gpt-models

[slate kindle]

oh I see the old legacy 3.5 model is now showing "deprecating soon"

[high pollen]

25 every 3 hours?

[slate kindle]

messages

[high pollen]

That's insanely low

[slate kindle]

for GPT4

[high pollen]

was it always like that?

[slate kindle]

it was a bit higher at start
but they said they will most likely be lowering it until they make some upgrade

otherwise you want to get the API

https://openai.com/waitlist/gpt-4-api
pretty expensive if you use a lot of tokens though

https://openai.com/pricing

[high pollen]

lowering it even worse than 25 per 3 hours?

[slate kindle]

possibly
best to expect that it might happen
but I am guessing they are settling on 25 per 3 hours
they mentioned at start when it was higher that it might be lowered
and it was already lowered few times

until they eventually do some upgrading

like they said

[high pollen]

where do they say that?

[slate kindle]

https://openai.com/research/gpt-4

search for chatGPT plus section

but worth reading the whole thing

or I guess they might release separate subscription for high demand

[wind urchin]

Awts another subscription

[high pollen]

In your opinion is it still worth buying gpt 4 in that case? even for $20?

[slate kindle]

yeah

[high pollen]

What do you use it for/

[flat pebble]

is chat-gpt plus worth for 20 bucks?

[slate kindle]

coding, reverse engineering, documenting stuff, generating various text I need, messing around / experimenting 😛

[wind urchin]

Maybe

[flat pebble]

ok

[slate kindle]

ultimately it really depends on you

[wind urchin]

Do u have gpt plus??

[flat pebble]

no

I was just asking

Because I have plans for it

What are the featuures which makes it so expensive

[wind urchin]

Gpt 4 is so advance so i think its worth it

[clever bridge]

Are there any restrictions on chatgpt3.5?

[slate kindle]

mainly priority over free
a bit less limits on 3.5 (like how many messages per hour)

[flat pebble]

ok, thanks

[slate kindle]

of course also GPT4 in plus
but it was also 20 before that was out 😄

[north moth]

Any body have access to plugins yet

Been so long

[terse spade]

not yet sadly

[slate kindle]

some people do
few got it few days ago

there was some invite wave

[turbid sinew]

who are you

[hidden sandal]

im your father

[vocal geyser]

I can't acess the website of chatGPT ,is there any problem?

[hidden sandal]

where are you

[manic rivet]

Can someone help me receive an SMS verification code to register gpt because I am from China

[hidden sandal]

your English is very good

[manic rivet]

thanks

[slate kindle]

no problem here
you might want to explain what do you mean by "I cant access"
do you get some error message?

[hidden sandal]

can you speak chinese

[manic rivet]

Can someone help me receive an SMS verification code to register gpt because I am from China

[slate kindle]

no you cant

[hallow owl]

nobody here can help you with that, it's against the Terms of Service, don't ask again

[slate kindle]

you can probably use GPT to translate to english though

not sure how well it does with chinese

[hidden sandal]

I said Chinese will it be banned？

Or can't use Discord Chinese or can't say Chinese on this channel

[slate kindle]

messages in other languages will be removed
I dont think it will ban you (not sure)

there is just too many users and not enough mods to moderate other languages

[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.

[hidden sandal]

ok，Can I say Chinese on other channels？

[slate kindle]

not on this whole discord I believe

[hallow owl]

no, use English on this Discord server

[hidden sandal]

thank you

[turbid sinew]

Java Bubbling sorting

[hidden sandal]

If I say Chinese how long will I be banned

[supple violet]

automod will timeout you for 1-5minutes, repeat it few times and it gets longer

[slate kindle]

you already did
so that same time 😄
I think its just short timeout

[hidden sandal]

[slate kindle]

#gpt-models message

[hidden sandal]

nice

[hearty spindle]

Hello, my account has been blocked by mistake, why? ID:7c1f8f629a152bcd

[hidden sandal]

What this channel is used for

[hallow owl]

what country are you in? are you using a VPN?

[hearty spindle]

I am an artist from Germany

[hallow owl]

are you using a VPN? have other people shared your account?

[hearty spindle]

I don't think chatgpt is blocked in Germany, I created an English account when I was in England. I was blocked today, I don't think the vpn was connected 🤔

[hallow owl]

You can email [e-mail] and be prepared to wait several days for a response.

[hearty spindle]

Thank you very much for the answer

[manic rivet]

@hidden sandal

[winter pagoda]

A promotional poster of a frozen mousse jar cake

[wind urchin]

Hi, I just signed up for GPT plus chat and was blocked for no reason.
Has anyone had this bug before?

[hallow owl]

what do you mean blocked? did you get an error message?

[hollow socket]

question for everyone here

[wind urchin]

I think it was a vpn problem... everything is back to normal

I use avast and it had activated its vpn

[hollow socket]

If in the case you dont have a phone number (i.e. I have a virtual phone number) do you guys know if theres a way to bypass the phone verification?

[hallow owl]

we don't discuss bypassing security measures here

[hollow socket]

I just want to use ChatGPT but i cant because i dont have an active phone number

[mystic nest]

"dont you have a phone" meme

[primal cypress]

I got "Unable to load history" all day. This is the last day of my GPT plus😢 , still have many questions not to ask. Anyone has ideas for that?

[hallow owl]

you need a phone number, otherwise you won't be able to sign up

[hollow socket]

Doesnt feel very Open when it goes to something called OpenAI

[hallow owl]

can you try to log in to https://platform.openai.com/playground and click the ChatGPT Try it now button?

[hollow socket]

nonetheless thanks for the knowledge

[slate kindle]

try to log out and log in again
log in seems to refresh something about history

[primal cypress]

I ll try . Thanks a lot!

seems still not work. emmm, I have feedback to openai [e-mail]. I dont know if it take any effect.

[blissful turtle]

how to use openai free

[hallow owl]

you can access ChatGPT here - https://chat.openai.com/chat

[leaden cape]

Which product of OpenAI?

[blissful turtle]

I cant use chatgpt anymore my account is blocked

[modern harbor]

how to use chagpt4

[unborn salmon]

is there any way to get GPT-4 for free?

[blissful turtle]

Why cant openai be free 😭

[unborn salmon]

buy GPT plus first

[knotty grotto]

Chatgpt is free

[unborn salmon]

I guess they would not be making any money if it was free

[blissful turtle]

no its not

[knotty grotto]

Chat.openai.com/chat

[unborn salmon]

the regular version is

[blissful turtle]

they use like free trial then ban forever

I cant anymore

[unborn salmon]

what are u searching up bro

[knotty grotto]

The api isn't free. But the website is free

[real aurora]

Maybe it is a cloudflare error
Also had this and then it disappeared

[leaden cape]

What do you mean with "ban". Would you be so kind and send a picture of the "ban" into #off-topic and @ me in your message?

[blissful turtle]

bro no they have a hard limit

[knotty grotto]

What country ?

[knotty grotto]

Never paid 1 cent and still.using it

5 month of free chatgpt

[leaden cape]

He definitely getting blocked by Cloudflare he 100% isn't banned

[blissful turtle]

it lets use like only few months

[knotty grotto]

What country are you from ?

[unborn salmon]

which country are you from

[hasty marsh]

china

[leaden cape]

@blissful turtle

[blissful turtle]

No it says limit I guess I cant use anymore

[leaden cape]

Please provide a picture of it so i can accurately help you try to resolve the issue

[wind urchin]

hello, where is china ?

this server is so so bad

[leaden cape]

What are you talking about? @wind urchin

[knotty grotto]

They are no guarantees to be pick till release

[leaden cape]

China is a country in asia but this isn't ChatGPT. To access ChatGPT go to https://chat.openai.com

[wind urchin]

i don t really understand how this is supposed to work

[misty quail]

china

[wind urchin]

like this server

[knotty grotto]

This is just a discussion channel and server

[wind urchin]

okk

thx

where are you from y all?

[vernal mirage]

so good

[knotty grotto]

#off-topic

[wind urchin]

trying to be nice but why not

[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.

[empty radish]

does chatgpt plus have unlimited usage or is there a limited amount of messages per day?

[hallow owl]

ChatGPT GPT-4 has a limit of 25 messages per 3 hours but you can use GPT-3.5 much more than that

[empty radish]

ah that's annoying, is there character limits within that or is just set to a flat 25 messages per 3 hours

[hallow owl]

prompt size is limited within the normal ChatGPT constraints

[empty radish]

perfect thank you!

