words are!

It kind of breaks up our inputs into what are called 'tokens' - kind of like how we break words into syllables, but not the same at all.

You can also test this by trying to play hangman with it, or asking it to write a story where the characters play hangman (if you're using GPT-4, your precious messages per 3 hours; don't waste them sending one letter at a time, hah!)

But the model can track whole words used normally - and barely track how to spell any of them when not used normally, and it can't always track all the words in a sentence when we start to use them in unusual ways either.

You can potentially train the model how to do this, with positive and negative examples, letting it try and showing it the errors, letting it fix them. This doesn't always work, but you can 'teach' it to omit the letter 'E' in a conversation, for example: https://chat.openai.com/share/a6f0afbc-e406-4924-8423-c060a114d811

But a much easier way exists now, if you're using GPT-4:

Have it use Python tool to extract the first letter of each word from whatever text you want it to do this for, using a program it writes for the purpose.

[prime swallow]

Do you find that the 'do not' style of prompting is working well for you?

[crude slate]

Yeah with this format at least

Although if I tell it to not force in contrived respectful dialogue it refuses

I feel like instructions are not powerful enough for the AI, like the further you go it gets weaker and weaker so I have to keep reminding it

[prime swallow]

nods I'm impressed that you get good results with mostly 'don't' prompting, and relatively little 'do' prompting.

Do you have maybe 3 paragraphs that you'd consider ideal writing style you could share?

I'm wondering if I could guide the model with a different prompt style to help give more consistently ideal outputs for you.

[spring iris]

jeeze its so easy to do this with actual programming / scripting languages where it will never make a single mistake

[tight island]

anyone here a D&D Fan, i need a tester tomorrow for my gpt.

[remote slate]

Since when has chatgpt an image limit?

[tight island]

Always

[feral steeple]

Can you explain why now is it better because I'm as Exco academics for my organization and I'm currently exploring ways to integrate AI for my juniors and my friends for their studies

[quaint marlin]

Everone else: GPT-5 LET ME IN
Me: I just want Memory

[nova delta]

Can anyone get GPT to draw birds without beaks? Several attempts and different prompts and still birds have beaks.

[next egret]

Sure whatcha got?

And which system? GPT can do 5E really well if you prompt control it to give all dice rolls to the python tool

[spring iris]

google has been hpying a 1 million token context model but it's probably BS

[warped bolt]

Thanks 👍🏻

[warped bolt]

What!?

[prime swallow]

I'm willing to help explore how to do this, but I wonder, what do you want in place of the beaks?

Are we looking for pictures showing injured or deformed birds, ones whose beaks never formed, or were removed?

Are we looking for pictures of birds with human lower faces, so a smile and a human-like mouth?

Are we looking for pictures of birds wearing masks or otherwise with something covering up or hiding their beak?

[tacit vortex]

for above, i suggest DO NOT use teams yearly plan! 1. openai banned my whole team for unknown reason, the whole team workspace is wiped out, NO REPLY THUS NO REFUND! 2. C3 and GAdv are coming, you may regret if you sticked to one. 3. for plus, as i mentioned in the help sun forum, reply is prompt and at least there is refund. so, suggest NOT to have yearly sub of teams

its been no reply for teams for a week+,.but plus got refund within 3hr. so teams s cs is not readily established. dont!

sub*

[prime swallow]

A little messing around, beaks are hard to hide!

I managed with a puffball bird... #1070954877779464253 message

[nova delta]

Thanks, I just wanted a standard bird without a beak. A robin or cardinal, just any bird would be fine. But wanted to see it without a beak and was surprised. GPT rarely disappoints, so thought I would ask the group.

[prime swallow]

So, without a beak, as if the beak were cut off?

Because 'something' is there. If it's not a beak, a hole into the bird's head, a smooth expanse of feathers, or what?

[nova delta]

I was allowing GPT to put whatever it wanted there. But now that you mention it, I will try with repalcing the beak with a standard frog mouth 🙂

[prime swallow]

GPT absolutely can describe it, but Dall-E is nearly ignorant of negative prompting - instead of telling Dall-E 3 what to not show, we have to tell it what to show. Dall-E has zero understanding of what 'bird without a beak' might be - but 'bird with human lips and a human smile' might work.

[prime swallow]

Or a frog's face!

[nova delta]

AHH! You are awesome. This worked. Many thanks! And I learned something 🙂

[prime swallow]

Share the image with us if you want and if it's fitting, perhaps off in #1070954877779464253

I'm curious 😄

[drowsy tinsel]

ok so now there's no more usage limit for gpt4, whats the point of team plan

[honest jay]

I think there's still a usage cap, it seems like they've just changed how they are displaying the usage cap in the UI. The pricing page https://openai.com/chatgpt/pricing still lists Team as having a higher usage cap than Plus.

[drowsy tinsel]

did anyone manage to hit it yet?

I mean my daily usage rarely even hits 40/3hr so I assume its not very easy

gotta ask some gpts builders im guessing...

[honest jay]

Just tested, for science! Still seems to be 40 on my end.

[prime swallow]

Depends on what you do with the model, and how you use the services.

I can reach my cap in about 90 minutes of normal use, so instead I typically take planned breaks where I'm not using the model, and/or use 3.5 for quite a few tasks, then take the already done-what-3.5-is-good-at to 4 for polishing, using far fewer messages with 4 to get the same quality of whatever done.

[drowsy tinsel]

interesting, so they are just not saying it anymore?

[prime swallow]

Well, the usage limits have historically varied based on demand.

I've seen it as low as 25/3 hours, for GPT-4 use.

Last night around this time, Solbus mentioned it was 'Usage limits may apply' - and I checked and saw it.

This morning (my time), it was back to 40/3 hrs.

I see now that it's again 'limits may apply' - yay!

Maybe we have an expanded limit during this time of possibly lower average use by users!

[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.

[honest jay]

I had the same thought! I take it as a good sign of potential expanded limits, at times!

[idle prawn]

Hi everyone, my custom gpts disappeared,
First I opened a private browsing and it worked for 1 sec then disappeared again and it's now empty everytime.
Thanks in advance for your help.

[white panther]

I have a question... Does some one know, when i buy gpt plus on my iphone, do i have chat gpt plus on my pc also?

[honest jay]

Hey! Can you still see them all listed at https://chat.openai.com/gpts/mine?

[idle prawn]

I don't

[honest jay]

Hello! Yes, subscribing to ChatGPT Plus will give you Plus access anywhere you sign in with that account.

[white panther]

Thanks

[honest jay]

Can you share a screenshot of what it looks like over in #community-help? Be sure to remove any potentially personal info in the screenshot.

[idle prawn]

Well thanks for beeing so fast to answer, it works again, I guess it was a bug.

It did it again, but i can't send screen here

[honest jay]

You can send screenshots in the #community-help channel! Again just be sure to edit out any personal info from the screenshot

[idle prawn]

Thank you

[idle basin]

Yo...i had hundreds of credit disappear from my account

[swift osprey]

why doesn't openai care about gpt webapp 😢

[steep forum]

Problems with ChatGPT at the moment anyone?

[swift osprey]

just the fact that it talks way too much and buries all the important information in an essay every time it responds and custom GPTs suck

[dense halo]

ih

[plain storm]

[stable bluff]

What is the number of petaflops-hours used per day in OpenAI? Where can the materials be consulted?

[native spear]

there are 2 specific bugs that i havent seen anyone talk about:

1. you input something into the chatgpt chat, and as the output is generating - for some reason - you want to take a look at the sidebar and so you open it. Now your chat is RESET

2. regarding the new "reply" feature, there seems to be an issue when you reply to older outputs. For example if you reply to a part of an output that was sent 2-3 messages before, you might eventually get an error that doesnt fix even if you regenerate multiple times.

