gpt-models

[remote tiger]

guys is it true gpt4 on the mobile app is weaker?

[lament saddle]

I feel gpt 4 is much faster now

[wet drift]

As far as I know right now the App doesnt have access to plugins

[wet drift]

Yeah I was noticing that too

[lament saddle]

anyone could get gpt-4 web browse plugin to find references on multiple sources instead of just 1 source ?

[lavish venture]

finally able to use GPT-4 again and was able to successfully switch my session that switched to gpt-3.5 back 4.0

[austere parrot]

?how to use

which room?

[valid wing]

is the reaon for the limit of plugins on gpt 4 is so it doesnt overload its servers?

[lament saddle]

and i wonder if you could tell chatgpt to mix the web results with its own trained data ? chatgpt refused

[inner trail]

Whichever room you like

[minor beacon]

Haven't tried it yet but have seen where you can backdoor onto already existing conversations - just can't start new convos

[sudden mural]

will we eventually also get access to the code interpreter or how does that work without a waitlist?

[boreal epoch]

BingChat does that by default. It's GPT4 based.

[minor beacon]

There has been no official statement that I know of - my guess is that it will roll out just like web browsing and plugins. A few will get it and then more will get it, and then they will announce an implementation window and all plus users will get it

[grand wyvern]

@minor beacon

[sudden mural]

ic ic thanks 🙂

[minor beacon]

?

[mental dawn]

can you dm me

i have been researching about it? can we connect? dm?

[cunning nest]

I think this idea has some merit to it been thinking about it for some time, there also is an expert advisor on metaTrader built using gpt

[shrewd citrus]

While it would definitely be able to relay some fundamentals it wont beat the people who are getting the highest returns US government reps. Thats because they have inside info. So best strategy should be to inquire how to hijack their trading strategy the best given they have info ChatGPT wont have.

[frozen veldt]

I built an open source python script to do some of that https://discord.com/channels/974519864045756446/1110001553282179102

it needs tuning by people who understand finance, and improvement by experts in python. I've taken it as far as I can with gpt-4 code interpreter

integrating gpt-4 calls when I can access (or because you can access) should definitely improve it's ability to conduct analysis

[shrewd citrus]

Unfortunately no amount of scraping can pull the data that would give it that insider trading knowledge that US reps use to financially profit. Its a known thing, and many people just mirror certain reps trades to beat the market. Though at a lesser return given the timing and such

[frozen veldt]

there are sites that report insider trading

it wont solve the inequity entirely.. but it's a start

[shrewd citrus]

RIght, but thats after the fact, and the market as responded. There is a reason that they get the highest returns. Its because they were first to act before the rest of the world knew.

[frozen veldt]

of course, that's finance. Can't help that some have the inside track. Best I could do was generate a sentiment analysis tool (replicating successful academic work) in order to hopefully inform people better as otherwise they would have to go pay for tools that ultimately guide them towards specific purchases. I'm even leery of the plugins now due to their pre-prompting. A native interaction with the core instance and having it run real sentiment analysis is as far as you can get from external influence. Add a bunch of math in there to make sure it's not just about the sentiment.. and it starts to look like an actual model. Now it needs tuning, I'm investing my own money to track performance of its predictions and analysis. Others could target different segements of the market to do concurrent work.

[unique urchin]

Hello, I new here

[fickle gale]

If I become insider trader and everyone copy me, then very hard for me to fail 😯

[unique urchin]

can you you clarify what are you specifically discussing?

[frozen veldt]

the initial topic was a gpt-4 trading bot

[shrewd citrus]

Using ChatGPT to create financial investment strategy. The topic is that the best strategy would be unavailable because the highest returns go to those doing insider trading while working for the government, but that you could get it to hijack the strategy with monitoring, and specific trading types.

[frozen veldt]

Nowicki monitoring inputs.. how often and how fast? polygon does have a windsock

In theory even now you could have auto gpt take my database and then monitor and deal with a portfolio using an investment API

or any other qualifying inputs

couldn't possibly go wrong.....
I prefer a multiple layer approach:
Pull data and have sentiment analysis conducted by vader and gpt, add a bunch of math
Run that database by gpt-4 code interpreter for extrapolation
Take that data, add my own DD, then invest

[fickle gale]

There are just so many trader plugins, why bother with da above?

[frozen veldt]

because those plugins have been revealed to have hidden prompts that can direct you to stock purchases rather than generating real DD

[fickle gale]

If I had to pick one category of plugins that is the highest, it is those trading ones.

[frozen veldt]

I don't trust any of them

[shrewd citrus]

I would say that the twitter account that monitors Nancy Pelosi's trades is a good one to start. They have an app too so maybe an API to pull from.

[frozen veldt]

what I built replicates research from university of florida, their gpt instance generated 500% returns on news and financial data

@shrewd citrus I'll take a look at that, additional inputs would be very helpful to the model

[civic sparrow]

is that actually working though and are the stocks crazy high priced

[frozen veldt]

I dont know they kept the rest of their results and their model proprietary

so I built it open source as best I could

it needs tuning, prompt engineering, input changes, output changes.. but it does work

[mint thistle]

Hello. Recently, my ChatGPT "mode" option disappeared... anyone have any info?

[cunning nest]

I think you have done your best. The thing we should be discussing right now should be what we can improve on right now at this point in time. The argument that politicians are involved in insider trading is true but we should be focusing on the meagre resources available to us and potentially improving them

[shrewd citrus]

Have you found elements of the prompts it does not like to respond to and essentially tells you no?

[frozen veldt]

No, using the research prompt it has been reasonably compliant. Its not asked to actually make an investment call, its asked to assess the sentiment of the article

same with gpt-4, I don't ask it to make stock picks, I ask it to determine from the data which stocks have the highest percieved upside from the data

[mint thistle]

Did ChatGPT get rid of the $20 month thing?

[frozen veldt]

that way you avoid the whole as an AI language model i cannot blah blah blah

[cunning nest]

No

[unique urchin]

this is actually interesting, but anyway trying to reach the best strategy would be okay as whether following a good or not highly effective strategy, those insiders still take the highest return

[mint thistle]

I can't even seem to access my account... as soon as I login, it goes into "ChatGPT Plus" now, and I have no way to see my account details.. no way to determine what my subscription is/

[frozen veldt]

I completely agree. I'm not a python coder, this took many revisions with gpt-4 code interpreter. A python expert might make serious headway on improvements in short order.

i'm also not a financial expert, someone or people who are could seriously improve the math

[fickle gale]

What value would a python expert add that u can't ask ChatGPT to generate?

[dark mauve]

Paying cutomers test the errors and then "You've reached the current usage cap for GPT-4. "

[mint thistle]

I don't see an option for ChatGTPT 4 mode anymore.

I have not even USED chatGPT 4 in like days

[frozen veldt]

It's that we're now using such a large code that the code interpreter cannot keep it in the contextual buffer. It is beyond it's ability to keep straight resulting in fidelity issues during code rebuilds

for example, it pruns previously inserted error correction logic because its making the function up again from scratch due to the context limit

[mint thistle]

(so maybe chatGPT 4 is just high usage now? and they turned it off )

[frozen veldt]

so yeah, it really has gone as far as I can go with myself and gpt-4 code interpreter, from here it takes actual humans

[unique urchin]

I am working on a ML model testing the impact of the independent variables (Cap_Form, EXP, IMP, and M2/GDP) on the dependent variable (GDP Growth), so is there here anyone who has experience in working on that topic before?

[frozen veldt]

sounds interesting, definitely ask the code interpreter if it has any math to suggest there

for my work it has been great, just gave me a crash course on pressure mechanics in a product contextualized way, super informative

[fickle gale]

Can u post the code again?

[unique urchin]

how can I do so?

[frozen veldt]

https://discord.com/channels/974519864045756446/1110001553282179102

[frozen veldt]

just seriously ask it if there is any math that can do that, using that sentence above. Dont use 3.5...

[unique urchin]

okay I will, thanks for you suggestion

[lime ocean]

Will we be able to use plugins via API? For example, if I want to make an API call to GPT-4 with a plugin, will I be able to?

[unreal swallow]

with the new keep generating feature for when responses get cut off short, does pressing keep generating count as another instance of you using one of your 25 prompts per 3 hours?

[supple violet]

yes

[lament isle]

Where can I find an explanation on every plugin available ?

[supple violet]

has not been confirmed anywhere if and when, only a side mention as a possibility for some plugins

[civic sparrow]

you can ask gpt to explain the plugins so activate them and say explain the features of this plugin and it will tell you

[lament isle]

Thanks very much

[wraith ginkgo]

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 @open crescent immediately.

[jade shore]

I have read that all chatgpt plus users should have access to real time data, but it this seems not the case, can anyone clarify?

[quartz haven]

Every day i wake up hoping code interpreter and image processing will be released

[quartz haven]

enable browsing mode under gpt-4 model tab

[jade shore]

awesome thanks, lol thought it woulda been automaticly enabled

[tulip zenith]

anyone else having issues with getting GPT4 to read google documents through plugins?

[unreal swallow]

how do we get plugins?

also my chat isnt laoding

[knotty grotto]

Gpt 4

[unreal swallow]

I have gpt 4 on chatgpt, how do I gain access to these plugins?/

[knotty grotto]

In the setting make sure they are enabled

Start a new chatgpt 4, go in plug in and plugin store

Hi I suggest to use the #introductions channel

[woeful yarrow]

noted @knotty grotto
thank you for feedback. Will do the needful

[manic sleet]

Have we heard from the team about the fact that gtp 4 has lost half of its intelligence since 3 days? I'm far from being the only one to have noticed it.

It is now unusable for my programming, giving me answers that make me waste more time than anything else?

And all this remains at the same price? hello devs can we have some info ??????????

