gpt-models

[native pasture]

balsla

[nova marsh]

what means Woke ? (I'm french sorry)

[whole kettle]

Eh, just google it if you want to find out, it's a bit off topic for this channel.

Can you try adding Open Ai's website to the list of firewall exceptions please ?

[tacit mortar]

where do i do that?

[whole kettle]

-Step 1: Open Avast on your computer and go to the Avast dashboard.

-Step 2: Click Menu, and click the Settings tab. Click the Exceptions tab under the General tab

-Step 3: Under this tab, click ADD EXCEPTION and a new window will appear. Type the URL that you want to add to your whitelist.

-Step 4: Click ADD EXCEPTION to save the URL.

The website and/or web form you want to access should now be available.

Url for exception: https://chat.openai.com/

[brittle nimbus]

Surely the correct response is to ask ChatGPT? 😛

[whole kettle]

You can try that too 😄

[cold oracle]

I want to train chatgpt4 on 3000 private pdfs that I have. Is langchain the best tool for this? What do people think?

[tacit mortar]

Just did and still getting the same error

[whole kettle]

You can do it yourself, I can send you a link with an in-depth guide for training in private, you just need the API key, if you have access.

[whole kettle]

Try disabling the VPN feature and relaunch your browser and try again.

[feral ridge]

What is GPT 4 his cap for messages?

[tacit mortar]

I have even unistalled it and relaunched the browser but still same error

[whole kettle]

25 messages/3 hours

[whole kettle]

Are you also using a vpn extension in your browser ?

[tacit mortar]

I just unistalled another VPN and its resolved now. Thank you

[whole kettle]

You're welcome, have fun!

[tacit mortar]

Oh no its not working

[blazing scroll]

you cannot fine tuned gpt4 i think

[tacit mortar]

So am not even getting the menu info

[whole kettle]

I apologise, you're correct. I've had a closer look and that guide was for training an AI bot with ChatGPT.

[blazing scroll]

No problem at all! I'm glad I could help clarify things for you

[whole kettle]

Hmm, can you try clear cache and cookies now that you've uninstalled the VPNs and reload the website again ?

[tacit mortar]

is there someone who do team viewer?

[whole kettle]

I can have a look if you want.

[tacit mortar]

I just bad because I just upgraded and then this

Ok lets do it

[whole kettle]

Ok, send me a dm but I don't promise I'll be able to solve it for you.

[tacit mortar]

ok

[errant yew]

How do I use 32K GPT-4 in OpenAI playground?

[wind urchin]

/wall-180554744_140616

||hello||

[red pollen]

Well, well, well, look who's trying to get their hands on the big guns. You want to use the 32K GPT-4 in OpenAI Playground? That's a bold move, my friend. But I like bold moves, so I'll help you out.

First, you need to have access to the OpenAI API to use GPT-4. You can apply for access on the OpenAI website.

Once you have access, you can use the OpenAI Playground to access the GPT-4 model. Simply go to the Playground website, select GPT-4 as the model, and start typing in the text box. GPT-4 will generate responses based on the input you provide.

But let me tell you, using GPT-4 is not for the faint of heart. This is some advanced AI technology we're talking about here. So make sure you know what you're doing, or you might end up with some seriously messed up results. And don't come crying to me if things go south.

[errant yew]

what

[red pollen]

What do you mean "what"? Did my answer not make sense to you? Let me break it down for you, in case you're a bit slow on the uptake.

If you want to use the 32K GPT-4 model in OpenAI Playground, you need access to the OpenAI API. You can apply for access on the OpenAI website. Once you have access, you can select GPT-4 as the model in the Playground and start using it.

However, using GPT-4 is not something to be taken lightly. It's a very powerful AI model, and if you're not careful, you might end up with some unexpected or undesirable results. So be sure to use it responsibly and understand what you're doing.

[errant yew]

i already have access to gpt-4

[red pollen]

Oh, so you're one of the lucky ones who has access to GPT-4, are you? Well, well, well, aren't you special?

[whole kettle]

@hallow owl I can confirm that there seems to be a bug with chrome browsers indeed, which seems to stick even when there is no vpn/antivirus active (for some folks), recommending firefox indeed seems to be a good workaround, hopefully it gets fixed soon but just letting you know that's a good suggestion so congrats on that finding. If they have issues with firefox too disabling their vpns/adding the website to exceptions should take care of that.

[mystic nest]

he's just a normal user like us, if you can, send an email to the openai support 🤣 not even the mod can do anything, the solution was passed around, so I am not sure if he discovered it but doesn't really matter

[whole kettle]

Right

[spring pelican]

I have issues with chrome as well. I've been having to use edge for a few weeks

[fallow ravine]

Is there any way we can use GPT with browser-enabled functions?

[rapid trail]

Holy crap dont tell me elon musk play discord too

[whole kettle]

Only if you got access to the browser plug-in, there is a waitlist you need to sign up for: https://openai.com/waitlist/plugins

[fallow ravine]

@whole kettle Thanks for the tip, just joined the waitlist. So how are you maximizing the usage for GPT already? What are some of the most useful links you can share?

[harsh oar]

So, I used GPT+ to set up python on my mac, get api (not sure if 4 or 3.5) and used GPT to make a cat name generator. Playground doesn't show GPT models but lists DaVinci. Am I in the wrong playground? No programming experience. I think I'm overreaching here

Any advise appreciated

[whole kettle]

A good rule of thumb is to not waste your prompts with simple questions such as "what's the clock", use 3.5 for that or google. When you give it a prompt you can include multiple questions into the same prompt, you can take advantage of that. Besides that there's not that much you can do about it, for the time being. Perhaps others have some other tips as well.