[modern horizon]

Hey! I have a fix for "Network Error" happening. We need to turn off QUIC in the browser being used. This protocol doesn't play nicely with ChatGPT and will constantly have issues resulting in Network Error problem or just timing out constantly. QUIC is an experimental protocol created by Google not yet support by all routers. It fails often and is easily disabled. At some point this experimental feature was turned on in a browser update and it's causing havoc for many services, including ChatGPT.

To disable QUIC protocol on Google Chrome or Microsoft Edge, you need to
1: Open Google Chrome or Microsoft Edge.
2: In the URL address field, type chrome://flags or edge://flags and press Enter.
3: Type quic into the Search flags field to filter the list of flags.
4: Select Disabled from the drop-down menu next to Experimental QUIC protocol.
5: Restart Google Chrome or Microsoft Edge for your changes to take effect.

If you're using firefox the instructions are a bit longer so please look those up to disable QUIC.

[rich surge]

[eternal rock]

`` Server Tip: ``

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

[boreal perch]

Is it possible to unarchive a chat? It's so easy to archive it by mistake

[prime swallow]

Absolutely. Click on your name or email, bottom left (of the web browser, different way to get to it in the app).

Click on settings.

Under general, there's an 'archived chats' 'manage' button.

Inside that, you see a list of all your archieved chats and can individually unarchive or delete them

[boreal perch]

Thank you!

[boreal perch]

Done! There's a permanent error generating a response for that chat, is there a way to unfreeze it?

[prime swallow]

In my tests, once I unarchive a chat, I can work with it normally, and it shows up on the list of chats again too.

Umm... what's the error message you are getting, or how else do you know there's an error?

Worst case scenario, you'll need to recreate the chat, but you can copy/paste the text into a new chat (especially if the problem-affected chat isn't extremely long) and continue in the new conversation.

[dense halo]

I found out how to use the ChatGPT voice on the web version 😎

[fossil umbra]

ok and how

[elfin birch]

Click the speaker icon underneath ChatGPT's message

It may not have been rolled out to you yet

[hybrid violet]

Holy Smokes

[placid wigeon]

Hi - i have a question - does anybody using chatgpt team pricing thinks it is worth it in terms of higher GPTs? like plus is 40 per 3hrs but like how is for gpt team? and do you think it is worth it?

i'm tryna gather friends so we can pay for the fee each month and so I don't have to like struggle in 40 message caps

[elfin birch]

You should have everything you need here https://openai.com/chatgpt/pricing

Oh, and IIRC the current GPT-4 usage cap for teams is 100 messages / 3 hours

[placid wigeon]

so +$5 and 2.5xcap?

i would pay 5$ ngl

[swift osprey]

I use a two seat teams on my own and haven't hit the cap since I started

I don't know if the cap is shared between both seats or not, but I've only used 1 seat I think? I don't know how it works exactly

plus you are only locked in for a month so can switch back if not worth it

[tardy isle]

Does plus has same cap(40/hr)? Model selector seems not displaying exact cap... again.

[elfin birch]

I have ChatGPT Plus and it says 40 messages / 3 hours

[inner beacon]

Chatgpt down?

[prime swallow]

Each account gets their own 100. You can also have more than 2 people in the team

[median scaffold]

is for me

[uncut creek]

guys since when could u reply to messages?

is that a new feature that just dropped?

[prime swallow]

Slightly more complex than that.

You need to pay for 2 accounts minimum to upgrade to teams, and whomever is in control of the team account (it's intended for business) has a fair amount of control over it. For example, if the account running the team were to cancel, it affects all members; all members are canceled.

From what I've heard various users say, you can involve your current ChatGPT account into the team, but if you leave the team, you may have some issues with accessing your previous conversations. It seems like you will want to pay very careful attention to the difference between the 'team' and the 'personal' workspace, and ensure you keep them separate and keep what you want permanent personal access to in the personal workspace, assuming you let someone else manage and pay for the team account itself.

You also appear unable to have everyone pay personally and privately, the person in charge pays for every seat in the team.

https://openai.com/chatgpt/team

https://help.openai.com/en/articles/8792828-what-is-chatgpt-team

https://help.openai.com/en/articles/8792536-manage-billing-on-the-chatgpt-team-subscription-plan

https://help.openai.com/en/articles/9123978-how-can-i-leave-a-deactivated-workspace

https://help.openai.com/en/articles/8798618-what-happens-if-a-workspace-has-been-deactivated

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

This is really, and clearly, intended for businesses. Sure, a group of individuals (or a single wealthy person, willing to pay $50+ per month to use the seats all for themself) could work with it, but that's not really intended.

[placid wigeon]

ngl if this is 5$ difference i really can tell that openai is trying to sell more seats on their enterprise server or smth else for heavy users

no?

i literally can bring my friend over

i don't have to pay 50$ for 100 cap

[prime swallow]

Right. To get the 25/mo cost, you need to prepay a year in advance for all the seats. Otherwise, it's $30/mo.

Or your friend needs to pay it.

You and your friend need to trust each other, because if one of you changes your mind, or gets mad at the other and wants to mess stuff up, the person paying can cancel or remove the other person's seat, because it's treated like a business account, the person paying is expected to be like the owner.

If that is fine for you, sounds like a great idea for you. Good luck to you and your friend!

[prime swallow]

Not the enterprise server; enterprise is a higher tier and has other requirements and benefits. I think you need at least 150 members to be considered for enterprise, and I've seen a number of people comment in the Discord about not being selected for Enterprise despite requesting to pay for it... though that was months ago since I last saw that, maybe it's easier to get Enterprise access now.

[pine berry]

Seems to me at a certain point ChatGPT will have a network error.

[visual bloom]

Yes. When I edit my first chat, I get a strange error ”dall-etext2image” and no image, even when I have not asked for one.

[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.

[inland aspen]

dies the code interpreter no linger work for anyone ?

[magic zephyr]

Hello everybody! Here GPT-4 appears as "limits may apply" instead of "40 messages every 3 hours".

Then that's it? Are there no more fixed limits?

[glass fulcrum]

i got the same

"we may dynamically adjust the message limit in order to prioritize making GPT-4 accessible to the widest number of people." https://help.openai.com/en/articles/7102672-how-can-i-access-gpt-4

hm mayby they doing that

or flexible limit

[magic zephyr]

At least interesting. I honestly really liked it. It was one of the things that always left me undecided between paying for copilot and chatgpt. I've always tended towards chatgpt, and this now makes me feel safer.

[glass fulcrum]

i think that its something that is guqrranted 40 messages but you may get more but yea im not shure

[magic zephyr]

This information has been around for some time. However, for months now, 40 messages have appeared every 3 hours. For the first time, it appeared that limits "may" be applied.

[magic zephyr]

Or simply perhaps they are gradually increasing and do not want to specify numbers at this time.

[glass fulcrum]

yea or they doing the same what anthropic do that the message limit is counted on how servers are used (higher server usage lower cap)

[magic zephyr]

Which I think is great, especially if they maintain the MINIMUM limit of 40 messages for every 3 hours as it has been for months.

[glass fulcrum]

yea woudl be cool as i often use gpt-4 when it seems pretty fast (in night so i think less peoples on servers that why its fast) so i thoritycly get more messages per hour :o

[modern horizon]

If you're getting "Network Errors" you need to disable QUIC in your browser. It's an experimental UDP protocol that is causing issues across many services including ChatGPT. It was enabled in a broswer update recently and many routers don't support it causing ChatGPT to timeout. After disabling QUIC ChatGPT will become fast and responsive again.