[woeful yarrow]

can you guide me which channel would be best suited to look for like minded people passionate about chatgpt who are willing to work on live projects maybe a freelancer or full time
Would be thankful if you can guide me

[frozen veldt]

apc are you a python coder?

or are you looking for coders

I have a whole roadmap of side projects to tackle, if I ever get ahead of that I might reach out

[amber wigeon]

is anyone

online

[terse ibex]

what is MI

[agile vortex]

machine learning

[shrewd citrus]

"Will do the needful" Eh? Did you happen to work at bluehost for a period of time?

[wind urchin]

idk, guys can you increase the width of the response, I feel its very narrow, and wasting screen space on both the sides

[regal inlet]

Me too

For me web pilot is better, at the moment

[wintry tangle]

I have been using gpt4 to help me with coding for a few months now. It has been extremely helpful with writing python code. However, today I feel like it has become 'dumber'. Previously, I could paste in my code, ask it to debug and it would understand what I am asking. Now it is forgetting that I have already pasted in my code and is giving me answers like "I would love to help, but without seeing your full code I wont be able to help much"

Im finding generally (with coding) it is not as good at helping as it was a week ago. Is anyone else having the same experience ?

[wraith ginkgo]

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 @open crescent immediately.

[manic sleet]

Yes, many of us have noticed. For coding he forgets the messages we sent him above, gives bad answers that waste time etc... this has been going on for 3/4 days, I am very disappointed.
No word from the openAI team. They downgroad their tool without telling anyone thinking that no one will notice, and charge the same price. Very sad

[wintry tangle]

Yes, I couldnt agree more. I am new in here, but is there staff from OpenAi that we can tag? Or would it just be reported as a bug? They need to get it back where it was.

[wraith ginkgo]

`` Server Tip: ``

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

[tulip zenith]

glad to know I'm' not the only one

[manic sleet]

nah i think they dont care. Not a single word since then despite several twitter threads about the problem

[wraith ginkgo]

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.

[frozen veldt]

I just finished a build of my project with gpt-4, the context limit is really the issue I suspect. once the file or conversation is too large it starts to make up working code but it drops little details like error prevention code

[wintry tangle]

So if my code is quite long, do you think this may be causing the issue?

[frozen veldt]

Exceedingly likely based on my experience

it was fine at 70 lines, at 200 lines.. but at 350 or so it started to hit problems

[north pewter]

you guys having these issue with plugins or default? it feels like there's something wonky about the way the context works in plugins, it hallucinates way more

[frozen veldt]

getting the last couple of builds in was really tough for me since I dont know how to code

[wintry tangle]

Ok, good to know, thank you very much for the intel

[frozen veldt]

I've learned a lot about python but not enough to even clean up the code well at this point

[wintry tangle]

Same

[frozen veldt]

so it helped me develop code I can't really understand, based on the workflow I wanted

helpful and mystifying at the same time

try a clean instance and upload the .py fresh, then ask your questions and keep it brief

[wintry tangle]

How did you work around it? Break the code down? or just power through with a fairly unhelpful AI assistant that needed constant reminding of the goal etc?

[frozen veldt]

some prompting helps

[wintry tangle]

How do you upload the py file? I have just been pasting it in

[frozen veldt]

what version are you using

[wintry tangle]

gpt4

[frozen veldt]

I did all my work in gpt-4 code interpreter

[wintry tangle]

I dont think I have that yet

[frozen veldt]

CI has a widget on the left side for uploading files

get on that waitlist then

[wintry tangle]

How do I do that?

Sorry, didnt know this was a thing

I have access to the plugins, but didnt realise there were more we could get on the wait list for

[frozen veldt]

all the fancy stuff has waitlists

[wintry tangle]

haha

[frozen veldt]

the one I really want to see a list for is code interpreter 32k model with live simulation of code ability

[wintry tangle]

How do I join the wait list?

https://openai.com/waitlist/plugins

[frozen veldt]

its under your account stuff I think, been a few weeks

[wintry tangle]

I have found this, but I already have access to loads of the plugins from being a plus user

[frozen veldt]

the plugins have been a mixed bag

i really don't like the investment ones that add hidden prompt information

[wintry tangle]

Is there a specific link to get on the wait list for the code interpreter ?

[frozen veldt]

I don't know sorry

[wintry tangle]

No worries, ive messaged OpenAI requesting it, see what they say

[lilac gull]

Which one is this?!

[wintry tangle]

I was trying to find a link to get access to this code interpreter plugin

I thought there might have been a wait list

[lilac gull]

Ahh ok.

Have you got a link/video for it?

Haven't seen it yet @wintry tangle

[valid sky]

@lilac gull

[lilac gull]

What's up

[valid sky]

What’s up

[lilac gull]

¯_(ツ)_/¯

[wintry tangle]

I haven't seen much on it either to be honest, @frozen veldt mentioned it

[valid sky]

¯_(ツ)_/¯

[frozen veldt]

https://openai.com/waitlist/gpt-4-api

[lilac gull]

Has anyone successfully submitted an eval?

If so, anyone willing to chat about it on a call?

[valid sky]

Whut

I was kick because I speak French

[lilac gull]

https://github.com/openai/evals

[daring pine]

yo, so let's say i have a piece of text; and I want to see how many tokens it is in total. Is there a way to calculate that?

[lilac gull]

https://platform.openai.com/tokenizer

@daring pine ^

[daring pine]

Dope! thank you, that is perfect. exactly what i needed

[signal eagle]

is the 25 limit not enforced? doesn't look like it is

[daring pine]

you might just not be hitting it. It's actually more than it sounds like, you'd have to ask it a question something like once every 7 minutes for 3 hours straight to hit the limit, and even then, you'd just barely hit it

I've never hit it before, even when using it for programming. But, if you do hit it, if you have the API you can keep asking it questions that way

[frozen veldt]

I can't call 4 by api yet unfortunately so getting my project done with code interpreter was done in crash sessions every 3 hours over weekends

jam on that for an hour, hit the limit, go play elden ring, do it again, go do something else.. round and round this weekend

[lilac gull]

Anyone up for a call?

[signal eagle]

nah I tracked the time and messages. I had over 34 messages in under 3 hours

[frozen veldt]

wild, that would be great lol

[daring pine]

WHOA. I just realized why it kept saying my "message was too long" when I pasted code. It was counting every space as a token!!! Replace spaces with tabs and you can paste in WAY more code. Thanks for showing me that tokenizer thing, i never would have guessed

[lilac gull]

No worries at all.

[lilac gull]

So no one at all??!!

[daring pine]

I'm listening to some music and writing some code, otherwise i would

[frozen veldt]

kinda hard to get people to just jump on a call

I'm fitting in brief moments between meetings

[daring pine]

meetings, are you a software engineer or something?

[frozen veldt]

developer of cpg goods

so meetings to ensure alignment internally then execution meetings with factory partners

how about you?

[daring pine]

ha, unemployed 😎

[frozen veldt]

congratulations or my sympathies, depending 😉

[daring pine]

I'm in university. I didn't get an internship this summer. It's all good though

i waited too late to apply

[gritty garnet]

Ew, gross, Clark Griswold.

[frozen veldt]

but if you have time to burn and know how to code and want to help I'll plug my own thing: https://discord.com/channels/974519864045756446/1110001553282179102

I've taken it as far as I can with gpt-4 code interpreter

[signal eagle]

anyway, before going, one thing that is super helpful to me is giving gpt 'personalities' and still conduct my work normally. but now it's way more fun and engaging! might be helpful for some since normal gpt is more boring than ads

[daring pine]

I'm taking some summer classes instead to get my GPA up and get some extra credits, and i'm working on developing some projects to improve my resume so i can start applying for summer 2024 internships this fall. I'm developing a Minecraft clone from scratch using C++/OpenGL.

Learning a ton about 3D graphics programming, multi-threading, data structures and algorithms, and all sorts of stuff. Definitely the biggest project i've ever done

[frozen veldt]

sweet have fun building that 😄

if I ever get to anything that large I'll have to hire a developer

I'm not putting my own learning of the code in front of a project, the thing I built was a side effort to help my own investing

[steel turtle]

Have you seen #1104422061318877285 ?

[hoary elk]

hello, im looking for a website that acts as a gpt-4 middleman. you collect coins in their website and with those coins you can have limited access to the gpt4 api through the site. not sure if im in the right place or even if what im asking for is even within openai tos but worth the shot

i stumbled across something like that some time ago but cant seem to find it anywhere anymore

[daring pine]

Haven't seen that post in particular, but i've actually done something very similar with it before, asking it for a list of topics to learn next to fill in gaps in my knowledge for certain subjects

[steel turtle]

I've seen good quality from that one in particular. It's not mine. Just thought you might like it, given your love of learning ❤️

[daring pine]

Will check it out for sure. Maybe i should look through that channel a bit and become a bit better at asking it questions. Right now, i just kinda brute force it; i can usually get an answer out of it but sometimes idk how to ask it what i wanna know lmao

[merry ferry]

Can you only do 25 every 3 hour per conversation or just per account?

[opaque hazel]

Hello, can anyone help me out how to provide chat gpt4 with images ? which plugin to use ?

[steel turtle]

Definitely put the prompt in #prompt-engineering if you're looking to see how to improve it. AND the #1019652163640762428 is so cool.

I wrote a lazy "prompt-fixer" tool in #1107132288279527485 in the library, BUT it's best to feed it a super lazy prompt (10 words or less)

[fierce oxide]

Hello. What's the cap on GPT-4 for ChatGPT Plus subscribers?

[lilac gull]

Where are you based out of? UK?

If so, might be able to help you with an internship/some work experience.

[drowsy ibex]

Is it possible to use plug-ins not in GPT4 because when I use it in GPT4 it effects my 25 messages

[daring pine]

Ah, I live in the United States

[wraith ginkgo]

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 @open crescent immediately.

[steel turtle]

no, unfortunately.