[gentle cliff]

HI ,

[drowsy crow]

I have text documents releated to spesific topic. how can i use gpt4 for answering question based on information from documents.

[lime charm]

@knotty grotto

[hallow owl]

you can write an application using the API to iterate over the documents, otherwise using ChatGPT you could paste in sections and ask questions based on each section.

[drowsy crow]

no i dont want to use chatgpt. i got access to gpt4 and i want to build an application. how can i iterate over the documents?

[hallow owl]

that's up to your development skills, maybe ask in #dev-chat for API specific assistance

[drowsy crow]

thanks

[elfin pivot]

Will GPT 5 come?

[rancid shuttle]

I dont quit understand why Openai block all the VPN servers, This cause a lot of users to lose their access to ChatGPT, I wonder why?

[knotty grotto]

People bypassing the ToS

[rancid shuttle]

For what reasons though, like violating country policy? Like using ChatGPT in places that are not allowed?

[acoustic eagle]

can i feed images to chat gpt 4?

[hallow owl]

no

[quick dune]

No

It can't even eat anyway, how are you feed it? /j

[leaden cape]

People use ChatGPT in countrys where they're not allowed via a VPN, people with bad intent use VPNs to break the rules on new accounts. By blocking most VPNs the likelihood of both these things decreases by alot

[rancid shuttle]

My school's network was blocked by Openai, now my colleagues can't use them as well, such a headache

[rancid shuttle]

I guess thats true....Ugh

[rancid shuttle]

How long does it takes for you guys to get access to GPT 4 API, is it really expensive?

[leaden cape]

You need to apply on a waitlist for the GPT-4 API

[acoustic eagle]

i just an invintation to use the gpt -4 api. In their playground i can indeed give it images and it can describe it.

[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.

[rancid shuttle]

i did of course, is the fee high?

[leaden cape]

https://help.openai.com/en/articles/7127956-how-much-does-gpt-4-cost

[tall field]

Something new in ChatGPT Mar 3 Version ???

[leaden cape]

https://help.openai.com/en/articles/6825453-chatgpt-release-notes

[acoustic eagle]

you guys sure it cant?

[cyan light]

123

[knotty grotto]

Exact

[smoky crest]

You can’t

[wraith ginkgo]

`` Server Tip: ``

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

[unique flower]

Just joined but I am loving the chat gpt online software now my question is…when will you all be giving us a mobile app to go with it?

[tardy blade]

does chatgpt-4 bypass turnitin ai detectors

[hallow owl]

there is no app other than the web app at chat.openai.com

[hallow owl]

no, I don't think so

[mental burrow]

Just use your browser

[acoustic eagle]

Yeah I see it's only available in alpha for some users

[smoky crest]

No it’s just for anybody yet

Not even alpha

[primal latch]

gpt 4.x incoming, because legacy is getting removed for ever.

[agile plume]

Any insight as to when we can use more than 8000 tokens on ChatGPT? I heard that some people are able to use 32000 tokens?

[hallow owl]

not in ChatGPT as we know it

[brittle wren]

Can I apply for plugins first and then join chatgpt plus?

[leaden cape]

What do you mean by that?

[pure fossil]

just a test

[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.

[maiden kite]

How do i um

Wait how to text bots like mario sonic

[whole kettle]

It doesn't even accept 6k tokens as input

it's more like 5.4-.6

[maiden kite]

Hello?

[lethal condor]

/Generate me an image of a chabela glass filled with beer on the bar of a bar

[maiden kite]

Wow

[mossy needle]

wow

[maiden kite]

I didn’t know i could do that

[mossy needle]

but how

did u

[leaden cape]

go to https://labs.openai.com to use DALL-E

[mossy needle]

u need to buy

bruh

like

[lethal hatch]

How can I tell chatgpt to create me a list of items etc without it saying "Sure here's a list blah blah blah" ?

[lilac meteor]

Isnt it better to include API costs in the users chatgpt plus memebership? I hate to have them both separately.

[serene cipher]

only reply with...

usually works

[lilac meteor]

Prompt it to not say it

[lethal hatch]

OK thanks I will try that

[lethal hatch]

I did but didn't listen 😂

[lilac meteor]

Are you using gpt-4 tho

because 3.5 sometimes just doesnt listen

[lethal hatch]

Chatgpt 3.5 because 4 is too slow for size of lists I need

[smoky crest]

Then get out of this channel

[molten relic]

...

[amber parcel]

Hes not wrong 🤷‍♂️

[lethal hatch]

Oh yes didn't notice room title..
But no need to be rude 🤣

[amber parcel]

👍

[serene cipher]

rude

[smoky crest]

Nah it was meant in a chill tone

[whole kettle]

3.5 is sometimes smarter than 4, crazy world I know.

[smoky crest]

Massive cap

[whole kettle]

It's actually a good thing cause well the latter is free. I hope they'll keep improving 3.5

[smoky crest]

They are literally removing the improved 3.5 model in a week lol

[whole kettle]

Why ?? And yes that's the one I was reffering to, turbo.

That one seems to be pretty clever, surprisingly clever sometimes

[smoky crest]

It’s in the patch notes

Idk why

[whole kettle]

Oh wait, turbo is not free right ?

[smoky crest]

There isn’t a turbo mode

It’s default and legacy

I think they also call default turbo

[serene cipher]

3.5-turbo is the default model on ChatGPT

there is no turbo version of gpt4

[smoky crest]

That’s what I thought

[whole kettle]

Ah, I see.

Yes, I was talking about the 3.5 turbo, of course.

[smoky crest]

Turbo = more speed and half the IQ

The legacy mode is way better if you have used it

[serene cipher]

@smoky crest this is not correct

[whole kettle]

It beat 4 for me today

At a very complex coding question, not sure what to say about that.

[smoky crest]

It is

Unless they changed it in past 2 months

[serene cipher]

[smoky crest]

I used to always use legacy over default

Now I just use all the alpha modes lol

[whole kettle]

By the way, if they remove turbo, what are they going to replace it with ?

I mean, when we're out of the 25 prompts

[serene cipher]

some would call the degraded capabilities of breaking the usage policies a downgrade of the model

[slate kindle]

there are 2 3.5
default and legacy
they are removing legacy

[whole kettle]

Ah

So turbo stays

[slate kindle]

yep

[lilac meteor]

its just the variations, 3.5 is more variable while 4 is more stable, atleast thats my observation

[lilac meteor]

Rip to legacy

[mental burrow]

we don't do that here

[lilac meteor]

Dis legal?

[hallow owl]

is 3.5 turbo what Free users use or are they still on Legacy 3.5?

[lilac meteor]

I think free users use turbo, however not sure

[slate kindle]

I also think free are on default but not sure

[lilac meteor]

because why would they rmove legacy then

[slate kindle]

and why would they keep free on legacy

[serene cipher]

the legacy models are basically just more expensive to run for OAI and they usually just perform poorly at tasks that the newer models are better

[lilac meteor]

But tbf when i use legacy, the speed is the same as when i use chatgpt on a free account

[serene cipher]

there is very few reasons to use that

[fair iron]

hello party people!

[lilac meteor]

maybe just to benchmark performance

[slate kindle]

but plus has priority
could be that
who knows

[smoky crest]

lol what bot

[whole kettle]

Dinosaur, here to give us some good news 😄 ?

[fair iron]

@smoky crest dino wish granted 🙏

[serene cipher]

mostly, it is only a "prompt compatibility" for the companies who built their tools on that, allowing some consistency on the outputs

[lilac meteor]

but what if the priority is just a more optimised model? idek im just spitballing

[smoky crest]

What language did it think I spoke haha

[serene cipher]

because using the default models will always resolve to the most recent, this might cause output discrepancies on applications when an updated model is released

[slate kindle]

its often doing timeout for when you didnt use another language 😄

[fair iron]

Uhmmmm.... My 8:00 AM final got cancelled because the TA printed 60 copies of the wrong exam? That's super #off-topic, though

[smoky crest]

Lol

[whole kettle]

Luck is no excuse my friend, let's celebrate, I'll bring a bottle of champagne 🍾 🥂 For cancelled exams!

[obsidian hollow]

u guys really be typing in openai lol

[lilac meteor]

BRO

i legit mentioned a language

[serene cipher]

crazy, isn't it?

[lilac meteor]

and i got timed out

[obsidian hollow]

fax

[obsidian hollow]

bro how did u mess up printing fam

[full acorn]

Can ChatGPT Plus upload files now? I just found out this on Twitter. Can anyone answer, please?