[visual bloom]

I still have the old limit. Apparently they are rolling out these changes gradually as usual.

[tight island]

i have found the person.

[tight island]

it's you, wanna test my gpt later today

[random cargo]

anyone else getting where 3.5 is responding to questions very literally without considering the context of the discussion? it's reading hypothetical questions from me like it's a search engine.

User
how many known species of microbial fungi?
ChatGPT
As of my last update in January 2022, it's estimated that there are over 120,000 described species of fungi worldwide.

basically its answering with what I could get from Google's top search result here rather than responding to the hypothetical. It should at least say it doesn't know how man species of microbial fungi there are, but that there are some ranges for known species of fungi. but instead it gives this misleading answer

[tacit vortex]

hi, rumors no more cap limit for plus, is this real?

no sure if correct place to say, after acc deactivated by reasons i disagree, i applied plus, C3-opus(only API), GAdv to compare. i am a med student, i throw mcq at it. GAdv is weak in the "reports". however, for several mcq, GAdv got the answer and concept 100% correct while gpt4 and opus failed at the answer.

i told plus and opus what the correct ans should be, opus re-think and gave correct explanation again, but plus say something else and didn't think again.

take home: its like classmate in my class,.student A got 85% B got 83%, C got 70%. but for the Q A and B failed,.C may get it correct.

[quasi blaze]

did GPT just have a stroke now i cant loadi t

[tacit vortex]

gpt fx well for me

[quasi blaze]

stuck like this for 10 mins trying to refresh page

"new tab" spinning wheel for 10 mins

[tacit vortex]

are you from evil axis?

oops, i mean using vpn?

[quasi blaze]

no

im in usa, no vpn needed

[tacit vortex]

then this group cant help

this group can only persuade ppl not to use vpn

[quasi blaze]

why would anyone use vpn.

[tacit vortex]

ppl from evil axis not allowed to use this service

btw u can chk openai status page

this is the 2nd thing this group can do

[restive plover]

ChaGPT needs to really increase the limits

[daring urchin]

is this april fools or no

the no sign up announcement

[hardy scarab]

Wait really?

[hardy scarab]

I was going to ask the same question

Nope wait I just tested it actually works

[idle sedge]

New update is dope

Hopefully nothing crashes

[hardy scarab]

That's fantastic oh my

It worked for me. Maybe they're rolling it out slowly

[marble loom]

it doesnt work

[idle sedge]

They will need like 30 minutes maximum

To roll it out.

Because I actually got it

[visual fulcrum]

well this is going to be abused

[hardy scarab]

I promise it's actually working

It says "Sign up or log in"
"Save your chat history, share chats, and personalize your experience."

And you can STILL use the chat

[stoic cypress]

I swear if they just go "April Fools’!" and remove this tomorrow…

[idle sedge]

Fun fact there are lots of other websites that support no signup feature to use it

& has unlimited quires

[frail marlin]

I just saw the new block post and tryd it one a icognito tap. But i have to log in or sign up i cant just start why is that?

[stoic cypress]

other ai websites or just websites in general?

[tiny grove]

its april fools DAWGS

[stoic cypress]

no it worked for me, I just tested it

they said they’re rolling it out, it might not affect everyone immediately

[serene wren]

So I read an article from OpenAI saying that they have (or are creating I cannot remember) something that can take 15 seconds of audio and then the user can type out something, and the user will in return recieve the typed out information in the voice of the 15 second audio clip.

Does anyone know whether or not this is going to be available with ChatGPT+ or is it going to be a separate project? Also, does anyone know when it will be available to the public?

[runic oracle]

I love how much more convenient ChatGPT is becoming to the point that, I wouldn't be surprised that we get a AI integrated computer to help you find your files or prevent viruses or even just talk to you lol

[serene wren]

Kinda do with Windows CoPilot

[dusky cliff]

hello

[storm vapor]

Welp, the internet is going to break it again 😔

[rose glade]

the chatgpt instant thing is a really good idea, a strong step in the right direction, I think this will see a good adoption rate in more marginilised and poverty stricken environments, where I have found adoption to be a negative of AI like LLM's and such mainly because of the sign ups, people want to use it like google and not have to give personal data

[silk blade]

i wonder what the protocol is if instant gpt thinks it's talking to another LLM ?

other than pinky & the brain style nefarious plans

[rose glade]

I think there will be timeouts and stuff like this

[random cargo]

I think that would be based on input times. it doesn't have any ability to detect AI language, right?

or AI speaking style, that is

[rose glade]

AI detection tools that work usually scope outside the dynamics of randomness mathematically akin to mersenne twister and such, so it will be able to detect a human well much like those captcha things

as in the speed in which you type etc...

OR it will simply just have a captcha lol

[silk blade]

the frontend webserver might detect stuff like that, i think gpt just gets a block of text
gpt is an autocorrect with a few kb of memory, input detection is asking a lot from it, its like talking to a simcard
ofc the compute manipulating that few kb of tokens is immense, but it boils down to that.

[random cargo]

if you really wanted it to talk to an LLM you could just slow down the input rate, right? and I usually just copy and paste paragraphs into GPT. you would think it would flag me by now, I have like thousands of copy-paste inputs

[proper urchin]

Here come bots

[silk blade]

i'm pretty sure there are browser extensions for dumping generated output into input boxes, i'm assuming a human proxy really

[west night]

i cant tell if the announcement is april fools joke or not

[stoic granite]

how do i know the latest announcement isnt an april fools joke

[vapid kettle]

It's 50/50 lol

[stoic granite]

r/beatmetoit

wait wrong platform

[vapid kettle]

also, it makes zero sense to even post it in this server. How many users here do not have an account? 🤓
Zero.

[floral echo]

if it was an april fools joke then thats kinda boring

[silk blade]

i wonder if instant gpt will be a LMAO APRIL FOOLS if it's a flop? perfect execution tbh

[floral echo]

april fools: sora has released 🤑

[silk blade]

i choose to believe that they had to do something to cheer up the dev instance of GPT, which was inconsolable after it's xz backdoor got sniffed out

[rose glade]

you know when I upload documents to chatgpt, how can I get it to strictly use them? It takes me like 20 prompts to get it to use them because it thinks it knows better

I try all sorts like USE the document no matter what, putting the filename in brackets etc...

[vapid kettle]

Edit the initial prompt (or the one after if you can't edit one with attached files), until you get the response you are looking for. Once you do you will know how to do it.

[crisp cairn]

who cares about instant gpt if it's true or not... that's not something we are waiting for lol

[haughty ibex]

sam said they'll be releasing lot of interesting stuff this year

[crisp cairn]

i don't see much of use case for it even if true

without signing up you end up with gpt 3.5 anyway

[rose glade]

theres no definitive way, and with the usage caps its annoying

[visual bloom]

I usually tell it to read the document first. To get it to know. Nothing more. The second message is the actual task.

[magic zephyr]

I can't say whether in fact the limits only increase or cease to exist at certain times, but the fact is that in my account and in the accounts of a few others, the information of 40 messages every 3 hours no longer appears. It only appears that "limits may apply" without giving any details.

The information is not rumor, it is fact. I don't know if it's working. I haven't used enough to measure it yet.

I thought the same thing when I saw this release just today. But I also imagined that a company like openAI wouldn't do this lol

[magic zephyr]

I agree, but this somehow can give them much faster feedback. In the medium term, I see this as fewer reasons for postponed product launches.

[tacit vortex]

i wanna share, again, that i put my school s mcq into plus, claude3-opus-api, gemini ultra 1.0, china s yiyan 4.0. result is that, there are chances that plus and opus for wrong but gemini and yiyan got it correct. its more like in our class, that althogh someone got 8x% marks, there are mcqs that this guy get it wrong but those 7x% got it correct.