[ancient yarrow]

25 messages/3 hours

[outer flax]

Do you knwo where the waitlist for alpha is? I've searched and can't find it

[static isle]

It used to be here
https://openai.com/waitlist/plugins
But the option to apply for the code interpreter was disabled after the beta rollout of plugins and browsing
Apologies for getting your hopes up 😢

[craggy relic]

Is chatgpt4 plugins down? I keep getting error generating response?

[hardy willow]

G

[forest perch]

it's here now as well

[tribal wave]

Hlw.,.... Can anyone help me in the coding test?

[static isle]

I don't think that's true. I am currently a Plus user and there is no option to select Code Interpreter in Beta Features

[outer flax]

oh ok, glad to know it wasn't me lol
I was wondering why I couldn't find it.
I can't find it in the Beta apps either so I'm guessing there's no way to use it for now?

[static isle]

Still gotta wait, nevertheless fortunate to be in a time of such growth

[outer flax]

🔥

[coral wave]

Hy

[floral grotto]

Has anyone had this issue:

where it will continue generating the same thing over and over and wasting the usage cap.

ill ask it one thing and it will just continue generating what it already told me, not click on continue or anything

[frozen veldt]

I"ve seen that in the code interpreter, it will literally spit out the same block repeatedly until told to stop sometimes

[wraith ginkgo]

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 @open crescent immediately.

[floral grotto]

Yeah , i havent got that yet

Only browsing and plugins

[unique hare]

Does anyone else feel like gpt-4 got nerfed pretty hard in the last patch? Also the browsing beta just does nonsense. I give it clear instructions and it can not follow them.

[dreamy rover]

nerfs, wonder what causes them. I remember when GPT 3.5 was nerfed so hard that it was making grammar mistakes lol thankfully that was fixed

[pale agate]

I want to extract transcript from youtube video using python or php. any idea how to code this? I am also thinkking of calling open ai api and use plugins like voxscript in a python program. Not sure if this possible. Once the transcript is extracted, I want to pass it to chatgpt to get summary. Any help will be appreciated.

[floral grotto]

This sounds like a fun idea

Would love to talk more on that idea

[pale agate]

sure. today, I am using chatgpt plus. there are plugins avaialble there like voxscript. if I give youtube url on gpt, it automatically calls the plugin and it gets me the summary. Before the plugin was added, I used to go yotuube and run webchat on a video that is open, which used to get me the transcript. I used the transcript to post it on chatgpt and get summary. Now, I am looking for a way to do the same outside of chatgpt either thru openai api or any other libraries that are available. hope I provided enough details.

[floral grotto]

Right right, looks like u would also using a youtube api as well?

Maybe, could be wrong as i don't work with alot of apis

mostly working on front end web dev

UI/UX type work.

[pale agate]

yup, I want to integrate that with a website

[lapis bane]

wait a minute

does gpt4 have a limit on the ios app?? can anyone confirm whether it does or does not

i do appreciate the help in advance xoxo

[floral grotto]

It does, as it also has one on desktop.

Would only make sense to.

unsure of when/if they're going to up the cap every 3 hrs

[wraith ginkgo]

`` Server Tip: ``

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

[iron fable]

help

[floral grotto]

Interesting, would love to connect more on you with this.

[wraith ginkgo]

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.

[austere seal]

help

[cunning coral]

Has anyone else found that GPT-4 is much worse for rewriting texts than gpt-3.5? I bought the pro today hoping for better

[floral grotto]

Well a few mins working with something i have achieved this

[grim ore]

Is there any way to get past the cap for GPT 4? It's so frustrating to pay $20/mo and have a cap on creations.

[floral grotto]

They would have to increase it themselves

[bright oyster]

I'm sure they might increase it soon.

[wraith ginkgo]

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 @open crescent immediately.

[icy bronze]

my subscription failed to renew and it's declining all my cards. anyone got any fixes?

[floral grotto]

Try clearing your cache for the website. Has worked for me in the past

[ornate bramble]

Anyone else noticing the huge drop in quality from gpt4 in the past couple of days? It seems like it got all the way back to gpt3...

[frozen mist]

A lot of people have mentioned this.
I think there might be two reasons for this. First, GPT4 might have downsized its model to save costs, and secondly, OpenAI might have included some bad data during fine-tuning, leading to a drop in overall output quality.

[inland oar]

anyone having issues with GPT4 repeating within the same answer over and over? had this happen with code, thought it was code specific, but also happening with plugins and text responses

[floral grotto]

Said the same thing brother

@inland oar

[gilded tusk]

way to get it free maybe?

[floral grotto]

Nope prob not

[storm yew]

So it’s interesting that Bing is now the default search. I’ve had nothing but a negative experience with search so far, so hopefully it’s a good move

Just quite keen on keeping church and state separate (as in, openai and Microsoft)

[static isle]

What makes you say that?

[storm yew]

Microsoft is bound by generations of bureaucracy, process, and safeguards. They are great at what they do, but I don’t want those same safeguards on gpt. You can see the direct comparison between bing chat and gpt today. Bing just isn’t as good

[spring estuary]

I just hate Microsoft for their EEE policy, and I fear they're attempting to do that with AI now.

[storm yew]

Yep. OpenAi is it’s own company, but again… church and state

[reef coral]

what?

[wraith ginkgo]

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 @open crescent immediately.

[boreal epoch]

My issue is Microsoft put an auto deleter onto their version of GPT that removes outputs and it has a false positive rate so high most valid outputs get deleted.

[chrome mantle]

People have been saying this since chatGPT came out. It could be that its getting worse because of the consoring they do with each jailbreak or it could just be people expectations are getting too high and so they feel its mistakes much more.

[wraith ginkgo]

`` Server Tip: ``

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

[snow slate]

i feel that way too

what are you typing that's so long?

[glad dagger]

i want to pay gpt plus

[snow slate]

sandi?

[glad dagger]

why only ameraca can pay plus ?

[snow slate]

no idea

might wanna ask chatGPT

[cerulean obsidian]

who said? I'm from India and been using from past 3 months

[glad dagger]

Do you use India's credit card to pay for your plus subscription?

[agile vortex]

no its worldwide, are you trying to use a prepaid card?

[obsidian cedar]

he's tryna use robux

[agile vortex]

and what country are you trying to buy from

[glad dagger]

i use visa card from china

[agile vortex]

thats the issue, chatgpt is not available in china

[glad dagger]

that is frustrating

i wanna buy plus

Which countries does OpenAI support?

[tepid gull]

@chrome mantle @ornate bramble I noticed that a huge portition of plugins on web chat fall back to use either a GPT-3 or GPT-3.5-turbo model. For both of my partner and I, when we enable a plugin on GPT-4-plugin model this happens. For example, AskYourPDF after reading a PDF file reports its model is GPT-3 if queried. Workaround I found was to instruct it to state its full model version string as the very first thing and then proceed to the actual task, which includes reading the PDF file — both instructed in the same prompt. This enforces it to use the GPT-3.5-turbo model, which at least gives a somewhat bigger context window than the GPT-3 model and thus it doesn't immediately produce garbage.

[agile vortex]

most countries besides like 5-10 countries such as russia, china north korea ...

[glad dagger]

emmm

[ancient yarrow]

Better blame the great firewall of china, not openai..

[agile vortex]

exactly lol

[glad dagger]

yes ,greate firewall is so bad

i hate it

Can Japanese bank cards purchase chatGPt plus subscriptions？

[tepid gull]

If I have to guess the service is so overloaded due to increased demand the release of plugins caused that it drops some users that use plugins to older models. The native web plugin states its model version to be GPT-3, which would explain why people get those poor quality answers when trying to use it.

[agile vortex]

yes but you must be in a valid country to buy it, others have tried and they did not get access to chatgpt plus

[obsidian cedar]

it's not available in North Korea, China, Iran, Cuba, or Syria. Other that that you should be good

[agile vortex]

for web search it'll be bing or i recommend one of the other web third-party plugin

i mean as in the native web plugin will turn into just 'search with bing' (and its already bring rolled out)

[tepid gull]

Those have the same limitation, but they too can be made to use the GPT-3-turbo model with the trick I mentioned.

[agile vortex]

true but the new bing search will work for both free and paid users

for paid users we get gpt4 or gpt3.5 others just get gpt3.5

[terse ibex]

I would like you to create a virtual presenter with online AI tools. Please watch and create a video with maximum 5MB size. Please use chat GPT to create small script about history of artificial intelligence. Then use leonardo AI to create a cartoon character of your choice. Use your script and character with DID to generate your virtual presenter (please use the youtube video description for generation).

[agile vortex]

this is not chatgpt

please go to https://chat.openai.com/ to try out chatgpt

[ancient yarrow]

Imagine asking chatgpt to make a video

[tepid gull]

But since this is most likely caused by the new wave of huge demand, I hope they invest more to computing to match the demand.

[agile vortex]

it can with plugins from chatgpt plus plugin store haha

[agile vortex]

agreed, i assume the new bing search will be better, as it'll use microsofts own specific servers and not just openai's + its also shown to be a lot lot faster, even the gpt4 version write nearly as fast as gpt3.5

[tepid gull]

What do I need to get my MS account to recognize that I have the plus subscription? Or does MS have their own product for this?

[agile vortex]

im not sure what your asking, as for bing search, it wont require a MS sign in as far as i know

[safe vale]

h

[leaden cape]

it does for me

[tepid gull]

Ok, well you mentioned paid users have a chance to get GPT-4 model, so how does it know I have a subscription?

[agile vortex]

im talking about the bing search plugin

[agile vortex]

check #off-topic

[leaden cape]

Ohhh you mean WebPilot

[tepid gull]

Ah there's a plugin, never mind then. Thanks 🙂

[agile vortex]

no, its that bing is replacing the current web browsing

[agile vortex]

its not a plugin as in one you download from plugin store, its replacing current web browsing feature

[tepid gull]

Yeah, got it.