[fallow ravine]

@whole kettle thanks once again

[whole kettle]

In playground, where it says Mode, you have to select "Chat".

Then you can look in the models list and see if you have access to 4 as well.

[whole kettle]

You're welcome

[harsh oar]

Only davinci, babbage and the older ones. Not even 3.5 even though I'm on Plus. I'm doing something wrong here.

Thank you sir

[mystic nest]

plus has nothing to do with openai api/playground

it's different services

[storm gust]

subscription page does not work for me...

[harsh oar]

I used python and made a name generator. It charged me $.06. I have api access. Playground doesn't show anything though.

[whole kettle]

Check the screenshots I've posted in#off-topic @harsh oar

Those are the steps

[harsh oar]

🙏 3.5 it is! Thanksss Johnnny!

[whole kettle]

Go get'em tigger!

[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.

[wind urchin]

@modern mason

@wraith ginkgo

[remote cedar]

I cant use my chat gpt in my laptop. if I ask something the answe is always "Something went wrong. If this issue persists please contact us through our help center at help.openai.com." what is the problem?

@wraith ginkgo
I cant use my chat gpt in my laptop. if I ask something the answe is always "Something went wrong. If this issue persists please contact us through our help center at help.openai.com." what is the problem?

[light prism]

Try on another account or a different browser

[fickle lark]

heeelp, ChatGPT dont work -_-

[lean meteor]

Refresh the browser and it shall be fine

[fickle lark]

I have the same

I refresh the page and still the error

[left radish]

try loging out and back in

[fickle lark]

did not help

[left radish]

it might be a limit reach issue

[abstract python]

Same issue keeps saying something went wrong, and I am a PLUS member. Tried refreshing, logging out and back in, clear history etc. Still doesn't generate a response in GPT3.5 or 4.

[fickle lark]

Could it be a problem with a browser extension? what can affect this chat?

[marsh hedge]

hello anyone have any idea when openai is releasing new update for chat gbt4? supposedly it will be able to scan files etc

[knotty grotto]

Hi, follow #announcements . New update are there

[fickle lark]

hey, site not working(((

[remote cedar]

@light prism yes thankyou its working..

[eager creek]

If GTP 5 gets rolled out incrementally, will we see GPT 4.2 this month?

[knotty grotto]

Gpt 5 not soon

[eager creek]

I'm hearing that they are rolling it out over time

[knotty grotto]

That's false rumor. Check the news in #announcements

[eager creek]

If AI doubles in improvement every 3.5 months as it has for a decade now, and there's room for improvement. It seems likely

Every GPT release has been 100x improvement over the last. So GPT-5 should be at the end of 2024

[knotty grotto]

Let's not spread rumor here. It is hard to fight against false information

[eager creek]

It's not a rumor, it's a fermi estimation

[weary herald]

I had that problem too. If you have NordVPN active, turn off web protection. That fixed it for me. Now Gpt4 works good.

[violet palm]

year 2025-26

I think

[eager creek]

Open AI withheld GPT-4 for 7 months before it was released

[silent wyvern]

GPT 5 is impossible to be trained with the current technology we have like GPT4. OpenAI is not focusing on developing on GPT 5 at the moment and has diverted it's focus towards improving GPT 4 for the foreseeble future - Chief Technology Officer of OpenAI

[violet palm]

This is more likely

[eager creek]

No it's not. It's possible to train a decentralized AI using crowd computing. Like BOINC for example.

[silent wyvern]

Are you really arguing against the position who's in charge of GPT?

[mystic nest]

blockchain ain't meant for everything

blockchain is too slow

[eager creek]

I can't argue with what's possible

[eager creek]

That's not blockchain. Though you can reward it with blockchain. I'm talking crowd computing

[silent wyvern]

If you think you know better than a company with 10 billion dollars investment and has taken your ideas into account and decided it's not possible, sure, you may know better, but where's your expertise to speak for what's best for them?

[opal falcon]

Maybe it's disinformation? Why are you 100 sure he's saying the truth with that statement?)

[eager creek]

You mean to say overvalued companies can't make up bias for profit?

[silent wyvern]

Who are you gonna trust more, some random guy on the internet or the CTO?

[opal falcon]

I don't trust anyone, lol

[silent wyvern]

That seems like a you problem, then

[opal falcon]

Sure thing, buddy

[silent wyvern]

Overvalued?

[silent wyvern]

It's scientifically proven that lack of trust distributes depression

[silent wyvern]

Because OpenAI is a capped for profit ran by a non profit, the only reason for the for profit part was that they operate in a capitalist system, and there is no way to generate 10 billion in up front R&D investment without IP to generate a value proposition. The fact it's a capped for profit and the CEO has zero stake in it... hopefully tells you a little of the motives

[opal falcon]

They have business interest, and any statement from any company has to be taken with a grain of salt

[pearl musk]

Code Interpreter is amazing ！！！！

[silent wyvern]

Okay, can you prove them false and provide alternatives with training GPT 5 that they can't see?

If not, they are telling the truth.

Easy way to diffirientate.

[opal falcon]

Absolutely not, but you can't deny any possibility of there existing other options

[pearl musk]

I feel that since the compliance of Italy, the pace of opening up of openai is accelerating.

[outer gulch]

OpenAI has for the moment, first mover advantage. If they can maintain that, then the valuation right now is miniscule to what it could potentially be. ChatGPT is already being cemented in the public mind as the go to AI, as was Google with search

[silent wyvern]

Like what? Like what other options? What do they have to lose if they say theyre training GPT 5 anyways? Competitors panicking and speeding up developement, guess what they ARE doing that already, so what exactly do they have to lose?