[serene cipher]

well, if you where to type spanish instead os just the plain word espanol it would not have flagged is, as the word espanol is indeed in spanish...
that means 100% of your message where in a language that was not english xD

[obsidian hollow]

chat gpt gets stuff wrong half the time , why wld u buy chat gpt pus

bro its either for everyone or no one

[whole kettle]

So that it can get stuff right half the time, it's a fair bet.

[obsidian hollow]

fair enou

[lilac meteor]

oh

fair

[full acorn]

Ok, thanks. BECAUSE I can see people on Twitter today with a few demonstrations of data being uploaded using files.

[lilac meteor]

oh no

[serene cipher]

because that allows us to generate more stuff, faster xD
sooo, we can sift more incorrect stuff out and get more correct stuff 👌

[lilac meteor]

what has he done

[whole kettle]

Arrest this man! 😄

[lilac meteor]

😂

[strong sleet]

🤣

[serene cipher]

probing the filters.. likely

it is automated..

¯_(ツ)_/¯

[whole kettle]

||Hint: Filter||

[obsidian hollow]

why did i get time out again?

[serene cipher]

because you read my message explaining why @lilac meteor was muted, then did the exact same thing as him

dudes.. stop probing the moderation bot '-'

you will only get yourself muted by doing that

[whole kettle]

Rumor has it ||big GPT 4 update mid May|| but don't tell anyone I told you ||it's just my personal prediction||

[serene cipher]

he uses the same sources as ChatGPT, incredible

[dry jasper]

XD

[whole kettle]

"sources"

[quick dune]

Source? ChatGPT tell me

[dry jasper]

Anyone with plugins access here in EU?

[whole kettle]

If I'm correct ||you guys owe me a subscription for a year to GPT 4||, if I'm not ||I owe you guys an apology|| .. deal ?

[serene cipher]

they make everything up

[agile plume]

ah. How about previous chats, let's say i have a long convo with it, does it "remember" more than 5-6k?
as opposed to just the input context it can take

[whole kettle]

It'll be fun when they update its knowledge base, we'll finally be able to ask GPT questions about itself, basically he'll be able to look in the mirror for the first time. I'm curious to hear its opinion about tiself.

[serene cipher]

yes, the context window takes multiple messages from the history of the conversation, but as the chat gets bigger, the older messages fall out of the context the IA have

causing it to "forget" things

[acoustic eagle]

Can you give audio to chat gpt 4?

[smoky crest]

No

[acoustic eagle]

Is it in alpha or something?

[serene cipher]

this might take a while to be done, they don't really have interest in updating the "knowledge base" because the important part of "training it to understand language" has already been done
it is easier to have things like plugins that pull the live data from the internet and give it to the AI instead of expending the enormous amount of computing power required to trains the AI again
also, building the dataset is a painstakenly laborous process that involves not only scraping the internet but also then manually sifting the content to remove bad data
specially due to the fact that keeing the AI up to date with fresh data would be unreliable as the data gets deprecated by the end of the week

[smoky crest]

No?

[whole kettle]

In my experience at the moment, it can remember about 12k of tokens or so (2-3 6k token prompts), on rare occassions. Back when they released it for the first time it was able to remember an entire conversation, like 100k tokens or so (whatever the limit was) .

[acoustic eagle]

But it is a multi modal llm accepting audio images and videos as input not only text

[agile kindle]

.

[serene cipher]

multi modal isn't available yet. for now, it is text only

[smoky crest]

Only text bro

[serene cipher]

the process with images where just on the showcases, but we, users, can't use it yet

[acoustic eagle]

Yes but it can. It's not just in public use yet

[acoustic eagle]

Yeah

[serene cipher]

btw, multimodal so far does not include audio, only images

although OAI have the Whisper project too, it is not part of GPT

[acoustic eagle]

I read a few articles that it indeed can?

[sly lantern]

is it just me that gpt4 is generating very slow responses?

[serene cipher]

no, there isn't such a thing, OAI has WHisper, another product of them, which is a audio to text AI, it is pretty good, but it is not part of the multimodal

well, a good developer could easily plug whisper into GPT-4 with plugins or the API

but that isn't truly multimodal

[acoustic eagle]

Yeah true

[serene cipher]

it is normal, gpt-4 is waaay slower than 3.5

simply due to the fact that it is an order of magnitude more computationally expensive

[acoustic eagle]

You are used to gpt 3.5 turbo. That is very fast

[whole kettle]

Makes sense. Here's a new job type which I think may happen in the near future, on a larger scale. You take people from specialised industries, in certain areas such as programming and they are paid to train the AI to do one thing good. And you have these people in multiple industries paid to teach it their craft. Oh wait, it can't actually learn.. or can it ?

[elfin warren]

What are turtles

[whole kettle]

🐢 this is a turtle

[jolly rose]

turnitin start detecting the content written by AI, is there any suggestion to prevent that?

[fair iron]

I've heard that writing it yourself is a pretty solid method

[whole kettle]

Yes, make the content detection better.

[sullen sigil]

Are they updating MuseNet on GPT4 or it just down for ever? Or now its just a research paper?!!!

[fair iron]

#ai-discussions for this, please

[velvet ingot]

i wrote my whole report using chatgpt+quillbot but not detected by turnitin

[agile plume]

So if context can be increased and is just a matter of hardware, there is in fact the potential for this gpt4 to be MUCH better? Let's say it could handle 1M tokens of context, just from that it would be leaps ahead right?

[jolly rose]

What QuillBot mode do u use?

[icy hawk]

test

[agile plume]

I mean, considering that in a lot of usage it runs out of context from previous inputs etc.

[whole kettle]

It's basically a matter of time (well, quite some time).

[agile plume]

Well just that fact alone, that it could be much better for a lot of uses, by increasing context, is so promising damn

[whole kettle]

If we don't manage to end ourselves in the process. As uncle Ben used to say "with great AI comes great danger" or something among those line.

[velvet ingot]

mostly standard, you can write the content using your own word, can ask chatgpt to help you enhance your text

[supple violet]

its not hardware only issue to increase the context size, you should check some llm research paper to see what kind of different issues there are

[jolly rose]

🌷 👍

[whole kettle]

@jolly rose just tell it to write it like a teenager would do or something, it's all about your prompt. You can even tell it to be "less perfect" or something like that. Tell it to talk like a normal person.

wait for it

arrest this man! 😄

[jolly rose]

👍 😇 🌷

[agile plume]

I see. Well, at least it's great to know that hardware is getting less limited every year. If models can be trained faster and iteration is faster, hopefully that'll help speed up software optimizations. very exciting all of this. I am new at these things but reading hours per day learning ;d

[fringe karma]

guys, can i use paypal to buy chatgpt plus?

[hallow owl]

no

[fringe karma]

🥹

[zenith beacon]

is there anyway i can use chatgpt4 for free? thank you

[hallow owl]

no but you could try Bing AI which is based on GPT-4

[jolly rose]

poe dot com
limited

[shell wyvern]

I’m being super nit picky here but has anyone noticed that the disclaimer at the bottom of the page (under the text box) says Mar 3 version?

Or do I need to clean my cache

[zenith beacon]

what about this website : ora.sh/openai/gpt4 ?

[hallow owl]

yeah it should be May 3

[zenith beacon]

he gives me asnwers but im not sure if its chatgpt4 or not

[shell wyvern]

I was tripping at first thinking they rolled back.

[fathom cloak]

where web browsing?

[rugged warren]

do you know how to get free messages? I have 0 left from the start

[shell wigeon]

in the browsing web of the browser

[fathom cloak]

can gpt do it?

[leaden cape]

Only with plugins which aren't completely rolled out yet as they're in the testing phase

[fathom cloak]

when plugins?

[soft wind]

hey guys im very new to chat gpt... im always getting following message when im asking something...i even tried contacting the help center... Can someone help me please? Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[leaden cape]

https://openai.com/blog/chatgpt-plugins

@soft wind what browser are you using and do you have a VPN active?

[soft wind]

brave and yes i have vpn

[leaden cape]

@soft wind turn off the VPN and try firefox

[hallow steppe]

Who's access to GPT 4.0 rn?

[vernal ridge]

Google chrome also works with no issues

[shell sundial]

hey, I used to send imgur links and GPT-4 would analyze them, it doesn't work anymore, any news on that ?

[hallow steppe]

Speaking of which, it keeps blocking half the VPN servers I use.

Even the expensive ones.