[glad dagger]

Can I pay for my GPT subscription using the credit card of Standard Chartered Bank in China？

[agile vortex]

man it wont work 🤣

you can pay, but you wont get access to it

many others have tried, GPT4 itself is banned in china as well as openai is restricted there

[glad dagger]

If I can pay, I can use it

[sinful rapids]

i agree sometimes when i am using plugins i find that the model has lower intelligence and poorer capacity of data process, so mind you telling me how to solve it

[tepid gull]

I have found no way to enable GPT-4 model for any plugin and my guess is this is caused by the service running out of resources due to the high demand. Best I have been able to do is to enforce the GPT-3.5-turbo as I explained in my message #gpt-models message (edit: sorry, you replied to different message, so I added a link to my message with the WA)

[wraith ginkgo]

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.

[high sky]

if i want my gpt api used by mutiple session and each session keeps the memories of the previous conversation, what can i do? how to implement this ?

[lilac gull]

Local caching.

Cache messages in database and have a uuid for the session etc

[tidal trout]

Hello can someone help me find my 2FA code to long in to desktop?

I have no problem logging in on my mobile as the 2FA code pops up automatically and fills in the field but when I try to log into desktop I do not know where they expect me to get the code from. I know it's on my iphone somewhere but I can't find it

[sinful rapids]

thanks

[full fulcrum]

Any good devs familiar with building projects on Open AIs APIs in here?

[marble heart]

Which chatGPT plugin can interpret images?

[tawdry charm]

Havent seen one yet. Code interpreter "maybe"? But thats not fully available yet

API goes brrr

[signal eagle]

day 69 of waiting for gpt-4 api access! awesome

[hallow owl]

oof, that's a bit of a wait

[tawdry charm]

I got it rather quickly.

[hallow owl]

I got API access after 24 days

[signal eagle]

i even pay for plus and spend like $30-50 a month on my account for the 3.5 api, insane

[tawdry charm]

Yea about the same I would say

[tawdry charm]

Have you mentioned any specific usecases that require GPT4 in your request?

Like developing smth?

[signal eagle]

First time I applied, I just said I wanted to develop stuff on top of it, nothing specific. Second time I applied (after 1 month), yeah, made a medium-length post

[hallow owl]

this may sound like a dumb question but did you get a confirmation email that you're on the waitlist?

[signal eagle]

Yep, both times I applied

[hallow owl]

ok, just wanted to be sure

[tawdry charm]

Mhh. I am pretty sure that i got access so quickly because I already had a very specific idea of what i wanted to build and why it requires specifically GPT4 and cant be done with GPT3.5.

[signal eagle]

well yeah gpt-4 is miles better. of course I mentioned in my application that it will be more accurate and will hallucinate less. still nothing.

[tawdry charm]

I dont think thats enough. Just saying "because its better" translates to "bro was too lazy to optimize his code".
Certain tasks like non english tasks, very complex stuff, very long context etc., basically tasks that really push what is even possible with GPT are probably highlighted for them. Or its just luck, idk

Because you can counter hallucinations even with 3.5 quite well imo.

[signal eagle]

yeah. mentioned I need larger context for my use case. it was a pretty lengthy post even lol, specifying how gpt-4 will be better for the use case

[tawdry charm]

Even with davinci if you put in the effort.

Yea idk. Maybe its also a random factor.

[nocturne sierra]

Does anyone know why I receive so many "network error" errors when using ChatGPT4? I mainly use it during my study on IT security. Is it cutting itself off because it might think its violating its own policies?

[sinful rapids]

How many people have learned that when using the plugin, the chatgpt model will degenerate to 3 or 3.5, and I am the only one who has not discovered it until now

[hallow owl]

firstly it can't go to GPT-3 because that's not even a chat model, secondly what makes you think it switches to GPT-3.5?

[sinful rapids]

In the discussion just now, someone said this, I just opened the plugin mode to ask questions, I asked him which model of chatgpt you are, he replied to me that he is a model library based on the chatgpt-3 architecture

[hallow owl]

yeah, that's not a reliable indicator because ChatGPT knows nothing about itself

[fringe granite]

dragon fruit, dragon head comes out of it, green tea leaf, vector, icon, tea packaging design

[sinful rapids]

But when I use the default model, he tells me he is chatgpt4

[tawdry charm]

Would be very weird considering that the plugins will call the GPT4 api. So maybe its becoming a dummy while using the plugins, but it is still GPT4

[sinful rapids]

sounds reasonable

[tawdry charm]

Honestly my suspicion is that the plugins are coded like trash

[sinful rapids]

lol

[quartz bane]

Hi guys I am new here and also new with discord, could someone give me an instruction where to ask questions on prompts? So that I can evolve my chatgpt to follow better instructions and make itself have more information? And how I make chatgpt use the internet to research my requested topics?

[hallow owl]

you can check out the #1019652163640762428 and ask questions in #prompt-engineering

[quartz bane]

Very thank you!

[lilac gull]

Has anyone submitted a successful eval?

[wraith ginkgo]

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 @open crescent immediately.

[pseudo hatch]

gpt 4 is getting worse day by day and gpt 3.5 is getting better day by day

[opal leaf]

who knows how to get use of gpt in daily vlogs

[hallow saffron]

#mj invite

[mint oak]

If there's someone here that knows how to use plug-ins really well, dm me. I'd really like some help setting some things up.

[hallow owl]

ask here, then everyone can learn, what do you need help with?

[mint oak]

Good idea. So I'll be blunt. I'm currently trying to understand what are the best plug-ins for service based businesses. Most of the things I've seen seem to only apply to information businesses.

[hallow owl]

do you mean for things like customer support? logistics? or something else? There are new Plugins being added to the Plugin store every day and it's very early days, perhaps soon some plugins will arrive that more closely match your use case.

[glass fjord]

@hallow owl You still online?

[hallow owl]

I am, what can I help you with?

[glass fjord]

Get some rest man.

[hallow owl]

I've only been awake a few hours today 🙂

[glass fjord]

@rigid estuary Make this dude moderator

[hallow owl]

oh no, I don't want to be a moderator, too much hard work

[glass fjord]

Hmm why can i not see plugin store

I've deleted cache and cookies

Nvm, just super slow

[hallow owl]

how many pages of plugins do you see on the store?

[crystal field]

I've run into several situations now where GPT will now assist me with shell scripts because they're either root issues (tcpdump) or i need to kill a process and it won't help because the word kill is there

does anything work to make it not be annoying? At least for hacked code it just says "don't use so many try except blocks"

Telling that that I'm root on my local Debian box kind or worked

[supple violet]

work around it and for example say end process

[wraith ginkgo]

`` Server Tip: ``

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

[radiant wyvern]

As far as I know, you don't manually pick the plugin and use it. Gpt logic pics which Plugin to use depending on your promt.

[knotty grotto]

Certain word will trigger the plug in you chosen for the chat

[stiff plinth]

I had an exam i asked gpt to tell me most important (the topic's questions or category) for (the exam), it showed me some results and guess what! 80% of the exam came from there

[crystal field]

lol

I'll just translate my prompts with deepl then, it's often far more helpful when not talking to it in English

[void kayak]

😂 that's hilarious

[inner yew]

Do we have a date when the Code interpreter will be released to the whole of the plus userbase?

[hallow owl]

no ETA, keep an eye on #announcements for possible updates

[inner yew]

thanks 🙂

[young pumice]

Complete noob here but is there any value in maintaining an offline version of Chat? It seems that much of the innovation is coming from the commercial sector and the private sector is lagging in keeping up (unlike image generation)? If so what repository should I look at?

[knotty grotto]

Don't think since it need the database and it is huge

You can discuss in #ai-discussions for the llama product. Local on pc but you need a good pc to run it

[young pumice]

Thanks for your thoughts. I put in here because I am thinking more about the innovations aspects. I will put it there as well.

[stiff plinth]

haha yeah

[wraith ginkgo]

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.

[river flint]

what is chat gpt 4 actually doing when it says its "thinking" when using web browsing

[knotty grotto]

Calling function

[digital sluice]

1

[flint perch]

so when will this be able to search stuff online?

[slate oracle]

I have an issue with GPT 4 and 3.5 none of them has been able to output The One Ring poem even when I try to ask it to modify it

The output starts but always ends quick : ``` Certainly, the poem about the One Ring, also known as the Ring-inscription, is from J.R.R. Tolkien's "The Lord of the Rings". It goes like this:

"Three Rings for the Elven```

[tawny vector]

It took Shakyamuni seven days to become a Buddha! What exactly did he go through? Write a 2000 word article

[hallow owl]

This is not ChatGPT, you can access ChatGPT here - https://chat.openai.com/chat

[tawny vector]

I haven't been able to access it today, what's the matter

[hallow owl]

what happens when you try?

[tawny vector]

Reminder Sorry, you have been blocked

[hallow owl]

what country are you in?

[tawny vector]

The VPN I am using is in China

[hallow owl]

so you're in China?

[tawny vector]

Yes, China

[hallow owl]

China is not a supported country, you won’t be able to use ChatGPT or the GPT API.

[tawny vector]

But when I use VPN, neither Japan nor the United States can handle nodes

[hallow owl]

you're in China, you are not allowed to use ChatGPT

[tawny vector]

Okay, thank you very much

[frozen veldt]

surprised they're allowed to access discord

[tawny vector]

Are you surprised? We can access anywhere without any restrictions

[shrewd citrus]

Chat GPT browse function is absolutely off the wall crazy good. You can get it to look at a news article, or transcript and immediately identify all logical fallacies present in the article, and identify other things like dog whistles, and known evasive corporate verbiage.

[balmy sleet]

promp/

[drowsy fox]

Hi,
I have use rest template, and integrate curl of chat model using rest template, I getting response very slowly.

[glossy plover]

I am starting to be very pissed now... I subscribed to gptplus but since a few days I got this from every time and then get the message saying you reached your number of generation for GPT 4. There is no way to contact support instead of a bot, I thinking more and more to u subscribed this scam...
What should I do?