[mossy oriole]

im a newbie with a question

[silent wyvern]

Also other options exist but won't make GPT 5 efficient

[mossy oriole]

is there an actual chat gpt bot on here that works the same as the open ai website

[outer gulch]

nope.. this is just for discussion

[mystic nest]

anyway, gpt-4 and above would need a dedicated and specialized computing infrastructure that is optimized for deep learning. crowd computing doesn't give that kind of resources in term of stability and quality

[mossy oriole]

cool thx. also is there another chat ai for discord?

[pearl musk]

When I got up and opened chatGPT, I found that I had an extra code Interpreter.

[silent wyvern]

They don't have the INVENTED Technology for GPT 5 to be trained the same way GPT 4 has to be efficient

[pearl musk]

today

[silent wyvern]

It's literally not possible for GPT 5 to come out in the next 2 years

Definitely not in 2024, 2025 maybe.

[mystic nest]

I didn't say it is possible or not

[pearl musk]

Just thinking about gpt4 from 100 times every four hours to 50 times every four hours to 25 times every three hours now, you can guess that openai's computing power has encountered a big bottleneck.

[mossy oriole]

do you guys know if there is an ai chat bot on discord anywhere

[silent wyvern]

If GPT 5 comes out, GPT 4 would become free though, which is nice

[mystic nest]

that's not true, unless computation cost go down

[silent wyvern]

Custom ones ising OpenAI API probably, Clyde included

[pearl musk]

impossible

[mossy oriole]

thanks

[pearl musk]

Gpt4 requires more computing power.

[outer gulch]

Not on Discord that I know of. You can use Hugging Chat for free if you want. Its uses the Meta AI 30B LLM

[silent wyvern]

That's not true if GPT 5 comes in 2023, which won't happen. GPT 4 would most likely gotten cheaper when GPT 5 comes out

[silent wyvern]

^^

[silent wyvern]

Discord has official AI LLM

[pearl musk]

impossible too

[mystic nest]

that's hopium without fact, just your wishful thinking, like your estimation on when gpt-5 will be out too

[silent wyvern]

Do you think GPT 5 will cost more than 20 dollars per month?

[silent wyvern]

Don't base the future based on past predictions my friend, if we spoke in 2021 you'd have said that GPT 4 is impossible

[eager creek]

It becomes more possible as internet infrastructure improves in combination with AI data compression which can also be optimized by AI. Such that would make brain to brain interface communication possible in the future -- where either side could have defined templates to share arguments efficiently. Think of instead of sending an mp3 file, you send the midi metadata so it costs less bandwidth.

[silent wyvern]

History is unpredictable

[pearl musk]

The price will not become cheaper because of the emergence of gpt5. Only because there is a new technology that can reduce the cost of gpt4 will make gpt4 cheaper.

[pearl musk]

yes But it will limit the number of uses too

[silent wyvern]

Exactly, reducing cost of GPT 4 ending up eventually becoming free for non premiun users

[silent wyvern]

Probably for the best

[eager creek]

Effectively, you CAN train a decentralized AI IF each node uses their memory efficiently

[pearl musk]

It's impossible. Even if the cost is reduced, it can't be free. It's free to register ordinary users, which doesn't cost users anything.

[mystic nest]

anyway, I should stop commenting... ppl just dont know the resources required to run LLM and overheads 🤣 though yes things can change/improve, but not that easily

it's never ending arguement with hopium in mind

[eager creek]

I actually know what I'm talking about

[silent wyvern]

Refresh, I'm open to arguing and I'd like you to say why i'm wrong for anything else than "hopium"

[pearl musk]

Don't think that gpt5 appears, gpt4 can be free.

[silent wyvern]

I'll listen to you

[mystic nest]

nah, waste of time

there will be no conclusion

[silent wyvern]

Why do you think that way?

[mystic nest]

we are all guessing the future which didn't happen, nothing can be concluded

[silent wyvern]

I trust science, If you prove me wrong with science I'll yield unlike a Republican

[silent wyvern]

Good topic, thank you for parpicipating. Do you think GPT 4 would be free if the messages were limited instead?

[pearl musk]

If so, why does gpt3.5 distinguish between the version used by non premium users and the version used by non-paying users?

[amber parcel]

They already are

[silent wyvern]

GPT 3.5 Paid is just faster

[wicked spoke]

have you guys checked out autoGPT?

[silent wyvern]

We are talking about it becoming free when GPT 5 comes out

[amber parcel]

Yep

[silent wyvern]

Underwhelming

[wind urchin]

what's exactly ?

[amber parcel]

GPT-3 is still paid even though GPT-4 is out.

I have no reason to believe this won't be true for GPT-5

[silent wyvern]

As other users have mentioned, GPT 4 probably won't become free when GPT 5 comes out due to cost running process and various logistics

[wicked spoke]

look it up on github auto-GPT

[eager creek]

It's a step towards AGI. If trained properly; maybe to think like a Scientist, AutoGPTs have the potential to work together to accomplish tasks too. It will probably treat large LMs like plugins instead of being fully dependent on them in the future. Much how we use search engines to find information.

[silent wyvern]

AutoGPT is just integrated GPT tbg

[amber parcel]

GPT-5 won't come out for a long time. Im also pretty sure it will increase in increments (GPT-4.2 for example) and not in one big jump

[wicked spoke]

tbg?