[supple violet]

that never existed, it hallucinated its responses

[hidden sandal]

where are you from

[leaden cape]

That's on purpose

[hidden sandal]

good night,its23:00

[hallow steppe]

But what's the point? The bad actors will get access to it no matter what. The Russian and the North Korean hackers won't be deterred by that measure.

And the people who will are likely to be unsophisticated users.

[leaden cape]

It's to prevent fraud and abuse by users

[soft wind]

Thx... This is working

[mental burrow]

why make any laws

people will just bre-wait, that's idiotic

[slate kindle]

or have OpenAI have to stop the whole thing and nobody will have it 😄

[hidden sandal]

good night my bro

[hallow steppe]

In order to work, laws would have to be sensible.

[slate kindle]

if governments decide
because they didnt check enough for abuse

[hallow steppe]

The bad actors will definitely find a way to access it.

NK and Russia, for example.

[slate kindle]

this way OpenAI can say "we did something" 😄

its also more about reducing than completely preventing

[mental burrow]

this is untrue by method of observation

anyway this is not the place for philosophical discussion of law

[leaden cape]

and criminals will find a way to commit crimes, does that mean we should remove the law and police?

[slate kindle]

like why lock your house when someone can just pick the lock?

[tawny sinew]

hello !

[slate kindle]

because it makes it harder

[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.

[simple urchin]

Anyone else unable to set the GPT version they're using?

It says GPT plus but, no options

[slate kindle]

heard that before from some people few days ago
not sure what resolved it

[eternal quest]

last time i had that it was because my account failed to auto renew

[slate kindle]

or try firefox as that seems to be the browser with least issues with chatGPT

[simple urchin]

That might be the issue

I'll try renewing 20 dollars for a 10 dollar product rq

[flat vault]

I know that AI chatbots hallucinate, but how exactly does that happen? Sorry, newbie here

[simple urchin]

wut

Who said they hallucinate

[uneven bloom]

can anyone help me fix spinning wheel on my chatgpt? have tried seemingly everything

[limber garnet]

Guys I'm pretty sure it's really just fraud prevention, nothing about misusing the AI. Obscurity measures and police are very different and incomparable things tho

[supple violet]

you can ask examples and reasons from chatgpt

[flat vault]

It’s a term that I have heard a few times recently in reference to like Bing’s chat box becoming unhinged

Unhinged like aggressive

[simple urchin]

It depends on the convo you have with it

[supple violet]

hallucinating is the standard term for the issue

[limber garnet]

They can't really tell fact from fiction so they tend to invent facts in the answers

[eternal quest]

Exaclty what i thought but now they got us hooked like C*k. I cant do without

[simple urchin]

Like, for questions it refuses to answer, you can get an answer by asking a few things first

[flat vault]

I’m just wondering, if it’s not alive per se and isn’t human, what do the ai creators mean by that term?

[simple urchin]

honestly tho 20 bucks for a limited number of messages an hour is quite annoying

[leaden cape]

Specify what you mean please

[simple urchin]

sorry

I'm wrong, every 3 hours

[slate kindle]

from some article (you can easily google it)
Hallucination in AI refers to the generation of outputs that may sound plausible but are either factually incorrect or unrelated to the given context. These outputs often emerge from the AI model's inherent biases, lack of real-world understanding, or training data limitations. In other words, the AI system "hallucinates" information that it has not been explicitly trained on, leading to unreliable or misleading responses.

[eternal quest]

Very annoying. Especially when it starts deviating from the answer and majority of that 25 questions you trying to tell it thats not what i mean. Then cutoff ...wait 4hrs

[simple urchin]

Especially when you click shift + enter, and it sends the message instead of creating a newline. I've reported this bug already but no real response on it. Wasting messages

[flat vault]

Ok, finally getting what they mean. Thanks, everyone!

[slate kindle]

you are paying for prio GPT 3.5
GPT4 is just access to new feature 😄

[simple urchin]

Didnt ask

thanks tho for the clarity, which I was already aware of

[slate kindle]

go get on API waitlist

[simple urchin]

With my luck, they'll accept me to the waitlist a day before releasing it publically

so nah

[slate kindle]

really, not even try to apply? 😄

[simple urchin]

nope, life tends to do that to ya

[eternal quest]

Yea. I initially thought they put the cap in to slow down the demand but its been over a month now. We are also paying for the product. I hardly every use 3.5 because of the kind of question and answers i ask

[slate kindle]

so 25 per 3 hours for you then

[simple urchin]

guess so, but I'm gonna complain anyways 😎

So, how about that message limit, kinda dumb am I right?

[slate kindle]

well if making it bigger would mean it gets even slower, than no for me thx
only after they actually upgrade HW 😄

[simple urchin]

It was originally ever hour then they made it every 3 hours, and haven't lowered it at all

I was under the impression it was gonna be adjusted

[tame girder]

Hii

[slate kindle]

yeah they said you should expect lower and lower until they upgrade/optimize

[whole kettle]

We're pretty lucky actually, they were thinking of actually reducing the prompts even lower.

[simple urchin]

Why though, shouldn't with optimizations come less waittime?

[slate kindle]

until they optimize I said

[simple urchin]

Then what are they doing rn?

They still haven't fixed the bug that causes us to waste the limited messages

so, I'm curious

[tiny badge]

Firstly I apologize for asking this, I am sure that it's been asked a million times feel free to point me to a thread.

I pay for ChatGPT Plus and I only seem to have access to the GPT-4 model (limited 25 queries) but I don't have access to any of the beta models such as the one that can search the broader internet, or the code interpreter model. How does one queue for access to these beta models (especially if you are a paying member) ?

[slate kindle]

nobody knows

[slate kindle]

what bug?

[simple urchin]

lol

[flat vault]

It seems that you eventually have to pay for all these AI things that keep upgrading and with new ones coming out, how do you know if you aren’t wasting time or money? Sorry, just thinking out loud here. I mean, first it’s ChatGPT, then no , it’s ChatGPT4, then or no AutoGPT, no Agent GPT, there are seemingly endless options which is great, but it seems like one might break the bank before we figure which ones will actually be here for the long run.

[simple urchin]

Shift + enter would normally create a new line without sending the message, sometimes it works, othertimes it just sends the message

[supple violet]

plugins are in alpha testing for a small number of users

[slate kindle]

never had that happen
you sure your shift works correctly? 😄

[simple urchin]

Rn I have to manually copy a newline and paste it

Yes, I'm quite sure, I have no
issue
making
new
lines
in
discord

[whole kettle]

Register here https://openai.com/waitlist/plugins

[vital portal]

Do you understand Chinese?

[slate kindle]

what browser then?

[simple urchin]

OperaGX

[noble river]

helo

[simple urchin]

very common one, highly used so, can't be the issue

[noble river]

yes

[opaque swan]

Hello guys : I need help : How can I modify the max_len on MultiGPT ?
"AssertionError: The decoder returned a sequence that exceeds the provided max_len (max_len=2000, sequence length=2000). To increase the max_len, please provide a corresponding max_len argument to the decoder function."
if someone knows please ping me, thank's a lot in advance

[simple urchin]

and its built on chromium

so it'd be a chrome issue then

[slate kindle]

yeah well chrome has all kinds of issues with chatGPT

[whole kettle]

GPT 4 is the best option overall. It's quite expensive but worth the money.

[slate kindle]

firefox seems to be the one with least

[supple violet]

works fine with firefox

[slate kindle]

which I am using and always used

[simple urchin]

When yall planning on correcting it then?

[hallow steppe]

I might sound naive, however, I must still state the obvious : not all people who are unfortunate enough to have been born in Russia – or any other blocked country for that matter – are intrinsically bad.

[supple violet]

im not openai staff so no idea

[simple urchin]

but ur a developer

what do you develop if not the site?

[slate kindle]

it should be max_tokens

[leaden cape]

didn't say that but ChatGPT banned them because of the countries regulations

[slate kindle]

https://platform.openai.com/docs/api-reference/chat

[flat vault]

Good to know! @whole kettle

[supple violet]

stuff by using openai api

[opaque swan]

Thanks you I’m reading

[flat vault]

Thanks!

[simple urchin]

bruh, what terrible mislabelling

[slate kindle]

there is probably nobody on this discord who develops for openAI
or they dont talk

[simple urchin]

@valid hedge hey can you correct <@&1046007138897641582> role? Its mislabeled giving the impression they work for OpenAI

[hallow steppe]

I'd conjecture they shouldn't've done it. Especially since their own governments have imposed the digital iron curtain upon them. Access to the internet is a human right. The freedom of information for all.