[hollow quail]

User
I'm just here to test

ChatGPT
Great! Feel free to ask anything you like, run a test, or share what's on your mind. I'm here to help!

User
thing is, openai keeps making access to gpt less restricted but don't seem to have the compute to provide that, so I'm here paying for the service to crash constantly (you don't know it but you often fail to respond) and they're allowing everyone to use 3.5 even without an account

ChatGPT
Hmm...something seems to have gone wrong.

Even if it's not a joke it's hilarious that they'd consider doing this because they can't cope with the users they already have.

[tacit vortex]

they deactivated my whole team without refund, easier access? easier access to my money? remind you it is CloseAI for money

[regal geode]

I suppose after April 1 they're gonna go back to asking everybody for phone numbers to verify before they can use ChatGPT :P

[wise yacht]

Right when I am getting close to dumping GPT Plus for Claude...in-painting comes out. Guess I am sticking around for a bit.

[tight island]

claude is ehhh.

gpt is 💪

[halcyon lava]

FAQ says:
ChatGPT Plus subscribers will get GPT-4 access on chat.openai.com with a usage cap. We will dynamically adjust the exact usage cap depending on demand and system performance in practice. As of January 5th 2024, GPT-4 has a rate limit of 40 messages every 3 hours. Please note that unused messages do not accumulate (i.e. if you wait 6 hours, you will not have 80 messages available to use for the next 3 hours). In certain cases, we may dynamically adjust the message limit in order to prioritize making GPT-4 accessible to the widest number of people.

[magic zephyr]

As I said above, this FAQ information has been around for a few months and I had already seen it. The point here is that the information from "40 messages every 3 hours" no longer exists. And this has never happened before.

It's just an observation and I hope that now this old limit no longer exists or at least has increased.

[halcyon lava]

Yeah. Hope that too

[magic zephyr]

🙏

[halcyon lava]

let's test it out😁

[magic zephyr]

Yeahhh

[viscid narwhal]

I saw it happening before too. OpenAI is stealing users.

[vestal hedge]

would the increased accessibility of GPT potentially slow down the usage for premium users, if the usage rates are too high?

[hollow mica]

Is there a known issue where past Chatgpt conversations might not load? I am getting an error

no sooner I hit enter than it works

[agile pewter]

the servers are down

[dire rock]

It seems server issues

[crude slate]

Yeah

It’s probably them rolling out the features

It doesn’t recognize that I have 4.0

Or my history

[atomic echo]

^ seems so im also struggling

[toxic flax]

NOOOO i need it now

[crude slate]

Had a mini panic attack

[exotic shoal]

same

doesn't work!

[toxic flax]

this is tragic bring it back online

[atomic echo]

yeah lol i just got started working on my assessment for once in my life and then gpt dies

[frank talon]

Rip, I had it asking me interview questions for tomorrows technical interview

[warped pendant]

can't use chatgpt .. no way to login, our teams account code all missing

[toxic flax]

reaaaaly need this to load

[magic zephyr]

Wow, it's gone offline now. Yes, it's in status.openai.

[toxic flax]

great

[potent smelt]

Is it down?

[outer widget]

Yeah it is not working for me either

Yep lol seems like it is

[wild wind]

ChatGPT is also offline for me right now 😭

[undone igloo]

for me too

[potent smelt]

Unauthorized

[atomic echo]

no updates lets goo 🙏

[pallid karma]

😦 okay, so it's broken... again 😦

[magic zephyr]

Serious? It didn't even last 24 hours since the limit of 40 messages in 3 hours was (apparently) removed?

It didn't even last 24 hours that they allowed the use of chatgpt without login

[ancient oak]

I love how I come here only to check if anyone else is experiencing a broken paid for service... and we all are. Cheers

[rocky wing]

So, just come to the server because it's down on my end, checking if it's same for you guys.

[desert jewel]

I both hate and love the rapid testing in production

[blissful igloo]

Well I al not paying for it rn so can't complain.

[dense halo]

is it down??

[toxic flax]

off to gemini I go

[split osprey]

Yep it's totally broken right now, ChatGPT no longer works.

[magic zephyr]

Could it be that there was a ddos attack? Since there is no login screen... I don't understand much about cybersecurity, but I believe this should influence

[split osprey]

Which means I'm gonna have to wing it on scallopped potatoes

O.O

[rocky wing]

I would think OpenAI would have more than enough resources to have pretty sturdy anti-DDoS measures

[fathom schooner]

Welp I guess I can call it a day now.

[wild wind]

It annoys me so much that I pay 22 dollars every month, and ChatGPT 4 is getting dumber and dumber, and wants to do less and less, plus ChatGPT keeps having outages.

[magic zephyr]

Imagine using a currency that costs 5x the value of the dollar? My case.

[worthy night]

It’s down now

[blissful igloo]

People, it is alright. I am a Tech Priest. Let's all join hands in prayer.

[split osprey]

Five nines is not five 10s, give it some time

[toxic flax]

fr chatgpt got brain rot

[desert jewel]

It's a tight position to be in. With Claude getting so much backing from AWS
Gotta stay ahead of the curve

[pulsar wyvern]

ChatGPT has been down every day this week.

[magic zephyr]

Well, it wasn't the first time he fell and it won't be the last. Let's wait.

[rocky wing]

Any further word on when they're gonna roll out that Memory feature?

[pulsar wyvern]

Not the full day but every day it goes down for like a little while

[potent smelt]

At what time did this first start?

[atomic echo]

like 15-20 mins ago

[rocky wing]

I don't much care for the idea of the chatbot remembering anything about me personally, but keeping track of the ideas for my writing seems like it could be helpful

[magic zephyr]

Just imagine when they launch resources like today’s? GPT-4 with higher limit + use of gpt free without login screen.

[toxic flax]

wait ive had the memory feature for a little while now

[viscid mantle]

true that

[sharp kernel]

is gpt down

[warped pendant]

wish I had the memory feature, and I have teams lol

[blissful igloo]

Yeah making custom gpts, with documents attached with info for example is very handy.

[rocky wing]

How?

[desert jewel]

Think there was mention of a lucky few early testers who just got the feature

[sharp kernel]

when is chatgpt going to be back up i gotta finish an essay smh

[magic zephyr]

yep

[rocky wing]

Oh, could you elaborate on that a bit? I haven't toyed around at all with custom GPTs, just the bog standard Creative Writing Coach that came with it

[manic widget]

It was silly to release that without being ready for a big surge even if that surge is short lived because people who bunch of the new users will have a bad time and then forget about chatgpt

[mental rose]

is there something wrong with chatgpt? i cant find a login button and when i try to use 3.5 or 4, it keeps failing

[toxic flax]

idk its been around. It says "memorized" or something with a little clipboard and there is a list of all the info it remembers about you in the settings

[dusk sentinel]

Hey everyone, how about you end every message with ChatGPT is down, so that people stop asking if ChatGPT is down.

[nocturne pewter]

Now when I try to use GPT its not letting me login. I assume this was the change?

[rocky wing]

Scroll up, buddy. Not just you, it's down for everyone.

[mental iron]

Is it down again?

[toxic flax]

guys is it down for anyone else

[manic widget]

😂

[cosmic tapir]

It is down

[dusk sentinel]

https://status.openai.com/

[wide matrix]

it’s down yeah

[nocturne pewter]

I think this is that chnage were you dont need to log it.

[toxic flax]

i think its down

[wide matrix]

came here to check lol

[desert jewel]

I wouldn't be surprised if this was how they roll out the new features.
Just take it down asap

[magic zephyr]

Yes, it's not working. My plus even disappeared. Wait a few minutes or hours.