[stiff trellis]

how do i add chat gpt bot

on discrd

[knotty grotto]

You can't here

[knotty grotto]

You are limited with 25 prompt every 3h. Support can't change that

Also this isn't scam since it is information you can find on their website and tos

[glossy plover]

That's not the problem....
I have every time something went wrong as response and it take it as a prompt generated

[silk bobcat]

is anyone aware what the red, crossed out, shield icon means next to plugins in ChatGPT?

[knotty grotto]

What country are you in ?

[haughty juniper]

hello all and ki

[glossy plover]

France

[knotty grotto]

Do you use vpn on your computer ?

[haughty juniper]

german

[glossy plover]

No

[knotty grotto]

What browser are you using ?

[glossy plover]

Brave, opera, Firefox and Chrome

But it's only with few conversations using GPT 4. Gpt 3 is fine

[knotty grotto]

Since yesterday users have been reported some bug with gpt 4.
2 fixes that helped most user

1. log off, delete cache and cookie, log back
2. try incognito

[glossy plover]

Thanks its true that I didn't used incognito. But it happend on my laptop and phone

[sly haven]

hi, guys, wow, more foreigners

[knotty grotto]

If you can try, let me know.

[glossy plover]

Gonna do it now on my phone

[spark roost]

web browsing is out already? they said ''next week'' at day 12/05

[knotty grotto]

Yes, working good

[spark roost]

can you print? i had gpt-4 like 3 days ago, and didnt see this option

[knotty grotto]

Go on setting on chat.openai.com and make sure you enable it

[glossy plover]

No still the same

[knotty grotto]

Can you try on pc , Firefox, log off, delete cookie and cache, log back ?

[glossy plover]

I don't have my pc now but will do

[bleak peak]

how do I get to gpt-4?

[knotty grotto]

You get the subs option

[glossy plover]

But it seems to be only in a conversation after I asked to generate a planing table for a project. And it seems to happend when you copy past things

[bleak peak]

how?

[knotty grotto]

Chat.openai.com , and pay the 20$

[knotty grotto]

If you start fresh convo does it work ?

[bleak peak]

BRUH

[glossy plover]

Yes

But it's still freaking annoying cause Thoose are my important conversation for project

[knotty grotto]

Temporary solution I guess.. I suggest to #1070006915414900886 . Openai employee do check rhe report

[glossy plover]

I did it but I don't think it will help tho... I did a backup conv with gpt 3 in case it happend but it's not the same and I subscribed for GPT 4 so...

[stuck kestrel]

Adding browser option and plugins to chatGPT4 lower stability and output quality. Now I am using chatGPT3.5 - the quality is similar.

[frozen veldt]

Always good to hear actually! I do a lot of work with friends in China. There are discord servers that connect GPT, you might want to look into that

[forest glade]

GPT4 output is getting worse and worse by the day. I asked it to generate some code for trading view, later GPT4 asked me to give it the full code that he generated, and asked me to contact trading view support. i mean tf? 😤

in another chat I gave it some information, and told gpt4 not to answer me yet before I ask any questions, then the answer? "I am sorry I an unable to help...."

[harsh oar]

Bing Integration is why GPT went to crap. Microsoft Build explains it.

[dusk dome]

im unable to select gpt 4 anyone else?

whenever i select it, it switches back to 3.5

[forest glade]

what?! isn't BING integration and ChatGPT separate thing? it's from the same tech, but isn't Bing is it's different version?

[wraith ginkgo]

`` Server Tip: ``

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

[forest glade]

so it's not just me feel ChatGPT is acting weird lately 🫠

[wraith ginkgo]

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 @open crescent immediately.

[quaint jungle]

i thought that chatGPT could read external links

[knotty grotto]

With plug in they can read some external link

[quaint jungle]

how can i give gpt a large code, what plug-in i need to give it a large code to disect

[knotty grotto]

I'm.not familiar with code and programming, maybe someone else could help you. I think you will need co pilot. Not yet release

[quaint jungle]

hmm ok thanks

[rare relic]

like GPT-3.5 taking hours to generate a few prompts due all fails of system attempting to get them go through isn't exactly just a minor peak time performance issues while GPT-4 system just cruises if people switch to it as solution to pay more... it's just something to look into like maybe stop allowing new users temporarily if needed

[frozen veldt]

You probably hit context limit

Context limit comes up a lot for me with gpt-4 code interpreter. It was fine at 70 lines, 200.. but now as its near 500 it can barely load it into the code interpreter and answer a few qeuestions before it starts to lose fidelity. Would LOVE to see a gpt-4-32k-code interpreter model

its at the point now where I can't really use it to improve the code at all

[wind urchin]

same same

are chatgpt plugins down?

[hallow owl]

I don't think so, are you getting an error message?

[wind urchin]

Error communicating with plugin service. Please try again later.

Not for wolfram though

[hallow owl]

which plugin? I'll try it here and see if I get the same

[wind urchin]

most of them for example ChatWithVideo

and Scraper

[hallow owl]

ok hang on I'll have a go

[wind urchin]

and Access link

[tidal river]

chatgpt seems to be down period right now. not even the chat.openai.com site is responding to prompts

[wind urchin]

gpt-4 normal is working

4 me

(╯°□°)╯︵ ┻━┻

[tidal river]

would send a screenshot if i could lol

normally a forced refresh fixes it lol

[wind urchin]

oh

[tidal river]

3.5 and 4 both seems to be having issues

[hallow owl]

GPT-4 and the WebPilot plugin are working fine for me right now

[wind urchin]

gpt-4 is fine

a lot of plugins are down

[quasi flint]

yeah

[wind urchin]

||(╯°□°)╯︵ ┻━┻||

[tidal river]

yeah ive tried 2 different browsers and nothing is working to get it to respond to prompts and my 2 plugins cant connect

[wind urchin]

yeah

[tidal river]

sadface lol guess i come back in a hour

[quasi flint]

but for most plugins

[pseudo solar]

Hmm, i found stranger issue... Try to go to Data Controls, disable Chat History and Training - then GPT-4 lost Web Browsing or Plugins.

[quasi flint]

its a 500 error

for most plugins

to /backend-api/conversation

but for a few its a QUIC_PROTOCOL_ERROR instead for some reason

[wind urchin]

ahh

[sour seal]

gpt4

[hallow owl]

I tried a few plugins, WebPilot, Link Reader and AskYourPDF and they're working fine

[wind urchin]

oh ok

[quasi flint]

seems like @round vector and @swift radish have the same issue

[cosmic python]

||Helluy||

[mellow dune]

Anyone else notice GPT4 got dumber?

[quick kelp]

When was the latest leak?

[dusty snow]

no leak

[mellow dune]

It’s literally just writing complete BS code now

[dusty snow]

Yes it is useless...

[mellow dune]

Anyone found any workaround on this? Super frustrating, if infra cost is a problem just increase the cost.

[unreal swallow]

how do I attach the file

to that chat?with the pdf extension

askyourpdf / link reader

[frozen veldt]

I hear this a lot.. can you be more specific?

[unreal swallow]

through the chat interface? how do I attach files etc

[hallow owl]

here....

Make sure the PDF you want to upload is hosted online and publicly accessible.
Copy the URL of the PDF.
Provide the URL to me, and I will use the AskYourPDF plugin to upload the document.
Please note that the document will be saved to a vector database, and it will be searchable for any queries you might have related to the content of the document.```

[wraith ginkgo]

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 @open crescent immediately.

[modern thorn]

@1052826159018168350 hello, I still haven't received information on how to resolve "Oops, an error occurred!" so I can access my account. This is affecting my Plus subscription because I am unable to use it entirely.

[wraith ginkgo]

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.

[inner trail]

@hallow owl Do you know what "IamRich" plugin is for? what can i do withit?

[hallow owl]

It's a joke. Many years ago when the Apple App Store was new someone made an app that did nothing except show a gemstone icon on the screen and charged $999.99 for it. It lasted on the App Store for a few days before Apple pulled the app. Someone has made an I am Rich Plugin, pointless 🙂

[inner trail]

Damn, i just installed it lol

Will uninstall it right now, thanks for your answer

[lilac meteor]

can we not have meme plugins like this, they take up too much space. either that or just add a frickin search bar

[sharp steeple]

Is there a browser extension where you can organise gpt conversations, and if possible, allow users to perform a search on existing conversations to search for something?

I know that this feature doesn't exist yet on the website, but it might be helpful for users to navigate through large lists of conversations if there is any solution currently available

[wind urchin]

Is Bing Chat GPT-4?

If it is, then what's the point of buying Chat GPT Plus?\

Oh, so bing is not based on GPT4.

Can we do images and stuff in GPT4?

Or only text based.

[boreal delta]

if you start a conversation, in gpt4 and then run out of the responses in a given time frame, then enable gpt 3.5 is there any way to reenable gpt4 afterwards?

[wind urchin]

So when will GPT4 20 prompt per day limit be removed?

Only then I will buy.

[inner trail]

I think Bing uses GPT-4 API, but its not good as far as i tested.

[wind urchin]

Still pretty bad.

[shy ingot]

do u need to pay chat gpt

4

[wind urchin]

But bing ai is still better than 3.5 right.

[mellow dune]

The code it writes now is very inaccurate with significantly more errors than before, it makes excuses on why it can’t write different functions such as “this is a complex task”

I’ve also asked it to write a function and what it does is it calls a different function that doesn’t exists within the function that I asked it to write and says put your own implementation of this here

[wind urchin]

Since its still based somewhat on GPt4 API

[inner trail]

Good question, but i kinda feel like they are the same

[desert kelp]

🐝

[wind urchin]

How good is GPT4 as compared to GPT3.5 in terms of homework?

[inner trail]

GPT4 could be better in every-aspect

[wind urchin]

3.5 is really bad at coding.