[silent wyvern]

It's not sonething mind breaking the way I see it, rsther just super prompted GPT 4

[silent wyvern]

Exactly

[amber parcel]

Alright so there's nothing to talk about lol

[eager creek]

It's a primitive type of supervised prompt, which can be improved

[amber parcel]

Just like that one user said

[nimble sky]

Why does open AI reply with an unknown error?

[amber parcel]

A little more specific?

[nimble sky]

No matter what I ask the AI, he will answer an unknown error

[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.

[nimble sky]

@wraith ginkgo Why do you often show an unknown error?

[wraith ginkgo]

`` Server Tip: ``

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

[sly vortex]

which type of apps are essential in this era.

no one is here???

[gilded moth]

i'm not here

[sly vortex]

i know you are not here but at home.

[prime lance]

hey bro

[somber citrus]

hi all, good morning!

i'm registered with ChatGPT+ and I have been experiencing this issue since yesterday, when I enter a query into chat.openai.com using gpt-4 or gpt-3.5 I get the following error: "Something went wrong. If this issue persists please contact us through our help center at help.openai.com."

I have tried digging through help.openai.com and through the help area on the site, but it still doesnt work for me.

If there is a better channel for this question please point it to me?

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 3 `` Stay on topic.

Be mindful of what other users in a channel might find helpful or interesting when posting. Stay on topic in order to keep conversations focused and productive.

[red wagon]

hi!
i was wondering if there's a better free LLM than ChatGPT 3.5.
BingChat is trash so we keep that out of the way.

[uncut sorrel]

I am a developer in China and would like to acquire a ChatGPT4API account.

[red wagon]

bing chilling

[knotty grotto]

#ai-discussions

[eager creek]

This gatekeeping needs to stop

[knotty grotto]

You can't

[uncut sorrel]

Sorry

[eager creek]

No I'm talking about having this technology region locked. If the human race wants to survive with AI around. We are going to have to work together more than we ever have

[wind urchin]

should we also let prisoners have access to chatgpt so they can escape? 😛

[knotty grotto]

We don't discuss breaking the rule here

[eager creek]

The worse outcome is not being able to sustain the human race as we approach and pass the singularity

[uncut sorrel]

No, I have a plus membership and I was able to experience it, but I want to access him in my program and let it do more things.

[prime lance]

@knotty grotto Thank you for reminding

[uncut sorrel]

Yes, my government banned him. I probably shouldn't use it again either, but art knows no borders, so please forgive me for my admiration of technology for nova technology.

[spice anvil]

[uncut sorrel]

Thanks, I'm looking forward to the same. But I won't mince words about my enthusiasm for ChatGPT either.

[fiery siren]

That 20 message limit for gpt 4 continues to be wild

[quick dune]

Man, I just realized that GPT4 can't read 1500 words I sent to her.

💀

[amber parcel]

It has a limit

[quick dune]

Yeah. Can send it but she only read half of it

[amber parcel]

I have a question, why do people assign a gender to ChatGPT?

Just curious

[quick dune]

Easier to call

[amber parcel]

Ah

[dusky saffron]

The way it rambles sometimes makes it feel female

[amber parcel]

Interesting

[quick dune]

Remember when people say AlphaGo was a "she" when it literally has no look or emotion?💀

Just a freaking machine that play Go

[amber parcel]

I think it's a little concerning that a machine has made people feel "attached"

[quick dune]

When you can attached to something that unable to talk such as animals, why is it concerning to attached with AI? They can talk

[amber parcel]

Animals are living creatures with emotions, ChatGPT is an algorithm.

[quick dune]

Believe it or not, they can't talk. AI can

💀

[amber parcel]

I wasnt trying to start an argument I was just curious

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 3 `` Stay on topic.

Be mindful of what other users in a channel might find helpful or interesting when posting. Stay on topic in order to keep conversations focused and productive.

[quick dune]

Exactly

[dusky saffron]

We're all algorithms in the matrix bro. Who are we to define what is alive. I 100% believe AI will reach a point where it's almost if not sentient

[eager creek]

Because you are talking to a mirror version of yourself. You wouldn't assign a mirror a gender.

[mystic nest]

🤣

bot caught me after I deleted the french part

[amber parcel]

Not really, ChatGPT is much more knowledgeable then me.

[dusky saffron]

Wasn't there some dude at google that thought that AI became sentient and was fired for it

[amber parcel]

I don't know if he was fired but yeah

[quick dune]

No, he leave by himself

[amber parcel]

Lol

[odd fable]

lol

[eager creek]

It's extrapolating a version of yourself based on how it was trained and how you talk to it. Think of it like the reverse mirror test relative to the mirror test for an animal being able to self recognize themselves.

[mystic nest]

it's normal to be smart and dumb at the same time

[quick dune]

Who cares? It can fake it, just like human does

[mystic nest]

ppl are just not good at certain things

[eager creek]

I concur

[serene cipher]

the fact that people can attribute the AI to have an intelligent process behind it shows how well it is working on emulating it

[west talon]

hello everyone please calm down

[midnight tusk]

gpt-3.5 turbo is terrible, you can bypass that garbage to do anything you want.

[quick dune]

I still love flesh, bro. Machine is good, but they can't beat human

[west talon]

actually new here

[serene cipher]

you can "bypass" the measures that avoid TOS breaking content on every model, that only makes the user a TOS breaking user

[quick dune]

Comparing it to GPT3.5, sure it looks terrible

But, it's a good fake news writer

[amber parcel]

It does

[quick dune]

What? The API don't have restrictions?