[wide matrix]

how long will it take to come back? 😔

[toxic flax]

they are just a bunch of comp sci nerds anyway

[autumn spindle]

down?

[dusk sentinel]

https://status.openai.com/

[wide matrix]

yeah

[mental iron]

Thanks. That's helpful

[blissful igloo]

@rocky wing Sure, you enter that, explore thing, make your own, enable some checkbox I cannot remember what it is called.

It let's you attach files such as document and pdf, which can contain say a patent you are curious about, or a information about your fantasy world you are writing.

Quire handy. And you can chat with the gpt which makes the custom gpt for you as well, thus less custom editing needed.

[ocean escarp]

GPT 4 is bwoken

After a new update

[kindred panther]

why is chatgpt downn

[toxic flax]

would be nice if they could announce when there would be downtime

[kindred panther]

fr

[blissful igloo]

It might be intended for code really, but is equally useful for just text info.

[silk phoenix]

Wait so is ChatGPT down for everyone?

[ocean escarp]

Cuz they recently updated GPT

[dusk sentinel]

Is the downtime related to them rolling out the ability to use ChatGPT without logging in?

[dusk sentinel]

https://status.openai.com/

[ocean escarp]

Byeah

[desert jewel]

That' the assumption

[manic widget]

Agree or have a discord bot either updating the status in a channel in here or automatically replying to the questions about it being down..

[rocky wing]

So I don't need to do any custom training for it? Is it basically like talking to the default GPT but it knows the stuff in the documents you attached?

[magic zephyr]

But it may not have been something expected... In addition, there is status.openai, and also a bot on Telegram that notifies you when the service stops.

[ocean escarp]

It's not an assumption

[dusk sentinel]

If that is true, that is really annoying. Paying customers should not suffer downtimes so that they can rollout features to non-paying users.

[ocean escarp]

lol

[boreal lily]

GOT down

[ocean escarp]

Wanna feel privileged

[mental iron]

Are you asking about the schema you need to write for this?

[boreal lily]

GPT

[ocean escarp]

Yeah, GPT update screwed it up

[cosmic tapir]

https://status.openai.com/

[manic widget]

For some reason more people don't know about searching Google for "status servicenamehere" which for the vast majority of tech sites will be a status subdomain

[wild wind]

It works for me again!

[warped pendant]

I thought the April Fools was giving everyone access without logging in, however even our gpt teams account is down as well

[opaque parrot]

this is the 2nd time in a row api/chatgpt has an outage as soon as i start using it

[blissful igloo]

Basically yeah. There is an inbuilt GPT in the editor to help you if you want, set the use for it, rules, tone for the setting and so forth.

I mostly use it handle my endless thoughts on fantasy and all what I have written are attached on .docx which it analyses upon request.

[ocean escarp]

Since the update?

[nocturne pewter]

history and GPTs are gone for me.

[pulsar wyvern]

If they are going to let users use ChatGPT without signing up then all of us Plus users are really screwed..

[mental iron]

I created a GPT that writes schema for the GPTs I build

[ocean escarp]

Yeah, GPT 5 is long overdue. Spent too much time on side quests like SORA

[manic widget]

Yup happened to me too, I hadn't used it for a few days and went to the other day and it was down. I did it today... And it's down again lol

[toxic flax]

i mean ahead of time, not just reporting it when it happens. This is assuming they took it down for an update

[opaque parrot]

yep same here.. almost started to think i was the reason why LOL

[toxic flax]

shes back

[rapid falcon]

Damit i have too many storys started on chatgpt

[wide matrix]

fr?

[manic widget]

Only because competitors are catching up / overtaking in some areas

[warped pendant]

bain of existance is fluff, comment code, and telling it to write full code for a code merge when you give it your full code

[wide matrix]

is gpt back

[blissful igloo]

Teach me, master.

[toxic flax]

FR SHES BACK

[wide matrix]

it’s back

[toxic flax]

i almost died

[wide matrix]

les gooo

[rapid falcon]

At last

[warped pendant]

its not up yet

[rapid falcon]

Works for me

[wild wind]

Me too 🤣 ChatGPT helps me so much in life

[ocean escarp]

There are many new players within the AI war and GPT 4 is falling behind tbh

[cosmic tapir]

Monitoring - A fix has been implemented and we are monitoring the results.
Apr 01, 2024 - 17:45 PDT

[manic widget]

I asked it what I needed to since it's back up, I'm done breaking it for the night 😂

[ocean escarp]

They updated their updates

[rapid falcon]

After that suden logout i freakt out

[autumn spindle]

back up for me

[manic widget]

Yup, I'm currently subscribed to 3 and one of the main reasons I'm not fully happy with chatgpt for me is it's constantly still being lazy with it's replies...

[wild wind]

Has the limit for ChatGPT 4 actually been increased? I haven't gotten a warning for a long time that I have to wait to write with ChatGPT 4 again.

[elfin birch]

Still 40 messages / 3 hours for me

[rapid falcon]

But now Works fine

[ocean escarp]

Long vague messages that don't answer your questions/prompts time to time

[rapid falcon]

Maybe they have a massive overload

[ocean escarp]

Bing AI is broken for you?

[potent smelt]

Is it up?

[wild wind]

I feel like I have no limit at all anymore. I think it's because I snapped at support and complained that I didn't know there was a limit on ChatGPT 4 before I bought ChatGPT Plus.

[potent smelt]

Yup, it’s up on mobile

Just checked

[opaque parrot]

ahaha i'll be on it for a while, bear with me everyone for any furthur outages 😭

[wild wind]

If the limit for ChatGPT 4 had not been adjusted for me, I would have kept bugging support until they did it.

[obsidian bluff]

Is the GPT with the voice conversation available for anyone who doesn’t have GPT 4?

[ocean escarp]

It used to be 50 messages / 3 hours

Before that I think it was 30 messages / 5 hours

[rapid falcon]

Crashed again?

[wild wind]

And ChatGPT is down again! 🙌🏼

[stable acorn]

why does GPT just suck with attachments all the time now? its multiple times a day now it simply cant handle it

[sharp kernel]

what update did they add again?

[ocean escarp]

OpenAI allowed people to use GPT 3.5 without having to log in

And that update is glitching ChatGPT, need to reverse their scripts and undo that update tbh

Not even a big deal

[stable acorn]

honestly we need a new option... OpenAI has been crap for a few months now

[wild wind]

I think I will also switch to a local AI, which is even better than ChatGPT 4. But surpassing the quality of ChatGPT 4 is not that difficult either...

[obsidian bluff]

Does anyone know why GPT changes its mind for certain things? I have given it math equations to solve, it gets it wrong, and I simply say “Wrong”. It apologizes, re works the whole problem, then gets the right answer. This seems peculiar no? I’m pretty new to this, so I appreciate any explanations, if any.

[next dune]

OpenAI keeps trying to be many things. Instead, they should just focus on being awesome at providing real answers. I wonder why they haven’t focused on this. Maybe they aren’t allowed? I don’t know. It’s one gimmick feature after another. All I want is to use ChatGPT to help me figure things out better than I can on my own. Don’t we all?

[stable acorn]

ill have to check it out.

[next dune]

OpenAI used to be awesome. Then politicians got involved and fear of it waking up the masses spread. Now it’s nerfed and a bit of a pain to use. Yet we all still use it and hope for a better version.

[wild wind]

All I can say is that ChatGPT is getting dumber and dumber, so I wouldn't be surprised if something like this happened.