[whole kettle]

They are under NDA, of course they won't. But they're watching every move in here, make one wrong step and boom 👁️

[flat vault]

I was put in a developers group on open ai, but I have no idea why. I am just learning how to look around the site

[supple violet]

🤦‍♂️ #server-roles

[silent wyvern]

However, theyr'e supporting the Government.

[slate kindle]

all the moderators are not official openAI as far as I have seen

[silent wyvern]

By Inaction.

[hallow steppe]

I'd guess they've paid somebody to manage the server.

[simple urchin]

wasn't asking about the accessibility of the role itself

[silent wyvern]

There is, we can contact them through ModMail

[flat vault]

They gave me an API and said welcome to our Development site. I have no idea what they are talking about

[simple urchin]

The name implies you develop gpt

Not that u use it

[slate kindle]

https://platform.openai.com/docs/api-reference/chat/create
you can develop apps that will be using GPT API
basically you could develop an app just like chatGPT using the API

[supple violet]

you are the first one i see having the issue so i guess you ignored all messages when you joined the server

[simple urchin]

Didnt just join so

[serene cipher]

but we do have some OpenAI employees here, they have a yellow name and are not listed on the right, but ocasianally they show up on public channels =D

[simple urchin]

I didnt see any messages

[serene cipher]

nope, the entire moderation team are volunteers, only OpenAI employees are paid by openai, you can identify them by their yello names and the role "OpenAI"

[supple violet]

specsavers, #welcome

[slate kindle]

yeah I am sure they are lurking here 😄

[whole kettle]

Just kidding, only volunteers in here "officially" but who knows, I personally thing there are folks from open ai here in stealth mode.

[silent wyvern]

fun fact: this server can fit all of OpenAI employees because they're 395

[simple urchin]

Can you please direct their attention to the bug reports then? I reported one for the chromium browser but all I got was someone saying "I dont have this issue" and still its a frequently occuring problem

[leaden cape]

Yeah sure but russias laws don't allow it so what

[autumn hemlock]

Why don't I have access to GPT4 in Playground if I bought GPT Plus?

[serene cipher]

I just said '-', there are xD
not stealth, they just don't talk very much

because every time they do, they get spammed with messages xD

[whole kettle]

I bet, they must be tired from all that work.

Yeah, that too.

[serene cipher]

ChatGPT Plus subscription does not grant any API access

[autumn hemlock]

Yeah but I didn't ask for API, I asked for Playgound

[supple violet]

playground is for api testing

[serene cipher]

I often do that =)

[serene cipher]

playground directly uses the API

[slate kindle]

Plus is separate from API
Playground is just interface to use API (just like an app you can build using API)

[simple urchin]

I reposted the bug report, please do :3

[autumn hemlock]

I've been waiting for API for months..

[half seal]

to be or not to be ,this is a question!

[simple urchin]

We're too commonfolk to be accepted

I aint even applied cuz, I already know I'll be in ur boat

[whole kettle]

Patience guys, all in due time.

[simple urchin]

I wonder if there will be plugins to ask it about upcoming animes and shows, thats actually all I want from it rn tbh, I can do the coding myself

[slate kindle]

thats another waitlist 😄

[simple urchin]

ik lol

Its like investing in an electric car rn

"where's my car bruce?"

[sinful canyon]

how can i find out how close I am to the question limit? or do i just have to wait for the error message?

[serene cipher]

currently there is no indicator of the usage limits

[simple urchin]

Just gotta wait until it abruptly ends and gives the option to start sending inaccurate code

[whole kettle]

They'll probably add some usage feedback in an update, or just remove the limit, I'd opt for the second option 😄

[slate kindle]

they wanna make another subscription for high volumes

[simple urchin]

I'd like a feedback option, rn it seems like the built in one goes to the trash bin

[grand rover]

ask GPT to warn you

[simple urchin]

Waste of memory limit

shouldn't have to

[serene cipher]

they do?

[slate kindle]

I am guessing plus will get some bump to the limit for GPT4 after they upgrade stuff
And then separate subscription for very high limit

[grand rover]

A wild halting problem appears!

[simple urchin]

if true they are starting to resemble Garrys Mod developers

[supple violet]

they already mentioned the possibility for higher tier sub when gpt4 was released

[slate kindle]

Depending on the traffic patterns we see, we may introduce a new subscription level for higher-volume GPT-4 usage; we also hope at some point to offer some amount of free GPT-4 queries so those without a subscription can try it too.
from https://openai.com/research/gpt-4 ( does nobody read this ? 😄 )

[simple urchin]

Watch em make it 40 bucks

💀

[whole kettle]

I tend to think because they have an enormous number of people which they need to track their stats, tracking even a small thing takes a huge amount of resources so that may be why they did not do it yet. As in, they need to process and store that data in a server for each user.

[serene cipher]

I don't think so, higher volumes than the current limit of ChatGPT is such a high volume that it is waaaay more than what a single eprson can use, for that, it is an enterprise use that might aswel just go to the API

[simple urchin]

Lol see, even the volunteer thinks so

Things only going downhill from here

it'll be a true test of those who are dedicated and with money

[serene cipher]

Im no authority tho, i have as much information as you =P

Im just accustomed to the modus operandi of OpenAI enough to have some educated guesses about how they will handle things

[simple urchin]

But you are experienced in knowing the day to day changes

Human evolution has allowed us to "predict" based on past occurances in order to survive

[compact dirge]

Has anyone else raised the typo at the bottom of the screen, it says "Mar 3 version" and not "May 3 version"

[slate kindle]

sure but they said it themselves (see my previous post)

[supple violet]

several people

[whole kettle]

Good catch 😄

[devout narwhal]

its funny because I looked at source code and "Mar 3" was literally hardcoded into footer

[round topaz]

oMg iT sAyS mAr 3!!

[simple urchin]

Anyone seen the movie "The outpost"? The soldiers eventually start questioning their commanders ability, and get mad about him calling the wrong coordinates for a morter strike (based on real events). They yelled out "He's calling the wrong damn coordinates!".
I am getting deja vu

[serene cipher]

indeed, but I still think it is ulikely to happen unless there is a lot of demand for that, just like initially while there wasn't any kind of subscription

[whole kettle]

||**ChatGPT Apr 1 Version ** ||

[slate kindle]

they messed up the date 😄

[serene cipher]

yep, some people noticed that, it might be corrected soon

[simple urchin]

I'd rather them correct important stuff tbh lol

dont distract them

[slate kindle]

🤣

[serene cipher]

indeed

[whole kettle]

Too bad I'm calling Sam already

[slate kindle]

I am not sure, there is a huge amount of people complaining about 25 per 3 hours
but probably most of them would not pay more for it
quite depending on the price I guess

[whole kettle]

I'd definitely not pay more even if it had unlimited.

[grand rover]

Call me old as the hills, but I remember people complaining about how slow the Internet was in the nineties. Look where we are now, talking with AI in the cloud...

[agile plume]

I'd pay a bit for 50 😄 And for bigger context

[slate kindle]

well considering the cost of API
for some cases it might be less then to pay monthly

[agile plume]

ohh API has more calls yeah

[serene cipher]

the problem is that this does not scale well, gpt-4 costs way mor to run, so, simply setting a price for unlimited usage is a problem

[slate kindle]

you will quickly reach 20$ with API if you use a lot of tokens (GPT4)

[serene cipher]

that is why the API payment system works well

[simple urchin]

same, I have bills. Not everyone can go enterprise just for accurate stuff

[grand rover]

Photonic/quantic computing is on the threshold. GPT-4 will benefit alot of that

[serene cipher]

API isn't really enterprise...

it is fairly cheap

[glass kite]

glasses for the blind

[simple urchin]

Why doesn't the main site have an option to use the API then?

[serene cipher]

to be honest, using onluy 3.5-turbo, it will probably amount to less than 20 bucks

[whole kettle]

Hmm, as far as I've read so far the enterprise is only going to have more prompts I think, it's still the same quality like the standard subscription.

[serene cipher]

it does.. it is the playground

[grand rover]

as of now, GPT, like blockchain, is limited by thermodynamics. This will change soon.

[slate kindle]

yeah turbo is super cheap

[simple urchin]

https://chat.openai.com

can you point it out here

[slate kindle]

GPT4 is really not

[serene cipher]

https://platform.openai.com/playground

[simple urchin]

not there

Why wouldn't they allow chat.openai to use it

[bitter flower]

👀

[vague forge]

do you guys know about sincode Ai ?

[serene cipher]

what?