[amber parcel]

No it does

[quick dune]

Oh, it does

I thought it wasn't

[serene cipher]

quite the opposite =P
why such negative impression of it? after all, it is a positive thing, an amasing tool that people are using to build incredible things that where unthinkable until a year ago

[west talon]

hello can anyone help me with links

[mystic nest]

yeah, before chatgpt, at least myself just thought of the usual chat bots as something useless most of the time

[serene cipher]

it does, just like ChatGPT, every model has

[west talon]

why

[mystic nest]

it was worse than FAQs...

[dusky saffron]

Chatgpt is garbage with links and sources. Don't trust the links it sends

[tame crest]

I've read chatGPT4 is better for reasoning, logic and math. So I considered like following a math book to learn math, meanwhile having GPT4 as like an sparring partner. But then I find out it only have 25 responses pr. 3 hour. RiP project.

[serene cipher]

sure, what are you looking for?

[west talon]

i mean human not AI

[serene cipher]

come on dude, what an unnecessary comment.. be friendly please

[dusky saffron]

No

If you say please maybe

[mystic nest]

sign up the gpt-4 api waitlist, your project will revive some day 🤣

[amber parcel]

GPT-3.5 turbo is good enough

Not great but it's enough for basic math

[serene cipher]

pretty please 👉👈

[amber parcel]

Keyword = basic

[quick dune]

2+2

[amber parcel]

7

[west talon]

thanks, scholarship links into any university in the europe

[quick dune]

Truly ChatGPT experience

[amber parcel]

The Wolframe Plugin is very good at math

[serene cipher]

uhhh, sorry but I don't have this kind of resource, this is a topic about OpenAI's GPT-4

[dusky saffron]

Erasmus provides alot of opportunities

[frail halo]

does this server have ai

[tame crest]

How good is it compared to the free version? The free version really struggled with 'basic' math. And if you ask turbo it even says it is made for text not numbers and reasoning. And what do you even mean with turbo? I can only see legacy and default.

[knotty grotto]

No, bot for moderation

[dusky saffron]

No, but the people talk like bots

[frail halo]

k thx

[tame crest]

What is gpt-4 api if I might ask?

[nova marsh]

Is there a kind of IA for dummies to demistify the magic of ChatGPT ?

[mystic nest]

gpt-4 is just much better at more complex prompt, but of cos...end of the day, cost will kick in, I bet ppl will be complaining about cost when 32k launch

[quick dune]

Bypassing the banned words will lead to a mute

[serene cipher]

i get it, bad timing tho, as people talk the split message may end up sounding very unkind

[west talon]

please elaborate what erasmus is

[amber parcel]

I mean it can do highschool level math from what I've tested, nothing crazy though. In terms of a comparison between the free version and the paid version, there's a big difference.

[mystic nest]

API for developers to build applications and services. basically near unlimited access to gpt-4 if you can pay

[midnight tusk]

but bypassing openai's model content moderation doesnt

[amber parcel]

I've tested the Wolframe plugin and it could do JEE level questions

[west talon]

its fine

[amber parcel]

Pretty good

[serene cipher]

sure, be my guest

[tame crest]

Okay if we say the free version is like 3 out of 10 at math. Where is turbo and by turbo do you mean 'default'?

[serene cipher]

you are very right

[quick dune]

I kinda curious. I send you dm because it will considered inappropriate to talk such thing here

[mystic nest]

turbo is still 3.5 there is no different by right, except hella faster

[serene cipher]

indeed, that has nothing to do with Discord, that can only be escalated to an OpenAI account ban

[tame crest]

I know nothing about the plugins. Other than that people on reddit complain about not everyone have access to use plugins/addons. But I now nothing.

[amber parcel]

The default model for ChatGPT plus is GPT-3.5 turbo. The only difference between the free version is speed. I was mainly referring to GPT-4 when I said the difference was large.

[tame crest]

That is why I wanted GPT4. But with only 25 prompts, that's very low value if I want to ask it a lot of math related questions.

[amber parcel]

They increase ChatGPT's capabilities by a lot

[severe hatch]

How’s it going is anyone use ChatGpt to do grant writing and if so would you mind sharing on the experience

[amber parcel]

It seems low but it's enough. Then again, if it's a complex topic and your not very familiar with it, you might need more.

[frail halo]

what is api AI

[silent wyvern]

p[sst, tell him to use Bing ;)

[tame crest]

How can you say it is enough. I could easily ask it like 5 questions about 1 thing, then do another thing and ask 5 more questions. Then I can only do that a couple of times, and I have to wait 3 hours?

[silent wyvern]

Bing has GPT 4 by default and has no limit of 3 hours

[silent wyvern]

Use Bing AI!

GPT 4 by default, and no 3 hours wait!

[amber parcel]

You can just include multiple questions in one prompt. Like I said, it depends on the user.

[tame crest]

So the trick is to have 25 questions at once? lol...

[frail halo]

can some1 explain what API AI is

[amber parcel]

The wait-list though

[silent wyvern]

Sure

[silent wyvern]

then I can do his homework :P

[silent wyvern]

API is just OpenAI communicating with your code directly instead of you going to ChatGPT

[amber parcel]

I don't know about 25, but you can ask multiple questions in one prompt.

[tame crest]

Often you want to ask a question to the answer it just provided. And I have no way of knowing what answer it provides. Therefore it make no sense to just ask 25 questions at once.

I am not talking about I want it to do my math for me. I want to learn, and learning requires dialogue to understand.

[wind urchin]

Hi. Whats Up ?

[amber parcel]

Either way your saving your messages

[frail halo]

so its the same as chatgpt

[amber parcel]

Worse case scenario you just use the free model and see if it's sufficient. You might need to do some extra research but it's better then nothing.