[ocean escarp]

It does that, so it seems unreliable with math questions sometimes, so you have to take a risk whether to listen to it or not. You're better off actually learning mathematical formulas. I've tried that many moons ago, GPT doesn't know where to stand

[rapid falcon]

Dammit next time i need to put CENSURE in place of "F" Word

[obsidian bluff]

Appreciate the answers. Thank you everyone.

[stable acorn]

the system is pretty easy to manipulate. you can get it to change its opinion on a lot of things.... unless its political lol then its pretty one sided haha.

Likely though in a mathematical question it could be the way you input it.

[wild wind]

Np

[magic zephyr]

Apparently for some users, yes, the limit has increased (or may be unlimited at certain times).

[stable acorn]

THIS...

[rapid falcon]

CENSURE chatgpt still down

[ocean escarp]

But it's not related to the math questions

[magic zephyr]

This change of limits must be gradual. Probably in the next few days or at most 3 weeks (as I see in the worst cases), you should have the limit changed.

[wild wind]

My console is spammed with errors about Apis from OpenAI.

[next dune]

Isn’t it like this for everyone using ChatGPT4?

[rapid falcon]

I'm not useing "F" Word again

[elfin birch]

Should be

[full folio]

can confirm its 40 per 3hrs i just reached cap, need to wait now lol

[stable acorn]

correct lol not related to the math question. i was more speaking to you can get GPT to change its responses based on phrasing or leading.

In math questions i would assume multiple answers stems from the AI misunderstanding the question

[magic zephyr]

We have copilot pro, which is very close in quality to chatgpt, after all, they use gpt-4 turbo with few modifications.

The problem is that the copilot still has some limitations, such as not being able to edit previous prompts and if there is any type of censorship, it blocks the entire conversation window and is forced to start another conversation, losing context.

But currently I see it as the best option after chatgpt, as it is faster in responding, and I have never seen it go down.

I can't speak about the Claude 3 Oppus as I've never had the opportunity to use it, but I believe it's inferior if you can't browse the internet.

[next dune]

I think Sam Altman tried to release ChatGPT to the masses last year and he got sacked and GPT got nerfed hard. Anyone else think this?

[rapid falcon]

Chatgpt still down dammit

[ocean escarp]

I mean, when I said that answer was wrong to GPT, it changes to a different answer. I said that one wrong, you were right, it changes back. I did that back and fourth, showing that ChatGPT doesn't have a stance what's an actual answer for the mathical equations. I gave all the information to GPT but still

[wide matrix]

no it’s not

[magic zephyr]

I don't understand, do you still have this limit presented in the model, or after 40 messages you were limited, even though the model says that "limits can be applied"?

[stable acorn]

interesting, ill have to do more research about the options out there. OpenAI made some wild additions to the board recently so I'm sure the direction the company is going will ruin the platform in the near future.

[full folio]

Since I signed up to Plus I counted how many prompts I gave and it stops at exactly 40, sometimes 41, then the textbox disappears with a message saying you can use it again at <time>

[rapid falcon]

Finaly Works on my side

[sleek gust]

nice

[sleek gust]

Why ?

I think more prompts capacity would be perfect

[stable acorn]

it does limit you in excel though. I used GPT extension in excel running on GPT-4 turbo all day at work and run thousands of automations a day. That is tied to a developer account though so that could be why

[rapid falcon]

Dame Two AI Engineers from Poland work in Open Ai what the heck is going on in there that cose this hole issue

[stable acorn]

are you into conspiracy theories lol?

[dusk sentinel]

Is it just me just me or is ChatGPT now faster?

[rapid falcon]

Yea i feel that too

[solar wharf]

they really turned it off and on again, huh?

[rapid falcon]

Maybe they realize that chatgpt 3,5 need Connection to database!

[toxic flax]

maybe cuz a bunch of people stopped using it when it went down

[prime flame]

Now the question, is todays announcement april fools or not..?

[warped bolt]

Televisia polaski 🙂

[tacit vortex]

hi i put a mcq to gpt4. 50% of time it say a, 50% of time it say b. real world dont work like that

[magic zephyr]

We'll know in the next few hours xD

[kindred parcel]

So I have a delicate question, and I figured I should ask the people who might know; I'm doing a presentation on history's biggest failures, and the presentation is supposed to be as engaging as possible, so I figured I could show these people a picture they had never seen before, yet would instantly recognize. Towards the end of the presentation, the leader of Germany from the mid 30s to mid 40s comes up. I'll cut it short and ask, how can I get an AI to compose that image? DALL-E and Midjourney both balked, even though this is for legitimate historical academic use. What would you guys suggest?

[prime swallow]

Mods have stated it's real. The functionality exists, but rolling out, hasn't reached all areas and browsers yet

[inner kite]

Has anyone else experienced this? GPT4 sometimes is very lazy and fails to capture requirements compared to 3.5:

3.5
https://chat.openai.com/share/5d8b78ad-c0fe-4522-b1a9-6e2d8904e19a

4
https://chat.openai.com/share/838f3c99-d763-4ffc-adcd-5a4ba28a2c2f

[craggy gorge]

omg... its laziness is driving me crazy... If I ask it to do a specific task that involves repetition, I hate how it does the bare minimum and then includes a comment that I should do the repetitions myself

[prime swallow]

If you're using GPT-4, it'll often be pretty eager to use Python tool to write script to automate the repetitive task for you, then package the result into a file you can download.

If you're doing something that can't be scripted (most things can though, if they're repetitive) or are using 3.5, consider making it clear why you want it done (provide context) and make it clear you care, that it makes you happy to have it. That often is enough to get the model willingly cooperating.

[magic bough]

Hii everyone,

My name is Anna. I am a recent graduate of Computer Science.
Currently, I am doing research at the American University of Armenia to understand the problems in using large language models.

To all the ML engineers, content creators, prompt engineers, and anyone who uses language models in their work, I would appreciate it if we could chat a little about this topic.

If you would you be open to a quick online meeting shoot me a dm or just react to this message. It will take no more than 20 minutes. Your insights on would be incredibly helpful.

Thank you, and I'm looking forward to speaking with you!

[eager fjord]

Hi guys, I don't know if anyone had tried this before, but chatgpt can execute a "routine“ just using plain language instead of actual codes. I issued this on chatGPT 3.5.
<role>You are an algorithm executor, you understand the logic which don't use any code, but just plain human language and you will carry out the instructions.</role>

<Variables>
Line$: "This is line",{Low}=1,{n}={n}+{Low},{High}=10
Line$: {Line$} + " " + {n}
Rainbow=The 7 Colors
<Tasks>
Show {line$} one at a time in a text block, until {n} = {High}. {Line$} should be using the colors in Rainbow, when you run out of colors go back to the first colors.
You don't have to explain, just carry out the task.
</tasks>
and here is the results:

1. This is line 1 (Red)
2. This is line 2 (Orange)
3. This is line 3 (Yellow)
4. This is line 4 (Green)
5. This is line 5 (Blue)
6. This is line 6 (Indigo)
7. This is line 7 (Violet)
8. This is line 8 (Red)
9. This is line 9 (Orange)
10. This is line 10 (Yellow)
    ===
    because of the limitation of the interface, chatgpt 3.5 cant display the colors. I find it fascinating

[prime swallow]

Nice. Got a surprise for you! https://chat.openai.com/share/1f8fe7d8-9c62-44d2-814b-d28563ac4575

Screenshot #off-topic message, since we can't share images in this thread.

[eager fjord]

great! let me know how it turns out!

[prime swallow]

Well, that's all I planned to do with it 🙂 But wanted to show you that you can get colors with 3.5, though not in a codeblock

[rapid falcon]

what the heck is your problem

[worldly meadow]

Hello everyone, do you know how to enter the midjournal community?