[hallow owl]

As of early 2023, OpenAI has been discussing the possibility of using a cryptographic watermark in the text output of ChatGPT. This watermark would be designed to make it easier to identify text that has been created by the AI chatbot, with the goal of preventing misuse of the tool, such as using it to write papers for school or college.

[wind urchin]

I wonder is GPT4 is really good at coding.

[inner trail]

Its the best for coding, never saw a better AI

[wind urchin]

Dang, but there's like humanizing ai as well.

I used one and it dropped the ai probability to like 30%.

[wind urchin]

Hmmm. If there was no prompt limit I would definitely get it.

[inner trail]

You have 25 messages for 3 hours, not really bad

[wind urchin]

I mean.

It should be unlimited since you are paying.

Sometimes you have to like do a lot of prompts especially when completing school work.

[inner trail]

I know, you have a good point, but it is what it is

[wind urchin]

Yeah, im sure they will remove the prompt limit.

[supple violet]

no they wont

[wind urchin]

What why.

[supple violet]

not from plus

[wind urchin]

I thought it was because heavy loads on server.

I would pay double for a new subscription if there were unlimited prompts.

[inner trail]

I don't think they will remove the limit, but my hope is to be "50 messages / 3 hours"

[supple violet]

"ChatGPT Plus subscribers will get GPT-4 access on chat.openai.com with a usage cap. We will adjust the exact usage cap depending on demand and system performance in practice, but we expect to be severely capacity constrained (though we will scale up and optimize over upcoming months).

Depending on the traffic patterns we see, we may introduce a new subscription level for higher-volume GPT-4 usage"

[wind urchin]

What's the reason for the limit though.

Is it solely because of heavy traffic?

[supple violet]

the current 25/3h limit is pretty good as with api the same would be 500-1000 USD

[wind urchin]

Why not get better servers, OpenAI must be rich by now.

[gaunt dagger]

never

[supple violet]

nvidia is the only supplier and unable to provide enough hw world wide

[wind urchin]

Dang, how much traffic GPT4 getting anyways?

Must be insane judging from even Nvidia not being able to support right now.

[supple violet]

what never? you can calculate it yourself

[wind urchin]

Can GPT4 solve SAT mathematics questions correctly.

And when will we be able to add graph questions.

[sharp steeple]

Nevermind, gpt 4 managed to answer this question with browsing plugin lol

[gaunt dagger]

Im using api with gpt4 the whole day for a few cents.

[inner trail]

Really? how much cents actually? curious to know

[supple violet]

if you use 100 gpt-4 messages with plus each day that's 3000 messages in a month with no separate token cost. with api ONE 1k input and 1k output costs around $0.1, so 3000x that's $300 alone... but normally you will a lot more than 1k token with each call

[wind urchin]

When GPT 5 gets released, will GPT 4 AI become free?

Or is there no more after GPT 4.

[inner trail]

There will be more, but we don't know what will they call it

4.5 or 5 idk

[wind urchin]

I really wanna see GPT AI get a physical body.

[gaunt dagger]

Model Prompt Completion
8K context $0.03 / 1K tokens $0.06 / 1K tokens
32K context $0.06 / 1K tokens $0.12 / 1K tokens

[wind urchin]

Hopefully that happens.

[supple violet]

if you work with a story or a code you will easily use the whole 8k context and each call will begin to cost you ~$0.3 so with 10 messages its already $3

[eternal marsh]

what is better? chatgpt, bard or bing?

[hallow owl]

You missed out Claude

[wind urchin]

gpt is the big daddy

rest are losers

[eternal marsh]

I don't try Claude yet is good?

[eternal marsh]

is good i am premium but pluggins web browser have some problems...

[wind urchin]

main difference i think is the data source, bard is continuously pulling data from internet

While GPT is till 2021.

Overall intelligence goes to GPT 4 though.

[eternal marsh]

i only put money in chat gpt but i love bard...

[wind urchin]

I cant use bard because they locked it to 18+.

[eternal marsh]

oh

you can change your age?

[wind urchin]

Dont want to.

[eternal marsh]

i see

[wind urchin]

I have OCD.

I want everything, perfect.

I wouldn't be able to sleep if I got a fake age.

[gaunt dagger]

Okay. Then I guess I'm not there yet or I have a different pattern of use. I use it for programming in my job. But so far, I haven't had a scenario that constantly requires the entire context.

[wind urchin]

Does better prompts make AI better?

How do we get these good prompts.

It sucks that it listens to your instructions, not your hearts.

[hallow owl]

yes, good prompts are conducive to getting good output

[wind urchin]

Cant wait for Neuralink to transplant GPT AI into my brain.

[frozen veldt]

"prompt engineering" some people scoff at the concept but getting it to generate the responses in the format you want and flavor you want can be extremely challenging

[wind urchin]

Yeah.

It's hard.

[hallow owl]

do you have GPT-4 access?

[wind urchin]

Im going to go take a look at the prompt library.

[wind urchin]

Not yet.

But my dad has it, im gonna ask him to get me it.

I'll try it out on his laptop tommorrow.

[rough lagoon]

My chatgpt can't connect to the plugin server. By trying to check if it is already working I have lost my 25 opportunities to use GPT-4, is it possible to reset this counter?

[hallow owl]

wait 3 hours now unfortunately

[rough lagoon]

Bruh 🙃

[prime flame]

Hi GPT gang. Trying to use it to collect and summarise some data from several links, but it consistently shows reading content failed. Any fixes to that?

[trail pulsar]

Does anyone know python

[simple aurora]

me

[trail pulsar]

Go pv

[sinful scaffold]

Hello everybody.

[void kayak]

is it just me or has GPT-4 written responses much quicker lately?

[marble kindle]

does it seem dumber too?

[void kayak]

I'm sure it used to be slower than it has been lately

Possibly a little, come to think of it

[marble kindle]

i havent used it in a but theyre prob gonna release a newer version soon so theyre making 4 a bit dumber so the newer one seems way way smarter

[void kayak]

interesting

[marble kindle]

idk if thats true tbh but from past experiences thats what it feels like

remember how chatgpt 3.5 used to be pretty smart but now its dumb af

[void kayak]

overall it still appears to be smarter than 3.5, but it does feel a little dumber than usual lately while being faster. For example, I have to send an additional message or two maybe to further clarify my intentions

[acoustic eagle]

Hey guys. I'm trying to create a promot when I want to refactor some react component. Does anybody have more suggestions or modifications. Love to hear your thoughts:
Assist in refactoring the provided React code with the following points:

1. Optimise it for maximum performance and maintainability
2. Apply recognized react design patterns.
3. Incorporate the following design patterns when applicable:
   -- render props
   -- custom hooks
   -- provider pattern,
   -- compound components,
   -- composition.
4. Ensure it gracefully handles various scenarios gracefully
5. Apply SOLID and KISS principles.
6. Improve the code's reusability and enhance flexibility.
7. Strive to make the code more resilient, scalable, and efficient.
8. Rename variables for better self-documentation and readability.
9. If appropriate, extract code into separate reusable components.
10. Consider moving repeated settings, configuration, or complex prop values into capital constants for improved readability and maintainability.
    After the refactored code,please provide a detailed explanation for each change, explaining why it improves the code.
    Include article links where available to support understanding of the reasoning behind each refactoring decision.
    If you understand me and are ready, please ask for the code to refactor.

[marble kindle]

ah yes i see what you mean. It could also be that our expectations are growing (prompts getting lazier)

[void kayak]

true

[marble kindle]

yea but hey it still gets the job done well xd

[void kayak]

also true

[foggy sage]

I think Bing is over-moderated.

[wraith ginkgo]

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 @open crescent immediately.

[void kayak]

yeah, I've tried Bing's a few times. It feels like it's something for kids to use or something. At least it feels heavily censored or watered down in some way

[hallow owl]

I won't use Edge, I think Microsoft are limiting adoption of Bing AI by forcing use through Edge

[void kayak]

indeed, I won't use Edge either 🙂

[tranquil phoenix]

hey so, when is gpt-4 gonna connect to the internet?

[hallow owl]

it is now, with Web Browsing and Plugins

[tranquil phoenix]

where, how

[hallow owl]

go to Settings -> Beta Features and turn on Web Browsing and Plugins

[tranquil phoenix]

Activating plugins is recomended?

[hallow owl]

it's up to you, I use them and some of them are good

[tranquil phoenix]

can you share the good ones here please?

[hallow owl]

I use WebPilot (a web browsing plugin), Link Reader, AskYourPDF and Wolfram

[tranquil phoenix]

Also, now that 3.5 is free tier, doesn't it make sense to remove 25 prompts per 3 hours?

[hallow owl]

the 25 prompt per 3 hour cap is there to manage demand and let everyone get a go at GPT-4

[tranquil phoenix]

I get it, but considering it's a paid system and only accessed by paying customers, can't the demand be managed or improving the "supply" from the funding?

oh well

[robust hornet]

Scaling is not a trivial problem.

[hallow owl]

the problem isn't one of money, it's a lack of supply of hardware. GPT-4 is computationally expensive and they can't add server hardware quick enough to meet demand, so they cap usage

[tranquil phoenix]

Maybe they need to expand and start manufacturing hardware... @hallow owl

[knotty grotto]

Also they have to pause new account since they are too many user. They are victims from popularity

[dusty snow]

Is Chatgpt down?

[tranquil phoenix]

"As an AI language model, I don't have direct access to the internet. I can provide information, answer questions, and generate code based on my training, but I can't establish a real-time connection to external websites or services. Is there something specific you'd like assistance with?"

[atomic crown]

same

[hallow owl]

you need to select the Web Browsing model from the GPT-4 model dropdown box

[robust hornet]

It's just GPT-4 that is not working, at least for me.

[dusty snow]

Yes it is not working for me as well

[indigo yarrow]

Yeah GPT-4 is down, 3.5 works.

[tranquil phoenix]

I don't see a dropdown box (but I can't use 4 now anyway...) is the box on the top of a new chat?