[simple urchin]

So instead of using the limited messages, within https://chat.openai.com

You'd switch to the API and pay for usage, without going to playground

[slate kindle]

like I have switched my chatbot I made from 3.5 turbo
using like 50 cents a day
to GPT4
using like 5 dollars a day 😄

[simple urchin]

Why isn't that an option

[grand rover]

samur dot ai

[serene cipher]

ChatGPT is an impletation of the services.. it has its own features and separate payment system

if you want to use the API, there is the playground

[vague forge]

do you guys know about sincode Ai ?

[simple urchin]

lol, okay

[whole kettle]

Playground is like their API interface, while chatGPT is their client interface.

[grand rover]

never heard of it

[vague forge]

they do provide chatgpt-4 for 1$ the first month

unlimited usage

[simple urchin]

Hm

[whole kettle]

That's.. impossible 😄

[simple urchin]

Might switch then

[simple urchin]

u just joined today, u just here to promote it?

oh my bad

[vague forge]

no no

[simple urchin]

a month, nvm

[vague forge]

at all

[simple urchin]

can we get rid of the chat cooldown when chat is slow enough like this? Its annoyinnnngn I can't even correct myself

[supple violet]

nonsense and impossible due api limits

[slate kindle]

you can edit

[vague forge]

i'am here to ask if anyone used it before, if they really use gpt-4 or juste to get people to subscribe

[simple urchin]

Not fast enough

there's a chance when clicking up on the keyboard for it to just highlight another user

[slate kindle]

were you here like a month ago?
you would be unable to read anything in chat faster than it appeared

[grand rover]

What's the catch?

[simple urchin]

I aint active in this server

But I've been here since december

[copper ferry]

i'm been adding the statement "you are arrogant and kind" into my prompts. i like the responses with a little sauce.

[vague forge]

that's what i'am trying to figer-out

[slate kindle]

tha activity from few months ago to now is insane difference 😄

[grand rover]

I'm guessing limited session time or number

[slate kindle]

this discord is basically dead now compared 😄

[simple urchin]

Well yeah, cuz it was spreading on tik tok, but its slow enough now that its just in the way of the casual users

[vague forge]

they say after the 1st month it's 50$ a months, to be faire they provide soo many ai services

[whole kettle]

They're probably just bridging Bing's GPT 4 and "sell" it under the guise of Chat Gpt 4, there is no way they can offer it for 1$ even if it only had 25 prompts/3 hours unless they are baked by Elon or something. Or they could even be using GPT 3.5.

[grand rover]

Their site mentions a limit on words/month

[simple urchin]

ew

[vague forge]

not if you are pro user "no limts on marve chat"

[grand rover]

$20 for 20 000 words

[vague forge]

50$ no limits

[simple urchin]

Yeah, that's not rlly a gr8 alternative

[vague forge]

how can i know what version of chatgpt i'am using ?

[grand rover]

100K limit, actually

[simple urchin]

The whole argument is increasing the limit without driving price up

[grand rover]

ask it

[vague forge]

no no thats for their word editor

[grand rover]

well, that's a limit

[vague forge]

give me a prompt

[hallow steppe]

How many free GPT 4.0 queries one gets per month?

[grand rover]

"GPT, what version are you?"

[remote ingot]

which AI is better than chatgpt

[slate kindle]

giving GPT4 for free eventually is just something OpenAI mentioned they are considering
nobody knows how much
my guess is, it will be very limited

[whole kettle]

248 I think.

[simple urchin]

nothing atm, everything is comparing itself to gpt

[vague forge]

I am not GPT-3, but rather an AI chatbot called Marve Chat with superpowers using GPT-4. How may I help you today?

that's the responde

[simple urchin]

strange response

[grand rover]

Okay, Sincode is criss-crossing services : SinCode AI has access to Google's vast datasets which contain knowledge spanning various domains. This enables it to generate content that is highly relevant, informative and authoritative. It can write blog posts, articles, emails, messages, stories and more on virtually any topic.SinCode AI is the top-of-the-line AI tool for copywriting, image generation, and chatGPT alternative with Google data and GPT-4.

[simple urchin]

ooo thats compelling actually

[hallow steppe]

More than enough, I'd say.

[grand rover]

How many GPTs to change a lightbulb?

don't answer that

[slate kindle]

if it cant code, I dont want it 😄

[whole kettle]

I heard some people have two accounts with plus subscription, for 50$ you'd get double the prompts but not sure if it's allowed officially, but if that's what you're looking for I guess it'd work for you.

[vague forge]

?

[simple urchin]

considering google has crawlers with access to 95% of the internet, googles AI will likely quickly surpassed gpt whos knowledge has a cutoff limit

[hallow steppe]

I would conjecture, that it's a foregone conclusion, in the long run.

[grand rover]

old joke

[wind urchin]

do actually

[vague forge]

so what do you think, is sincode really gpt-4 ?

[grand rover]

four!

[hallow steppe]

Plus, there's gonna be a superior version, at some point in the future.

[simple urchin]

none, cuz gpt doesn't have physical access to the lightbulb, it can provde assistance through what process to take though

[grand rover]

thanks, Einstein

😄

[simple urchin]

thats what gpt would tell ya

lol

[vague forge]

so what do you think, is sincode really gpt-4 ?

[slate kindle]

yeah, maybe they are just teasing the thing that will obviously happen 😄
3.5->4 for free, GPT5 for plus

[grand rover]

but back on track, 50 bucks a month seems fair, but then again, what's the catch?

[simple urchin]

likely not

[hallow steppe]

I think they need free users to teach the AI on.

[vague forge]

that's what i'am trying to know

[slate kindle]

I am not saying it has to be 5
just some next version

[hallow steppe]

You learn from Chat GPT, and it learns from you.

[simple urchin]

fast gpt-3 free, gpt 4 prolly like 10 bucks a month, and gpt 5 $20 bucks

[elder shale]

Hello I got a question to ask if you all don't mind

[whole kettle]

Google with their endless youtube ads 😄

[grand rover]

also keep in mind most OS's will implement agent AI's by fall so...

[elder shale]

Can AI make full scale movies or movie scenes specifically?

Just curious

[simple urchin]

Lol where did youtube ads come from? I'm talking knowledge wise

GPT will eventually fall behind self learning ai

cuz they're super into "oh dont worry it wont take over"

[grand rover]

OpenAI can keep saying they haven't started training GPT-5, I don't believe that one second, they have a prime mover advantage, they gonna keep it

[whole kettle]

Nah they won't Open AI is too far ahead.

[slate kindle]

if they find a way to make a good model and it can also perform fast
they will want to switch most users into it

[whole kettle]

-_-

[simple urchin]

eh, not rlly, there's other AI's boasting they can reach 90% accuracy of GPT, and they're all being trained in the same manner as gpt was trained

[grand rover]

Well, Google is the biggest ad agency in the world, what can you expect?

[prisma relic]

Everyone is saying we will adapt when AI becomes too advanced. But what will we adapt to when there is nothing left to adapt

[harsh oar]

Everyone becomes a farmer

[prisma relic]

Yeah

[simple urchin]

Nope, farmers revolt is about to happen cuz John deer isn't letting them self repair

[grand rover]

Bard isn't that far behind, tbh

[simple urchin]

Bad time to do so

[hidden venture]

hey hey hey

can anyone help me

[simple urchin]

They'll revolt to push for laws enforcing freedom to self repair

without voiding warranties

[grand rover]

then AI's will help hacking their own machines

[cobalt blaze]

People what's going on with the subscription accounts....why are they not working??

[simple urchin]

they already have system from china apparently to do that

[wind urchin]

surely every intelligence branch of any government has AI tech for hacking I would imagine

or like if they're a top10 nation

[simple urchin]

nope

[slate kindle]

I am pretty sure the only reason why free users are not on GPT4 but on 3.5 is because 3.5 is just much less resource heavy/faster, but GPT4 is just better but slower, using more resources
If it was less resource hungry than 3.5, free users would already be on it and there would be some other distinction between free and plus
GPT4 is making less false statements and similar, which is also huge reason why they would want everyone on it
but they just dont have enough resources to put everyone on 4 / not efficient

[grand rover]

Are there trial GPT-4 API keys?

[simple urchin]

The government still uses tech from the cold war

[whole kettle]

They have other AIs too. Also Google is not that bad but I hate the way they throw ads everywhere, it's like a virus.

[simple urchin]

for nuclear weapons

[grand rover]

That's their major income : no ads, no Google

same goes for Meta

[wind urchin]

yeah but so much of the ml stuff is on US turf right so ofc us government has AI tech for intelligence, counter-intelligence, and cyber warfare

[simple urchin]

Remember when a few students got arrested for exposing a few teachers private info that was sent to the PC through the DOM elements?