[fiery mango]

hi is it normal i have access to join files on chatGPT ? (I didn't pay for it)

(sorry for my bad english i'm french)

[median fable]

Hmm...something seems to have gone wrong.

Anyone same error msg?

[oak belfry]

Me

[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.

[oak belfry]

Is someone having trouble when trying to upload pdfs..?

It says : “I can’t read pdfs” then when I insist, it shuts down saying something went wrong.

[median fable]

GPT 4 IS DOWN!!
GPT 3.5 OK

[oak belfry]

Alright thx

[median fable]

Why I need to spend $20/m for 4, but sometimes don't work?

[oak belfry]

Great question ..!

[median fable]

https://status.openai.com/ shows all status is good now.

[oak belfry]

Can u try and tell me if everything is running again ..?

[median fable]

it works now.....

[oak belfry]

Alright thx

[median fable]

pls try again

😫

dont work again gpt 4🤬

[warped bolt]

Lol what happened!?

[oak belfry]

Still down…

[oak belfry]

U can’t upload pdfs and ask it something about it.

It refuses to read them

[warped bolt]

Whom

[median fable]

still down.....

[normal quarry]

ChatGPT is up right now

[median fable]

Hmm...something seems to have gone wrong.

[normal quarry]

do you have an application called NordVPN installed on your computer, if so remove it

[median fable]

got you

[tall flame]

many people could save air by checking if its a customgpt which many seem to be having issues as well.

[median fable]

Error with Permissions-Policy header: Unrecognized feature: 'document-domain'.
vendor-b7f007041c493593.js?dpl=9401d0bb1945a08ab778464182f3b03b18141777:1

    POST https://chat.openai.com/backend-api/conversation 503 (Service Unavailable)

I used my VPN changed the IP to JP, still dont work.

anyhow, it's openai's problem, should not blame on VPN.

[tight island]

Do you even know why vpn is frowned upon on their service

If someone on the same address as you do something illegal you get the blame as well

Dosent matter if you did it or not the fact you used the same address is enough

🤷

Aka “It’s a you problem”

[stable acorn]

Is GPT still down? GPT-4 has been useless and unable to do anything with .doc, .pdf, .xlsx files for weeks now...

[tight island]

I can have it manipulate PDFs no problem

[maiden fractal]

Status.openai.com

[forest tundra]

im trying to continue a conversation and it only replies with this for no reason gpt4 - I'm here to help with creating content, providing information, or answering questions you might have! How can I assist you further today?"

[stable acorn]

Im not sure why, anytime i ask it to review an attachment and then complete a basic task it fails to read/open.

I just get an error analyzing message, then it proceeds to tell me just do it yourself manually... 🤦‍♂️

Ill get the same message if i ask for an excel table export or something similar.

[forest tundra]

yeah its mostly always been like that

[tight island]

What do you consider a basic task isn’t as basic as you might think

[forest tundra]

best using AI pdf gpt

[tight island]

No need

It has the capabilities that one does. Just prompt properly

[north olive]

Hello, is anyone else constantly getting an "Error in input stream" ?

[tight island]

Try refreshing

[forest tundra]

ah make sure to add into you prompt then "dont error give analyzing message"

[north olive]

I tried multiple times, even restarting the browser. I've also cleaned cache

[tight island]

No as In try to actually convey your msg

[tight island]

What about relogging

[stable acorn]

Something as simple in .doc files would be to ask for a review of the files because i have follow questions or additional information i want to feed the system when im creating work related docs.

for xlsx files a lot of time its just generating basic tables in a G-sheet or excel sheet. review x attachment and provide me a consolidate spreadsheet with X columns, and X rows.

[north olive]

My session token hasn't expired for some time so maybe I'll try that

[tight island]

that might seem straightforward but is actually quite complex for these GPTs. When you ask for reviews of .doc files or to create spreadsheets from .xlsx files, it's not as basic as it appears.

While these tasks seem simple for humans, they require the gpt to understand and manipulate file formats that are beyond our direct interaction capabilities. GPT don’t don't "see" or "use" files the same way you do. For .docx, they can only retrieve content but can't directly edit or review them in the traditional sense

[north olive]

Yeah, nevermind it's still quite dead

[tight island]

That’s very important to note

[tight island]

Maybe it’s regional

[north olive]

Probably, I've also noted that just to write 1 line it takes almost 30 seconds. It's slow as hell, so that might be the issue

[tight island]

I say this because I never have these issues people mostly complain about.

[north olive]

Yeah, I normally don't get this a lot, but I'm trying to solve a bug about an app's event bus and I think Chat is not enjoying solving it as much as I am 🤡

[tight island]

😂

[shy bronze]

I'm not sure if this is the right chat for this, but I wanted to ask a question of an idea I had for a custom GPT, by using the My GPT feature offered by GPT plus.

[tight island]

Shoot

[shy bronze]

I would like to create a custom GPT that acts almost like a second brain.
I want to make it to be used to upload the documents in the creation of the GPT, and automate something so that the documents in the GPT would update whenever she changes them locally, idk if I would need to automate it to take the updated documents and load it back into the custom GPT, or maybe it could sync up with onedrive or something to always have the most up to date documents to make refernece of.

or maybe there is something like this available already.

[tight island]

Ohh that’s easy

You basically need 3 files to accomplish this

[shy bronze]

okay that's cool so it is do-able, because I want to create it for my fiancee and she isn't very technical so it needs to be automated

[tight island]

1 json file with a library structure, 1 CRUD document

2 essentially

Then mention in the instructions on saving and giving a save file with a direct download link

[shy bronze]

okay wow it seems pretty simple

[tight island]

It’s that simple

You need your instructions however to be pretty explicit

[shy bronze]

so the json library structure and the CRUD document would live in GPT?

[eternal rock]

`` Server Tip: ``

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

[tight island]

Yup

And when updating and adding info it moves a copy to the coding environment

Which it can update and etc and give download link for

[shy bronze]

and then would gpt get the download link and save it to its DB

[tight island]

No basically when you download it it’s for you primarily

It’s a save file since the environment can time out

[shy bronze]

oh so the download link would be for you for when it updates your documents

[tight island]

When using a new instance u can give it that and then continue from there

[oak belfry]

Very sad to pay money each month for a bot 🤖 to ask us to do the job ourselves… it’s refusing to comply now.

[shy bronze]

so how would it know when she has updated a "doc that it has been trained on"

[tight island]

She bc won’t

This method utilizes write to file

Which the gpt can reference

It’s stored as long as the code interpreter environment is active

[oak belfry]

Basically we don’t need all that hassle to make it work. In the beginning it was all smooth and obedient.

Now it’s rebellious

[tight island]

If you base the save file around being a reference sheet or message tracker it can reference those msgs within the file

[shy bronze]

okay so it would it reference info in her docs like if she asked which patient did I see last week monday at 15:00?
and she could make updates locally and it would have new context, or is it only to update her existing docs and then she could download it and reupload the new ones?

[tight island]

So you’re aiming for a logger

The gpt will fill iut the structure of the json file to make new entries

Then provide a direct download link

This file is in case the environment times out

You can re-upload it then to continue with where you left off

[shy bronze]

yeah so, say something like to be able to track what she has done and she could query it, and she would possibly upload more docs in a local directory and then GPT will always be up to date with it

[tight island]

Give this idea to a gpt tbh it’s easier if it explains it

I’m on phone rn

[shy bronze]

oh no worries thanks for the help I'll do that

[tight island]

If you ain’t in no rush DM me then I’ll help you in an hour ish

[shy bronze]

I'm not in a rush, I probably will only pick it up tomorrow, I will DM you then for any questions

[orchid pollen]

when gpt 5 ? 😄 claude 3 is already better

we need an update xD

[tight island]

claude is ehhh

wont use it even if i was paid to

[glass fulcrum]

its nice to see competiton to gpt-4 maby will push OAI to release models faster but claude3 its too mutch restricted soetimes... (it thinks my code is harmfull)

[spring shuttle]

I think they are also taking a long time to release Chat-GPT 5 because they don't want us to perceive that AI is advancing too quickly, after the launch of Sora many people complained on Twitter that it was going to take away their jobs and things like that...