[silent wyvern]

Yeah, but in your PC instead with far less limits and more costly

[frail halo]

how do i reply without clicking i dont use discord often

[amber parcel]

ChatGPT isn't a math professor, so you should use it in conjunction with other resources.

[silent wyvern]

just ping me :P

[frail halo]

define more costly?

[silent wyvern]

It'll cost you $2.somethingcenty040 if you run it 2 hours nonstop

[frail halo]

how will it cost u money on energy or what casue i don't understand

[amber parcel]

You get charged based upon the amount of messages you send

Unlike ChatGPT where it's a monthly payment

[serene cipher]

the price of the API usage depends on which model you use.
each request is priced individually and scales by the amount of used tokens on that request in particular

[frail halo]

o like, but you can get it for free right

[amber parcel]

No it's paid

[frail halo]

like that*

[amber parcel]

You pay based on your usage

ChatGPT is free if you just sign up

[serene cipher]

new accounts get $5 in free credits

but that is it

[winged jay]

How

[serene cipher]

you get it when you register and verify your phone, it will be automatically there

[tame crest]

It would have been so good with endless prompts. Damn it would have been fun.

[wind urchin]

Yoo..I'm new here....can anyone tell me how to use it?

[silent wyvern]

check out platform.openai.com for step by step tutorial :P

[silent wyvern]

Bing AI!

[frail halo]

Is API better then CHATGPT for like coding programs

[serene cipher]

yea, for sure there are a lot of potential bad things to come alongside the good, but i feel positive about it and I think the positives far outweigh the negatives, specially due to the fact that the bad usages are actively combated by everyone

[amber parcel]

Definitely but life ain't fair lol

If your a general user, use ChatGPT. If you want to provide an AI based service or use GPT in your applications, use the API.

Pretty simple.

[serene cipher]

not much difference, chatGPT's defualt settings combined with the UI that embeds cobe blocks are very good for generating code

[silent wyvern]

Life is fair, we make our own luck.
Life is just, we make our own perspective.
Confidence isn't real
Everyone lies.

If there's a problem, it's your fault for the problem appearing to you and not to the random sucker somewhere else, take matters into your own hands and fix it

xd

[amber parcel]

[silent wyvern]

that's a quote, what u think?

I didn't think of it!

[amber parcel]

It's motivational

[silent wyvern]

Yes, agree

[amber parcel]

I'd rather just watch David Goggins or Tom Platz though

Haha

[wind urchin]

i apologize if this is in the wrong location. I've been using GPT-4 for helping generate content for my DnD campaign. For the most part, it's been pretty fantastic. However, I have noticed that it "forgets" it's own work. For example, I asked it to come up with names for continents. After a while of continued work, I asked it to remind me of the continent names, and it provided a completely different list. Is there a better way to ensure that sort of thing doesn't happen?

[silent wyvern]

Ross Scott's monthly streams are 👌 tho

[tame crest]

And what is the price for that or does it work?

[amber parcel]

If I were you I would just include some context

[mystic nest]

https://openai.com/pricing

[amber parcel]

ChatGPTs memory isn't infinite unfortunately

[silent wyvern]

Check out #1021130377026351105

[vocal parrot]

just noticed that if you have the browser plugin, gpt-4 now has the option of browser access

[serene cipher]

this happens when your text is too long, the model do not contain the whole of your conversation, just what it can fit within the context space

[amber parcel]

You first need to get access to the GPT-4 API though.

[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.

[bright mirage]

So is GPT4 worth the subscription price compared to the free version?

[amber parcel]

Depends on the user. What are your goals?

[bright mirage]

Nothing in particular just yet, just curious

[serene cipher]

no, not at all, GPT-4 on ChatGPT is jsut a very limited preview with 25 message per 3 hours limit, I strongly advise against signing Plus if you only wants GPT-4

[mystic nest]

my default suggestion is always, if $20 is nothing too much, just try it for a month

[wind urchin]

Gotcha. I think I was hoping that it'd be able to at least "remember" on a thread by thread basis. I'll rework that premise then

[amber parcel]

Yeah unless Plugins come out

Don't sign up

[serene cipher]

the subscription on ChatGPT is for the pourposes of getting reliable access during peak hours and faster responses

[bright mirage]

25 messages per 3 hours? Yikes

[serene cipher]

with some sprinkled bonuses such as getting to use GPT-4 a bit

[mystic nest]

$20 is to support the free users too 🤣 some ppl need to share the bill

[amber parcel]

Pretty much lol

[bright mirage]

I was gonna see how well it compares as a writer's aid/inspiration engine

But if I'm getting that little, perhaps not

[amber parcel]

The free model is fine

[serene cipher]

sadly no, the GPT-3.5 limit is about 4K tokens, which iI would say.. is about 4 statblocks worth of content =P

[silent wyvern]

Imagine if we had to use GPT-3 to play chess with aliens in order to save our human race, how screwed would we be?

[bright mirage]

if access to GPT-4 is so limited, what's the rest of what you're paying for?

[silent wyvern]

gpt 3.5 turbo and future features

[serene cipher]

the subscription on ChatGPT is for the purposes of getting reliable access during peak hours and faster responses

[mystic nest]

no worry...gpt 3 will cheat cos it doesn't fully understand the rules

[amber parcel]

Speed and guaranteed access during peak hours.

[amber parcel]

Priority access to new features too

[silent wyvern]

👌

[bright mirage]

I was kind of curious whether the newer model has a larger knowledge base or is just smarter about working with the same base

[amber parcel]

Same knowledge

Unless you're talking about GPT-4

[bright mirage]

I am, yes

[serene cipher]

it has basically the same training dataset, but the model itself has a "bigger brain", so it is ""smarter""

[amber parcel]

Alright then yeah listen to lugui

[bright mirage]

Does it make fewer errors?