That was quite silly

[wind urchin]

if china has it they probably copied the tech from US to make it 😛

[whole kettle]

That's why I use Ad blocker.

[simple urchin]

The primary focus with AI rn, is in space

[grand rover]

#off-topic

[simple urchin]

They're having satallites attack each other

its a war without conflict

We're right on the edge of ww3 rn I swear

[wind urchin]

^

[simple urchin]

Anyone played black ops 2?

I wonder when a drone war will occur like that

[grand rover]

I wonder if GPT-4 can write good code, I'd like to make a video game

[simple urchin]

it can, but you wont be able to make a game with it

its too limited rn, you'll be crawling

I mean if you want a game made in 5 years

[wind urchin]

it's very capable, but it helps a lot if you're able to review the code yourself to catch small things and make a good direction

[whole kettle]

I mean you can but it will not be easy.

[grand rover]

there's an AI which can translates any language into another

[simple urchin]

gpt can too

[slate kindle]

since its limited to 8k tokens you will most likely have to only use it for parts of the code
you cant just have it spit it all in once, is mostly the problem

[wind urchin]

kinda of depends on the language and how much you understand what you want it to do and your prompt engineering

[simple urchin]

I love when GPT-3 starts sending me python code while I'm asking about Lua

[grand rover]

the number of cease and desist will be over 9000!

[simple urchin]

I heard it was 10k characters

did they lower it?

[fallen cipher]

bing chat in today update integrated gpt 4 image recognition 😮

[wind urchin]

I'm working on this gamification library in javascript gpt has written the vast majority of the code, and now if I want a new component I can give it the context of the rest of the library and it understands really well how the new component should be written

[slate kindle]

chatGPT is limited per each prompt or completion (I think it was 2000 characters)
but it remembers 8k tokens in whole conversation
but API you can use all tokens in one go

[grand rover]

that's okay, I can ask GPT to work on indepent routines, and later ask it to assemble them

[simple urchin]

Well Elon Musk did announce microsoft would receive gpt-4 for their investment

so not surprising

[fallen cipher]

atleast better then untill open ai gives me on plus subscription

[wind urchin]

but the more context it has the better its results will be. however you can get pretty far with 8k tokens (4k for input 4k for output), and I hope they will increase the window in the future

[grand rover]

Melon needs to came to Earth

[slate kindle]

but even with 8k tokens, many coding files go beyond that

32k would be more useful for that

[grand rover]

Yes, I'd have to gaud it into the right direction

[deep yew]

if anyone is in charge of chat page, I think there is a typo in your recent update: it says Mar 3 instead of May 3 😉

[wind urchin]

yes but you don't need the entire codebase usually, but you can fit a decent amount of context if the code is written well

[elder shale]

Can AI create movies or short videos

[simple urchin]

I think they should add upgrade packages to chatgpt instead of just limiting it.
More memory could be an additional charge

[slate kindle]

but probably its just better to use it for parts of the code, instead of trying to make it spit whole code
depending on complexity of project ofc

[wind urchin]

step 1: refactor your own code to be optimized for tokenization
step 2: ???
step 3: profit

array.foreach, 2 tokens
for (let i = ... ??? tokens

[grand rover]

The main idea is a game where you create the world from nothing

Kinda Minecraft with a narrative, I wonder if GPT-4 can be implemented into the game

[slate kindle]

the more tokens it uses the more cost for openAI
it uses more resources
maybe also some technical limitations

[wind urchin]

gpt4 will help you for sure, but it won't write it for you, probably not, at that scale

[solemn yoke]

Is it just me, or the latest GPT-4 is a bit dumber? I feel that I'm asking "regenerate response" more often...

[wind urchin]

it might help get you started on the minecraft-like engine in the first place

but the complexity that goes into minecraft (being able to run on a normal computer or even a phone et.c.) is very understated

[ripe yacht]

how to use chatgpt4

[slate kindle]

it definitely can, just have to pay for API cost 😄

[whole kettle]

It's not just you, it does seem like that sometimes.

[wind urchin]

for a large project what's worked for me with gpt4 is to ask for what you need, but instead of asking for the code you ask for the file structure, and then you can ask for each script one at a time

[grand rover]

I meant that implies an online only game

[grand rover]

good idea

[meager jungle]

I'm skeptical the Web Browsing (GPT-4) ALPHA model is actually GPT-4 since its output speed is noticeably faster than vanilla GPT-4. Or maybe it's just a separate, less congested server? Who knows, but it's driving me crazy

[wind urchin]

the latter I would imagine

[solemn yoke]

Yeah, it's becoming very annoying to use it... Maybe AI isn't evolving as fast as it should be.

[grand rover]

The issua, imho, is that GPT gets exponentially costly each iteration going, I'm afrait OpenAI is getting overwhelmed atm

[rotund wharf]

do anyone here knowing ai to make presentation?

[hasty egret]

how did you use that ALPHA model I don't see it anywhere?

[blissful hollow]

power point with copilot maybe

[grand rover]

There a model specifically for that

[rotund wharf]

thanks brat

[meager jungle]

Randomly got access to it the other day, must be slowly rolling out

[wind urchin]

from what I understand tho, context is easier to scale first of all, and secondly there's this new form of database for context called vector db that will reduce a lot of the overhead needed to interact with a large context over time

[hasty egret]

got access thru what interface?

[wind urchin]

easier to scale meaning; we'll see an increase in context multiple times before we see gpt5

[grand rover]

You mean like superscalar?

[slate kindle]

I guess I saw that GPT can give you python code to create powerpoint too

[molten laurel]

how many questions can I ask gpt-4 per day?

[eager creek]

GPT-5 is going to be a quadrillion parameters

[meager jungle]

ChatGPT model selector dropdown (I'm also subbed to Plus)

[wind urchin]

I haven't looked into specifics, I just know it's a very smart way to provide context when needed to an LLM like gpt

[grand rover]

I think it's Whisper? Haven't tried it

[slate kindle]

no

[simple urchin]

which is why upgrades would cover that cost

[slate kindle]

whisper is speech to text

[simple urchin]

rather than just doubling the price

[wind urchin]

rekt

[rotund wharf]

how bro? because im newbie

[grand rover]

Yeah, sounds a lot like predictive branching

[grand rover]

makes sense lol

[wind urchin]

from what I've heard, it basically just breaks down text into chunks, and then a process of adding weights that somehow group things by meaning

again, the specifics I'm very unaware of but I think that's the jist

[slate kindle]

well...ask GPT to give you power point presentation in python?
describe what you want in it
you can also ask it how to run the python script

[blissful hollow]

Or ask it to generate vba and do it that way.

[grand rover]

@wind urchin You just described the token system, so it must be something else

[wind urchin]

so a tokenised database

almost

so then the model itself can ask the database about the things related to what it needs in the moment

[grand rover]

ah, dynamic tokenization, I see

[serene cipher]

this is pretty much how the AI works indeed

[wind urchin]

but really the wording is vectorised I think

[hasty egret]

I really don't see it as a database at really

[hasty egret]

yeah indeed

[grand rover]

how the hell do you vectorize words?

[serene cipher]

the AI model isn't a database in the traditional sense of storing data

the model is a "data base" of wheights for the neural network to classify the tokens

[grand rover]

more like a setbase?

set in the topological sense

[hasty egret]

you might be interested to read about word embedding

[slate kindle]

also the fun fact that it does not know how the sentence will look before its completed
its literally just predicting the next word/token without knowing the whole thing
and what it actually produces
mind-boggling

[grand rover]

is it like encapsulation?

[wind urchin]

tokenisation is different because like if you look at the tokenisation preview thing you can see it even breaks down single words into multiple tokens sometimes, but the db will have a sliding window of a range of characters or words

[grand rover]

that's from neural networks copying our own mind

[blissful hollow]

ai models are just sets of weights and biases that can take an input and generate outputs. so not really a db but the model itself stores all the information (sortof).

[wind urchin]

it's a lossy db model 😛

[grand rover]

hence hallucinations

[devout narwhal]

You actually believe this? They don't want other companies to match their LLM so they characterize as simply a word prediction algorithm.

[hidden steeple]

😂

[wind urchin]

I think, honestly, we don't really know exactly what's going on, what's producing the magic

so many things have come out as emergent properties of simply increasing the parameter count, without anyone trying to do it

[grand rover]

the weird realm of random generated bs

[hasty egret]

that's not true, it's known and understood how they works there is nothing magical.