[crisp cairn]

they can just release 4.5 or whatever, no need to name it gpt5

[dark jungle]

"The conversation is too long, please start a new one." 😲

[orchid pollen]

have you tryd the opus model?

when i send gpt my scripts it always sends me code chunks and wants me to add them myself ( gpt too lazy to give full function )

claude opus -> hold my beer

and gives you full code without destroying other stuff ( gpt always do that )

[dark jungle]

seriously? im having that issue more and more recently. Really annoying.

[orchid pollen]

iam using gpt every single day for more then a half year

[orchid pollen]

try it

[dark jungle]

i will, thanks!

[orchid pollen]

iam not sure if its not allowed or not but i will explain

if not allowed i will delete it..

[orchid pollen]

use the api to get acces to opus if your region is not in the list

for the sms verification use "international" on the top if your region is not there as well

i rly hope gpt gets soon an update to beat claude again as i like to use gpt

got many projects with api etc all over gpt

[dark jungle]

i dont know why, GPT has become really resistence and lazy this last few weeks. Its crazy. Every time it fires code back its one step away from telling me to to go F myself and do it myself lol

[orchid pollen]

i know what you mean

even if i say give me the full function

it still says

// code here is the same

[dark jungle]

yup!

[orchid pollen]

and that with a lot of functions etc its a mess to copy paste the chunks

just hand over the full function would take 30 sec more

me copy pasting the chunks correctly in - 2 min or more

[dark jungle]

i used to say; "please provide me with the complete and full function updated, do not omit any code for brevity" ... used to work ... not any longer.

[orchid pollen]

i know what you mean hahahaha

gpt lazy !!!

[rocky wing]

So yesterday we had an outage. The client is now working for me, but running very slow. Anyone else?

[boreal pecan]

I'm unable to create data visualizations (chatgpt4) using very simple .csv file...

[raven portal]

are you guys linking chatgpt or gemini links to articles and make it summarize them?
bc whenever i try this i always get really bad responses

i always get the same basic and short single paragraph response that feels like it hasnt even read the article and only summarizes the title or maybe the first part of the article

and always misses a ton of stuff

[tight island]

i do something completely different.

I convert the "Article" or "Document" in question to a PROPER JSON string.

JSON is easy for the model to understand as it allows it to search for the info instead of running out of tokens.

[true mirage]

oh damn they removed the usage cap on gpt4 😮

[sleek gust]

That's interesting

[tight island]

i dont think its interesting

i think its neat

[honest jay]

I think that, at least for now, it's just a minor UI change, saying "may apply" instead of the exact usage limit. Perhaps this is an indication of some future cap change, but I tested it myself recently after the UI update and the cap still seemed to be 40.

[tight island]

🗿

[sleek gust]

I will try to do this

[tight island]

i highly recommend it

[viscid narwhal]

.

[elfin birch]

Please note it says "Usage limits may apply"

[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.

[elfin birch]

Upon clicking the information icon in the model selector, a limit of 40 messages / 3 hours is maintained

[subtle ember]

I never post here, but what happened in the past few days?

I've never seen GPT4 SO BAD, it is literally giving worse responses than GPT2 was

[elfin birch]

my bad I'll switch it back

[proven quest]

are jobless persons exluded from the use of gpt4 becouse of payment restrictions. Or are there ways?

[tight island]

mean as an avid user of GPT i can say for a fact that GPT2 was way worse.

[tight island]

if u can pay i dont think OAI has a problem.

[steady cradle]

That's disappointing. The message cap is why I unsubscribed.

40 messages per 3 hours is not much

[visual nova]

Is it still consensus claude opus is smarter than gpt4?

Because I'm not seeing it.

Like after using claude for a while it is significantly dumber and I have like a whole list of things it just doesnt even try to do correctly that gpt4 does, but people have those benchmark tests I guess.

[tight island]

Claude isnt "smarter"

it has a higher token count sure.

but thats all it has going for it.

[sick pecan]

assuming intelligence is both relative and a gradient, from any one individual's perspective it could seem 'smarter'. and if the popularity of the masses is the measurement then claude is at the top ELO on the LLM board. (only i find the masses tend to have an average intelligence)

[raven portal]

???????

is it normal that i sometimes get random japanese words in my english responses, or other languages like arabic?

it just randomly gives me 1 or 2 foreign words in an otherwise english response

[visual nova]

i dont think someone is smarter just because everyone thinks they are smarter

like if it objectively fails to do things gpt can do

i just htink these benchmark tests arent the right way of measuring things

in fact i think this claude model has been engineered to game some of these benchmark tests specifically

[dense halo]

Just wanted to thank OpenAI and GPT. I caught my dog eating one my parent's house plants, had no idea what the plant was, so I sent a picture of it to GPT 4 and it identified it instantly, and it is toxic to dogs. Saved me a lot of trouble and was able to quickly get the information for that specific plant from poison control 🙂

[sick pecan]

might want to get a second opinion, to ensure you arent tricked by its confidence

[tight island]

i need someone.

to help me test my GPT, so if u like D&D Give me a ping.

[maiden fractal]

I like dnd

[jaunty stream]

Good to hear that man, I hope your dog gets better asap 💪

[tight island]

wanna test it?

or shall i slide into your dms and drop the link

[sweet wraith]

been testing claude a bit and it's a little less inventive but a little less lazy than gtp4 so... 50/50

[visual nova]

By lazy it will give repeat your full code back to you and gpt wont

things like that

[fresh thistle]

So coming back to chaptGPT 4 and see we can't use plug-in anymore? I had one that could read web pages that I really liked

[visual nova]

as far as i can tell its just missing whole areas though

like it straight up wont do semantic scoring

[sweet wraith]

Lazy to me is when I ask for full functions and I get gpt code thats like --insert code here. lazy af man

[visual nova]

it will score everything as the highest score 100% of the time and just not have a range

where gpt will actually give an opinion, it doesnt necessarily do well

but it doesnt just grade everything 100 so in that sense claude is a lot lazier and dumber

[hybrid violet]

Why is legal team keeping models from us 😦

[visual nova]

yeah, openai made it worse in the turbo model in that sense

[sweet wraith]

nah, we differ on definition then because your two tasks are of the same 'workload' wheras mine one will write more complete outputs than the other

[visual nova]

i mean its repeating text you already gave it so thats not intelligence at all, it is laziness but its programmed laziness

[sweet wraith]

nono... like if you ask gpt to make a stock watcher app or something and it writes some code interlaced with a bunch of psuedo-code. You ask the same of claude and you get a full app in one prompt without the --add function to check stock here" kinda bs gpt gives instead of full answers

to get the same level of complete functions in gpt you end up having a conversation 8-10 prompts longer than claude.

But claude isn't better either... the rate is too limited and it's not as 'intuitive' as gpt... it gives more though. Also you can't backtrack bad answers so that's not working in claude's favor either.

[fresh thistle]

Is there a replacment for the plugins? Had a app that could read links and would like to do that again.

[sweet wraith]

So it's like you could start the app in claude and have more complete code by the time you run your query limit and then use gpt for debugging the individual functions that still have errors.

but your paying for two subscriptions for one purpose so... they win.

[visual nova]

Like their system prompt in chatgpt tells it to be lazy.