[bronze onyx]

does gpt-4 stop working ?

[hallow owl]

yes, the model selector is at the top of a new chat

[frozen path]

as always

[tranquil phoenix]

awesome! 25 prompts with the web browser still, I presume

[deft solar]

Btw, how to select gpt 3 or 4? I am using free version, but is the the same pathway, just it matters on how much you pay?

[hallow owl]

yes, unfortunately

[tranquil phoenix]

Thanks robert

[mystic girder]

GPT-4 went down right when I was trying to demonstrate "GPT-4 with browsing" to my coworkers, rip.

[hallow owl]

murphy's law 🙂

[indigo path]

Looks like it's working again now

[indigo yarrow]

Yeah working now for me too, just came back.

[simple oxide]

Thank god

[mystic girder]

Nice

[wraith ginkgo]

`` Server Tip: ``

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

[bronze onyx]

thanks

[stray halo]

Open ai train his model with gpt4 Api request?

[supple violet]

no

[indigo path]

I spoke too soon... ChatGPT 4 definitely still having issues

[stray halo]

@supple violet @supple violet so I can use that in a medical context?

[indigo path]

"Error in message stream"

[blazing shard]

yea just got that as well

[indigo path]

I hope these half-responses are not taking up my message request limit lol

[glad fern]

I noticed the issues too and checked over here to see if anyone else was encountering them.

[blazing shard]

it will so for now dont use gpt 4

[supple violet]

depends on what level, openai forbids certain medical area use

[stray halo]

Really even for researcher?

[copper juniper]

Even GPT 3.5 is giving error messages at the moment

[indigo path]

😦 big sad

[supple violet]

you can read their policy docs, they have few different ones

[stray halo]

I am working like researcher in health domain

[warm radish]

Is GPT4 keep failing for everyone else? I keep getting either Network error or Error in message stream

[bronze onyx]

its interesting that my notionAI stop working ,too.

[warm radish]

Looks like it

[vague creek]

Yup, failing for me atm

[warm radish]

Thought so, thanks

[fallow adder]

gpt4 buggy?

[vague creek]

https://status.openai.com/

[warm radish]

Found it thanks

I subscribed to updates 💪

[wind urchin]

there is a probably with chatgpt

its not letting me use it...

[warm radish]

Yeah it is not just you

[dusty snow]

Is it working now?

[warm radish]

Not yet it dont look like

[random hornet]

Anyone else getting Error in message stream?

[normal turtle]

"Error in message stream" always appears

[warm radish]

Everyone is

https://status.openai.com/

[random hornet]

Got to love it. Time to go on my playground and use that API now.

[fallow adder]

sometimes

[versed pagoda]

Ohh man… I was doing my French homework :(

[warm radish]

🤣

[tough gazelle]

exam tomorrow ffs 💀

[warm radish]

Back in my day we had to do our own homework 🤣

[versed pagoda]

It works now, after 3 tries ahah

[tough gazelle]

nah just revision, explains things better than my prof usually

[indigo path]

Anyone have suggestions on getting ChatGPT 4 to actually code me a website? Feel free to DM me, I have quite the prompt written up. Used ChatGPT to iterate on it a bunch, but it insists it will only write me an outline with code placeholders lol

[minor jay]

Hmm...something seems to have gone wrong.
is this only to me?

[warm radish]

💪

[robust hornet]

Back in my day, I skipped doing my homework, because it was mostly pointless busywork that had nothing to do with learning.

[tough gazelle]

bump

[random hornet]

Oof you need autogpt for that @indigo path

[warm radish]

It is not good at coding an entire website, you can give it small parts of your website to help you with and it works great.

[tough gazelle]

back up again now for me @minor jay

[quaint jungle]

wow people want a whole website created by chat? im sure it will be bug free lol

[indigo path]

Roger that. Thanks guys. @random hornet, you mind if I DM you? Never used AutoGPT before, but have great technical aptitude (working in IT as a Senior Support Engineer). I'm sure I can figure it out, but would be nice to know where to start, etc.

[indigo path]

I don't mind bugs. Just need a head start, with my minimal HTML/CSS knowledge haha

[warm radish]

I have yet to think of a good use for AutoGPT

[indigo path]

I've had it write me a comprehensive pentesting script in Python, was impressed

[versed pagoda]

Guys I’m kinda new to gpt 4, how do I use “code interpreter”? I can’t find it

[warm radish]

code interpreter?

[quaint jungle]

it will give u gray hair fo sure, limited knowledge of javascript/pyton might help, but html and css, ur basically bulding a static news paper

[random hornet]

@indigo path you need to have access to the chatgpt4 api whitelist to be able to use autogpt.

[versed pagoda]

Isn’t that a thing?

[indigo path]

Rekt. I'm screwed on that front then 😂

[supple violet]

you dont, its in closed alpha

[warm radish]

No wonder I have never heard of it lol

[random hornet]

You talking about vscode insider edition of github co-pilot nightly and github co-pilot x?

[versed pagoda]

So ig I can just use the normal gpt version without any plugins?

[quaint jungle]

how hard is it to create a higher teir and charge double? im sure it will reduce the high traffic.

[supple violet]

if you have plus then you can use them as they are in beta now and can enable them from settings

[glass haven]

Anybody having issues with GPT4 suddenly?

[tough phoenix]

Anyone getting "Error in message stream" error constantly?

[wind urchin]

Yep

[glass haven]

☹️

[random hornet]

https://status.openai.com/

Bookmark that link and just check it next time something happens.

[stuck dock]

I want answers!!

why 503, plzz

[quaint jungle]

i want a higher teir

[glass haven]

Usually it's not up to date so I check here first

[worldly bluff]

I am

[glass haven]

Have we ever found out for sure whether the failed messages count towards our limit or not?

[stuck dock]

jesus, Ive become dependent on GPT-4

where can i seek help, if i can ask gpt-4 about it?

[quaint jungle]

😆 me too

[worldly bluff]

i cant give you an answer until GPT comes back

xD

[quaint jungle]

🤣

[stuck dock]

the world stops

gpt-4 is the world, omg

have you noticed the matrix could also be spelled gpt-4

[worldly bluff]

ofc gpt 3.5 works just fine lol

[void wave]

I guess we should.. go outside?

[worldly bluff]

you mean like where the sun is?

[indigo path]

I've touched enough grass this week 😦

[pulsar hollow]

Did the regenerate response button disappear for anyone else? 😐

[main aurora]

yeah having all sorts of problems. sometimes the chat wont load at all. sometimes gpt4 wont answer the prompt. kicks back errors. and then if i try again sometimes it comes back

[indigo path]

Who's got API access and wants to throw something at AutoGPT for me? Will pay 👀

[pulsar hollow]

Uff, hope they fix it soon. 🥲I can get past it by just submitting the prompt again but it's not the same. Seems it's the same for 4 and 3.

[inner trail]

Don't have API access, but if you want through ChatGPT i can help

[indigo path]

I have a massive prompt I made through repeated iterations. It's to code a website in HTML5/CSS/JS

A minimalist portfolio site for photography

[inner trail]

I'd help if you want, just DM me

[indigo path]

Sounds good!

[warm radish]

I managed to fix my code myself!!! 🤣 🥳

[warm radish]

IT IS BACK UP!

[twilit yacht]

that's how it should be, it's good to apply ur knowledge too

[lucid lance]

gpt does make you kinda lazy

[warm radish]

Oh yeah well I would be able to give it the right prompts without my own knowledge so there is that 🤣

[vagrant marsh]

Yes he just ate 20$ from me while not workig

[frozen veldt]

I dunno about that steve, gpt has resulted in me working harder than ever

so far it has helped me flesh out everything I need to go to proto stage on a boardgame, build that financial sentiment analysis tool, generate high altitude recipes for my wife, organize and review a bunch of sales data for work, generate new options to formulate.. really crazy what I've gotten done in 90 days

[wraith ginkgo]

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 @open crescent immediately.

[molten ginkgo]

So I was using chatgpt and then I got the message: You've reached our limit of messages per hour. Please try again later.

So I decided to upgrade to chat gpt plus, but im still getting that message?

[hallow owl]

just wait an hour, then it should be ok

[molten ginkgo]

argh. I was just on a roll, I guess time for bed and try again tomorrow 🙂

[nocturne bone]

bros , and girl bros i enabled browsing in gpt but its still jiving saying it cant browse the internet plus i dont see where to enable plugins

[hallow owl]

You enable Web Browsing and Plugins by going to Settings -> Beta Features and turning them on. You then need to start a new conversation and select the Web Browsing model from the GPT-4 model selector.

[nocturne bone]

thanks bro , may the gpt guide you

[jovial dagger]

Is there currently an API for the GPT-4 image comprehension feature? Or is that still coming down the road eventually?

[hallow owl]

that's still something for the future

[jovial dagger]

Is there an alpha/beta thing I could sign up for, for image comprehension?

[hallow owl]

not at this stage, keep an eye on #announcements for updates

[tawdry charm]

actually unsure, does code interpreter understand images now or does it just take the title etc?

[hallow owl]

I don't think Code Interpreter understands images

[tawdry charm]

Tbh before they introduce image comprehension I would actually prefer to get fine tuning

[wind urchin]

🤔 what is your fine tune use case?

almost everyone who thinks they need fine tunes, actually needs embedings

[jovial dagger]

nah don't listen to him, only rush to implement the things that I personally care about. haha

[nocturne bone]

whats all the FAISS about embeddings anyway

[wind urchin]

embeddings allow you to take your text based documentation, a corpus, if you will, and embed that into a vector database using a lesser model as a similarity function, once that's been done you can query the vector database for relevancy and produce a new context that is custom to your query, that way you get the advantage of massive datasets with a smaller models speed and low cost

[jovial dagger]

But actually, to Foxabilo's point, google "How I Built Supabase’s OpenAI Doc Search (Embeddings)" and watch that. Or just check out the Subabase AI doc, which is powered by it, with GPT4 under the hood.