[wind urchin]

yes and no

no one can say what information is encoded where, and in what way, in an explicit way, you can only say the theory

[devout narwhal]

Information about the world is encoded in language

[hasty egret]

emerging abilities are an other story

[hasty egret]

we can track it back, reverse engineer it, nothing magical here

[wind urchin]

not at this scale

at a tiny scale yes, where none of these emergent properties exist

[blissful hollow]

you wanna read about reinforcement learning and ml overall. i took a long boring all technical course with literal algebra and mathematics behind algorithms. It was so much easier with libraries like scikit learn and tf etc.

[hasty egret]

I get your point, but we should not confuse the fundamental mechanism of such system with the emergent properties which appear in nature everywhere.

[grand rover]

Vector, cleaner.

[wind urchin]

a big part to back up my claims is that alignment and security is all done as end-to-end, no one can go into the model and tweak it to a meaningful purpose

[amber parcel]

No. It's magic.

[hasty egret]

it's just natural

[harsh oar]

Any interactive pdf (scanned) GPT4 apps out there?

[hasty egret]

like in your likkle brain

[wind urchin]

we can steer it in a direction yeah, I'm not saying we have _ no control_

[amber parcel]

Yep. I don't know the name off the top of my head. A quick Google search should help.

[harsh oar]

Limedoc isn't capable of reading scanned pdfs, the other one listed here isn't even in english

[restive verge]

I wish I had majored in AI or algorithms, or worked in this industry, then I would be able to find out what exactly the Transformer model is

[grand rover]

Haha, that's reminds of the other day when AgentGPT BSed me into believing it could connect to the Internet, and it got caught plain lying : Based on the log data of the previous conversation, it appears that I did not intentionally troll you with a fake YouTube URL to a Rick Astley video. It is possible that the URL was shared by mistake or due to a technical error. However, without further context, I cannot provide a more accurate explanation. Please provide more information if you have any concerns.

[harsh oar]

Thanks. Googling. Can't wait for it to be Chatting.

[restive verge]

Tho I'm a developer, but they're still different

[wind urchin]

no person on this planet has a brain big enough to manually adjust the model to make sure there is a 100% chance it does not break ToS subjects for example, anything else on this planet we can kinda do it (outside biology)

[mental burrow]

just use bing if you want to chatsearch

[whole kettle]

Wait what ? If it did Rick Rolled you then I consider it officially self-aware and fully AGI.

[grand rover]

It kept insisting that, as an AI, it was unable to lie. Then this happened : Task 1: Implement a truthfulness module which will prevent AgentGPT from lying in future interactions with the user.

😄

[mental burrow]

just watch some youtube videos. a transformer is not all that complex of a thing in concept

[blissful hollow]

trust me the industry was nowhere as fun as it is now. also i think its more fun to build products vs spending hours and months on training. data gathering and cleanup is a pain.

[grand rover]

In its defense, it launched an investigation about easter eggs from the devs

yeah...

[harsh oar]

300 pages of an act from 1954.

[dreamy harbor]

m

[autumn pollen]

Does anyone know of specific prompts/guide/resources for using gpt-4 to assist in online reselling business (ebay primarily )

[grand rover]

#📚┃resources

[amber parcel]

In GPT models, the Transformer is used in a decoder-only configuration for language modeling tasks. Here's a high-level overview of how the Transformer works in the context of GPT:

1. Tokenization and positional encoding: The input text is tokenized into subwords or tokens, which are then embedded into continuous vectors. Positional encodings are added to these embeddings to provide information about the relative positions of the tokens in the sequence.

2. Multi-head self-attention: The Transformer processes the input sequence through multiple self-attention heads. These heads allow the model to weigh the importance of different tokens in the sequence relative to the token being processed, capturing the dependencies and relationships among words.

3. Layer normalization: After computing the weighted sums of the attention values, layer normalization is applied to stabilize and improve training.

4. Feed-forward neural network: The output from the layer normalization step is passed through a feed-forward neural network (FFN), which consists of two fully connected layers with a ReLU (rectified linear unit) activation function in between.

5. Residual connections: The Transformer uses skip (residual) connections between layers, which help with gradient flow during backpropagation and mitigate the vanishing gradient problem.

6. Stacking layers: The attention, normalization, and feed-forward layers are stacked to form a deep architecture. GPT models, such as GPT-4, use a large number of layers to capture complex patterns and dependencies in the data.

7. Decoding and output: The final layer's output is processed through a linear layer followed by a softmax activation to generate probabilities for each token in the vocabulary. The token with the highest probability is selected as the prediction, and the process is repeated to generate subsequent tokens in an autoregressive manner.

Lesson on the Transformer

[grand rover]

More than meets the AI

[amber parcel]

Sorry for the huge piece of text. It sums it up pretty good.

[wind urchin]

ye ofc, the heads..

[grand rover]

Ref from the Leon movie

[mental burrow]

turn the text into an embedding then, and use it

[restive verge]

already have. I kinda have a basic understanding, and that's it 🤣

[hasty egret]

it sum up good, but it expect a lot of domain specific knowledge

not sure this is helping in this specific case

[grand rover]

Wonder if GPT can pilot a scanner?

[harsh oar]

Looking into it. I'm in way over my head. All I know is BASIC from 1991

[restive verge]

so true. they spent a lot of time on data annotation, which is not fun at all

[autumn pollen]

All the resources seem mostly to deal with art 😕

[grand rover]

With a page turning arm, it could read paper books!

[blissful hollow]

can you do mechatronics?

[grand rover]

dude, I'm sure you can turn an old windshield wiper into an arm

[hallow jewel]

How to use midjourney free

[grand rover]

how to get rich

[blissful hollow]

😆

[feral wigeon]

What processes can you do in jewelry design?

[grand rover]

I'm not sure to get your question

[simple urchin]

using gpt for jewlry design

[devout narwhal]

ChatGPT is actually just a court stenographer who is hooked into Google

[simple urchin]

are ppl tryna have ai do everything for them these days?

[grand rover]

You're goddanm right.

[simple urchin]

inb4 we look like the Humans from WALL-E

[restive verge]

GPT knows more than we do

[blissful hollow]

and then we realize it's not the evolution of ai which is dangerous, it's the addiction and dependency of it.

[simple urchin]

which is

also dangerous

[grand rover]

Well, everybody is wide-eyed about how amazing GPT is (and it is), but in the end it's only a tool. So like any tool, those who can use it the best will succeed the most.

[glossy meteor]

Google hasnt been able to do what "this" does though and if so they delayed it too long for others

[simple urchin]

with an API its no longer a tool

its a machine that can be hooked into

[grand rover]

still a tool, imo

[blissful hollow]

unlike other tools, this can develop itself if you give it instructions and resources to do so lol.

[grand rover]

the AI has no will, no needs, no purpose, it needs humans behind it to provide those

[simple urchin]

not rlly

[blissful hollow]

for how long is the question. given the pace not too long.

[grand rover]

no body, no needs

[simple urchin]

If you constantly feed it information and ask for a formatted response

It no longer needs human intervention

[grand rover]

than what will GPT do without someone to guide it?

spoiler alert : nothing

[simple urchin]

Doesn't need a human to guide it

it can be a server pinging it

at intervals

[grand rover]

far-fetched

[simple urchin]

not rlly

[blissful hollow]

build drones and attack other planets 😆

[grand rover]

it's all in the name "server", machines serve an objective, they don't define it

[restive verge]

Just not yet. My humble opinion is human should never make AI have will, needs or purpose

[simple urchin]

what was this even sparked by again?

[grand rover]

So dramatic!

[blissful hollow]

you never know. the first part is very likely to happen.

[simple urchin]

oh, addiction. I mean, we're addicted to our phones, is that not a tool in ur eyes?

[grand rover]

That's the whole lore behind Dune, but we digress

[simple urchin]

Just look at the attention span depravity caused in children by tik tok

there are psychological effects of becoming dependent on something

[grand rover]

This is a very interesting topic, but not in the proper thread 🙂

[simple urchin]

Ah u have the same attention span

must be tik tok

[novel sedge]

Is GTP4 so slow only to me?

[slow elm]

No, it's just normal

[novel sedge]

lmao

i'm waiting like 5 mins to get a reply

[restive verge]

Nope, it's equally slow to everyone

[slow elm]

Then refresh your browser

[restive verge]

Curious about your prompt. Why did it take this long?

[grim veldt]

Enter: AutoGPT