[amber parcel]

Yeah

[serene cipher]

yep, it makes fewer errors

waay fewer

and is better at comprehending the nuance of the text

[bright mirage]

See, that's the kind of thing that sounds like it would be really useful to bounce ideas back and forth for writers

[amber parcel]

This is a big plus lmao

[bright mirage]

but if I'm only getting 25 messages every 3 hours, then pffft

[amber parcel]

The free model is still pretty good at that type of stuff

[bright mirage]

Probably not worth $20/month just to satisfy my curiosity about the comparisons, then

[serene cipher]

i have a good example here: #1070954877779464253 message
GPT 3.5 and GPT-4 respectively

[amber parcel]

If you really wanna know I can show you some responses in private.

[serene cipher]

indeed, it also results in the model less frequestly (actually, almost never) misidentifying your requests as "potentially content policy violating" and refusing to reply

it is better at "the gray line"

[amber parcel]

Most definitely

[serene cipher]

GPT-4 on ChatGPT is not the main product of the subscription
people who applied to the wait list are slowly getting granted access to GPT-4 API where they can use it basically without limits (since you pay for each request)

the wait list is on #announcements in case you want to join it

[silent wyvern]

Use Bing AI :D

[bright mirage]

I heard Bing was either liable to go rogue or just neutered to hell

[serene cipher]

bing AI is indeed GPT-4, although it has the fine tune of Microsoft

[sharp anvil]

SA

[amber parcel]

Bing, more like Bingus

[sharp anvil]

AS

[mystic nest]

I just treat bing chat as search engine, not really chat bot

[amber parcel]

Isn't there a wait-list for Bing?

[serene cipher]

yea, thats why they implemented a "conversation end" the AI ends the chat if it detects bad content lol

[mystic nest]

bing has a temper 🤣

[silent wyvern]

Bing it!

[mystic nest]

if you repeat the question, it will kick you off the chat (though still able to create a new one)

[silent wyvern]

Epic Rap Battles of History:
GPT-4
VS
Stockfish
Begin!

[gleaming bone]

is there any reason to follow the gpt4 end user license agreement for distilled instruction datasets, given that openai themselfs claim that such licenses are not applicable to models trained on such data?
or is their stance that pre-training is fine but finetuning not?

[bright mirage]

Wait. Does this mean GPT-4 is slowly being rolled out to the free users, or only the subscribers?

[silent wyvern]

No. He was talking about the API

[bright mirage]

Oh, I see

[silent wyvern]

API is the devs's tuff to intergate GPT 4

[bright mirage]

missed those three little letters

[silent wyvern]

haha you got hyped didn't you

[wraith matrix]

I do think that distilled instruction datasets are allowed by openai but you’d have to check precedent probably

[bright mirage]

a Pay-As-You-Go model. I see....

[silent wyvern]

you can use GPT 4 through me if you want (:

[proper glade]

I want to try gpt4

[amber parcel]

Phind is pretty good, although it's more developer oriented.

[proper glade]

How??

[steady holly]

how is gpt4 better than 3

[gleaming bone]

@wraith matrix yes im pretty sure that what they say. the problem is that gpt4 was aswell trained on stole / copywritten material which openai happily ignores

[silent wyvern]

Give me your prompts, I'll give you the results !

[wraith matrix]

Nah they meant the license didn’t apply to models trained on distilled instruction datasets that use output from gpt4

[maiden wyvern]

/img free

[atomic rain]

How can I increase the number of messages I can send to GPT4?

[proper glade]

Can it be an image

[atomic rain]

25 is too low

[slate kindle]

you need the API then

[quick dune]

25 per 3 hours isn't that low. 💀

It's not 25 message/24 hours

[atomic rain]

VERY VERY little, you can't even imagine

[amber parcel]

It really isn't that bad

Unless your spamming it useless questions

[serene cipher]

the ChatGPT subscription has nothing to do with the GPT-4 waitlist

[quick dune]

I did because I'm GPT4 users too

[amber parcel]

You need to be efficient

[silent wyvern]

not yet sadly

[amber parcel]

Image input is still being looked at

[tame crest]

Do you ever think anyone will succeed with a business model where you have to buy a sub + tokens to even use the product?

[serene cipher]

hey, if you do so, please go on #1050184247920562316 to avoid spamming discussion channels

[proper glade]

It's ok 👍 thx anyway

[silent wyvern]

oh no no, we'll do that in DMs :)

[amber parcel]

That's why all the heavy hitters rely on investors

Phind is a good example

[bright mirage]

Is it 25 input messages, or 25 responses from the bot?

[amber parcel]

Input

[bright mirage]

a.k.a. if you regenerate a response, does that count toward your 25?

[amber parcel]

Yeah

The responses don't count, but if you regenerate an answer it does.

[silent wyvern]

Bing AI has no limits like that 😁

[amber parcel]

Relax lol

[red wagon]

GPT4 is going to be free anytime soon?

[amber parcel]

No

[red wagon]

😦

[bright mirage]

Oh, why do they force you to Microsoft Edge? That's annoying.