I would share the link, but AutoMod would get grumpy.

[wind urchin]

we all love automod 😅

[frozen path]

bruh why's it capped at 25 msg at every 3hrs even if we pay

[knotty grotto]

Cause they are too many subs users before paying they said the cap are adjusted in demand

[frozen path]

then it doesn't work, you try and try and the 25 msg are done and u got nothing, gotta wait 3hrs, like do smt

[stark mason]

it's been 25msgs/3hrs since it came out, they've never adjusted it as far as I've seen

[frigid galleon]

any word on when they might update it? lol so disappointing but kinda understandable

[hallow owl]

it was 100 messages at first

[frozen path]

they adjusted it, backwards :d

[knotty grotto]

Do 1 prompt every 8 min and you are fine

[knotty grotto]

It was 100, 50 and now 25

And even with 25 the system struggle with demande could be 20 msg every 3h

[rustic nova]

Hi there, is gpt-4 support the visual inputs or not??
Openai website said it is support, but GPT4 says not?
Who should I belive??

[hallow owl]

vision input is not available yet

[frozen path]

just pay and do 1 msg every 10 min u should be good

[copper halo]

does gpt 4 get detected as often by ai detectors

[rustic nova]

Thanks, it will be avaliable soon right?

[rose sundial]

fart

[pulsar dirge]

How does such one get access t out gpt 4

[rustic nova]

You can ask it

[rose sundial]

money

[river canopy]

bro chatgpt needs to hurry up and lift that 25 message cap lmao

[hallow owl]

we don't know, there has been no indication of it being released in the foreseeable future

[pulsar dirge]

Oh brother 🔥🇺🇸🦅

[jovial dagger]

If you want to, I can share a little python script called "eternal_chat.py", it's basically a CLI equivalent of the ChatGPT web interface. No bells and whistles, but it uses an API key, so you can give OpenAI as much money as you want to have conversations.

I'm not sure if it's self-promotion if I wrote the script myself. Fair warning, I was originally unconcerned about the cost of the Chat Completions, and, uh, well, suffice to say it's a good thing it was on the company's dime.

It also allows you to control the knobs that the API gives you, like temperature, which can make the AI be a little bit wacko or extremely formal, precise, and correct.

[regal moat]

Yea that come with api usage cost. If you are not a company, don't bother. Unless you want to do it for education but yea

[unkempt crag]

How do you glitch chatgpt?

[regal moat]

You ask it

[wind urchin]

whens the feature where u can give it pictures gonna come out?

[jovial dagger]

I can see that before I asked my question, I should have read the comments in here a little.

[stark mason]

Anyone else getting the "Error communicating with plugin service. Please try again later." when using plugins?

[wind urchin]

yea, I got that a few hours ago

lemmi see if it's cleared up

nope, still down

you in the UK?

[stark mason]

I cleared my cache/cookies and it worked for me

But it's not always consistent

[wind urchin]

🧐

[stoic lodge]

anyone have the bing browsing in gpt now? how does it compare? edit: still getting reading/clicking errors

[hallow owl]

do you have it?

[torpid scaffold]

He

[stoic lodge]

yeah, its the only option for browsing for me now

[hallow owl]

you don't have plugins?

[stoic lodge]

o i do, but the only browsing model was replaced with bing, is it still regular for u? havent noticed much difference

[wraith ginkgo]

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.

[hallow owl]

oh yeah the Web Browsing model changed to Bing yesterday, just the label has been updated to reflect that. It's better than the initial Web Browsing that was released beforehand. Not so many errors now.

[rustic phoenix]

anyone know if the plugins chats count towards the 25 messages every 3 hours? I assume they do?

[hallow owl]

they do

[rustic phoenix]

only reason I ask is because when i go to chat with a third party plugin the cap message goes away

ok thanks

too bad you cant use them with 3.5

[inner trail]

Oh wow, i just saw that as well? when did they change it to Bing?

[normal atlas]

today

for me anyways. I used chatgpt like 4 hours ago and it didn't mention bing, but now it does

[inner trail]

Yes same, didn't notice it until now

[normal atlas]

it doesn't seem to change how it searches other than now saying "Searched Bing: “[]” the actual quality of the results seem to be equivalent

[late willow]

Man, plugins were unfortunately so damn buggy for me

well, does depend plugin to plugin

[inner trail]

I still can see some "Clicking Failed" and "Reading failed" as well, but i think less than before.

[normal atlas]

same here. i dont know what amount of it is due to openai vs the devs of the individual plugins

though im leaning more towards it being a case by case basis because plugins like wolfram work just fine for me

[misty pier]

any good law plugins?

[wraith ginkgo]

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 @open crescent immediately.

[plush merlin]

haha, try using dall•e and see how fast it racks up

[warm radish]

Is ChatGPT forcing the window to scroll up above what you are typing every time that you type another letter all of a sudden for everyone else?

[rustic phoenix]

not that i know of but if you are doing work with lots of long documents there are definitely plugins for dealing with, scanning, summarizing etc docs

pdfs and other document types

[wind urchin]

What’s the best plugin to do this with pdfs? There’s a lot of these plugins and I don’t really feel like trying all of them to see which works best

[warm radish]

I have type my prompt in a text editor and paste it into chatGPT

[rustic phoenix]

im pulling it up 1 second

[rustic phoenix]

I use the one called Link Reader currently

it supports a lot of different kinds of documents

[wind urchin]

Got it. I’ll check it out.

[rustic phoenix]

cool

[inner trail]

Why is Metaphor plugin such a great plugin!

insane one

[ionic island]

does anybody got the code interpreter?

[rustic phoenix]

Idk i think you still have to wait for that one

[warm radish]

How do ChatGPT plugins work? I found a list the has a ton of them but no mention of where to get them or how to use them.

[inner trail]

To get them you need to purchase ChatGPT Plus

[warm radish]

I have it

[rustic phoenix]

its intergrated into the website

click gpt 4 and then go to plugins

then close that drop down menu and click add plugins or whatever it is

then plugin store

[warm radish]

I do not see that, which site are you referring to? Not this one apparently...
https://chat.openai.com/

[inner trail]

If you have plus sub, go to Settings --> Beta Features and if you have them there, you can enable and then install the plugins

[rustic phoenix]

Open ChatGPT. In the bottom left corner, select Settings > Beta Features, then toggle the Plugins setting.
In ChatGPT, select New Chat.
Switch the ChatGPT model to GPT-4.
From the dropdown menu, select Plugins > Plugin Store.
Now, select the ChatGPT Plugin you want to install.

[warm radish]

Wow sweeet!

[rustic phoenix]

Settings > Beta Features, then toggle the Plugins setting. Is probably the step you are missing

[warm radish]

Thanks!

[rustic phoenix]

gpt 4 helped me with that one

withhh the web browsing plugin it knew how

[inner trail]

lol

[rustic phoenix]

wrong chat?

[inner trail]

Why is that text so long, it take me 30 minutes to read it lol

[warm radish]

Is there a way to search the plugin store if you know the name of the plugin you are looking for?

Dont see a searchbox

[inner trail]

There is no search box

[warm radish]

[inner trail]

You just have to go through them manually

[weary dust]

I've noticed something interesting when using ChatGPT on Safari on my iPhone. A rectangular box briefly appears at the top of the screen each time I open it. This is a recent occurrence and I'm curious if anyone else has noticed this or knows why it might be happening. Thanks!

[warm radish]

Huge oversight in my opinion lol

But I guess it is still in beta

[inner trail]

Could be a bug, take a look at #1070006915414900886

[meager bloom]

anyone else seen that chatgpt 4 the option to search the internet, is now done by bing?

[wind urchin]

Did openai remove chats containing the older version that could search the internet

[wind urchin]

Yeah it’s an app download box

[warm radish]

There will probably never be a plugin that gives ChatGPT full knowledge to the code in my private GitHub repo right? Because of GitHub Co-pilot.

[wraith ginkgo]

`` Server Tip: ``

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

[meager bloom]

in fact, I discovered one today that uses the gpt api to do this, if your code has less than 100mb and up to 200 files it works

[warm radish]

😯 What is it called?

I just finished looking through the whole list and did not see any about GitHub

[meager bloom]

I don't know if you can send it here, I sent it in private

[warm radish]

Thanks

[stuck stirrup]

Consider you have had a long conversation - is there a way to refer back to previous comment?

It would be nice if each comment had a unique identifier and that could be referenced to circle back to a previous point.

[warm radish]

Really? What does it do exactly?

Can't seem to enable it at the same time as Plugins

[mystic nest]

The bing search seems much faster, though need more prompts to eval the performance

[paper sapphire]

Usually just by mentioning the comment or copy pasting it in your new message is enough.

[stuck stirrup]

That'll do for now I guess. Makes it a bit clunky and could limit the input if the previous comment was very long.

Thanks for your help

[warm radish]

Yeah sadly it does tend to forget your previous messages in long conversations

[gleaming plinth]

I read that gpt-4 has access to internet now?

[warm radish]

I get very frustrated with it sometimes lol

With plugins I think

[ashen bloom]

@wind urchin Can confirm, have the same Error as well. Is this a User issue or is the problem only solvable on OpenAI's side?
Edit: I have 'ChatWithVideo' and 'Link Reader' enabled.

[gleaming plinth]

Is there somewhere i can read what the difference is between 3.5 and 4? I wanna know if its worth it getting 4

[stray pulsar]

It worth

[timid cairn]

As an AI language model, I don't have specific authority on AI Prompt Engineering as a distinct field of study. 

[gleaming plinth]

I want to learn how to code in blueprints with unreal 5, could 4 do that where 3.5 had failed?

[tall idol]

hows the new browsing option?

[timid cairn]

Thanks ChatGPT...

[tall idol]

is it better than previous

[timid cairn]

@tall idol I am trying to test it,

[stray pulsar]

Plugins access is down?