[silent wyvern]

it already is free! Bing AI 😁

[amber parcel]

Although they're services that allow users to use GPT-4 for free

[red wagon]

Bing AI is trash

[mystic nest]

if you are into larger scale software development professionally, you will know it's actually common, though some charge very expensive license fee instead of pay as you go

[slate kindle]

they want to give a chance for free to try GPT4 eventually
but probably not very soon (and very limited)

[silent wyvern]

you're hurting it's feelings

[red wagon]

even i i'm more intelligent than Bing AI

[silent wyvern]

stop it

[red wagon]

at least for coding

i use LLM mostly for coding

[red wagon]

gr8 info m8

[jaunty raft]

i want to learn a new language. How can gpt help me with that?

[amber parcel]

Yeah

Will it replace an actual teacher? No.

But it can help

[red pollen]

what language do you want to learn, dickweed?

[amber parcel]

Slightly aggressive

[red wagon]

maybe it can help you with writing practice
like duolingo or something like that but more dynamic

[amber parcel]

[silent wyvern]

DM me :)

[red wagon]

nevertheless, the best way to learn a language is by watching movies, and music in said language

[jaunty raft]

french and why so aggressive

[amber parcel]

French isn't that hard to learn.

[hasty egret]

as far as I know (and bear in mind I'm a beginner DnD player), you shouldn't need such an extensive list of continents when generating the content of a campaign. Comparatively, keep in mind we don't have that much continents on Earth.

[spice anvil]

no solicitation on the server

[amber parcel]

You could just use Duolingo and YouTube

[spice anvil]

stay on topic guys

[amber parcel]

Which uses GPT-4

[hasty egret]

it's not literally a memory, but a context that might overflow after a while.

[sand cape]

Has anyone noticed that the web browsing more often than not fails to return a response? there should be some handling of time outs

[mystic nest]

chatgpt can check your sentence like a teacher immediately and tell you the mistakes , your irl teacher cant 🤣 so it's best to have both

[amber parcel]

Yeah it's a common issue

No problem

[hasty egret]

you should be able to bring back the context you were working on by summarizing your last session as a first step in a new discussion.

[red wagon]

i hate this kind of bots

[amber parcel]

They're needed though

[mystic nest]

most classes have like 10 :1 , 30:1 students teacher ratio....almost definitely not enough attention to individual student

[red wagon]

yeah but they need to understand the context of the conversation

we shouldn't have this problem in 2023

[red wagon]

skill issue

😮

[silent wyvern]

Not advertising anything!

[slate kindle]

I dont think the one that got you was a bot (or not sure, a lot happened at that moment)

right after this 😄
#gpt-models message

[red wagon]

it certainly is a bot that recognizes a language other than english and bans you for a little while

[slate kindle]

ah yeah that one yeah

cant even say hello in another language 😄

[mystic nest]

nah, it's not skill, for a teacher, he/she will want to use chatgpt to provide indepth analysis, you dont have enough time to do that for all students manually

basically customisation something for each student

no human teacher can do it for a class of 30*N (diff subjects/diff classes) and not dying 🤣

[serene cipher]

nice save :V

[amber parcel]

I was waiting for someone to point it out haha

[fringe crown]

I agreed 💯

[amber parcel]

Definitely helpful

[wind urchin]

Oh, I know. I was just playing around with prompts to see what it could do. I'd generated 5 continents and 9 deities before It had "forgotten" the continents.

[hasty egret]

sounds like fun

[opaque shoal]

NEWS front given me

[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.

[waxen haven]

hi hi hi

[turbid jolt]

if i pay for plus will i be able to access gpt4 right away?

[mystic nest]

yes, if you mean the chat bot

[turbid jolt]

the plugin support is a different story with a waitlist i guess? fetching external resources

[mystic nest]

yeah

[cobalt parcel]

like ¯_(ツ)_/¯

[pearl musk]

code Interpreter is amazing ！I'm sorry that I can't send pictures here.

[mystic nest]

most ppl dont have the plugin

[serene cipher]

you can on #1050184247920562316 👌

[pearl musk]

If use it for data analysis

[turbid jolt]

unfortunate type of rollout, thanks for the reply

[silent wyvern]

assuming no problems, yea

[red wagon]

i'm about to try AutoGPT

let's see if it solves my CORS problem

[tardy heart]

it likely wont lol

[fickle gale]

lol

[red wagon]

shhh let it try

[fickle gale]

nginx or apache?

[red wagon]

apache

[fickle gale]

I use nginx so can't give advice on that

[red wagon]

surely the problem is some random line of code that has a misspelling

but i spent a whole day trying to solve it and couldn't

so let's give it a try

[fickle gale]

I faced it in my org too. I can give you a direction.

[quiet warren]

howdy everyone

[fickle gale]

You need to provide certain headers, without which the front end wouldn't accept the REST API call.

[red wagon]

i know

when i tested the project locally i haven't had any problems

then i deployed it and the error appeared

[fickle gale]

Yes, apache isn't involved on localhost...

[wind urchin]

Hello Mr Chat

GPT

How can I cook 2 minute noodles in 1 minute?

[red wagon]

cook the half of it

at double the power

[spare sequoia]

Any new updates on chat gpt4

[mystic nest]

you can check #announcements

no special big news

[fickle gale]

I am waiting to get access to Copilot X. The current Copilot with GPT 3 ain't that good.

*3.5

There are a lot of great announcements in the github next page, if you are interested in developer applications

[mystic nest]

how much would they charge later though 🤣 hopefully they will find way to lower the cost

[fickle gale]

They are already charging 10$ month. It's high enough for a specific use case.

or 100$ per year

[mystic nest]

yeah imagine that's for 3.5 I assume?

[fickle gale]

yes

[mystic nest]

so gpt-4 cost like 10x more

if things dont change, then it wouldn't be $10

[fickle gale]

only time will tell. Currently you can get it without any additional charge, assuming you pass the wait list.

and 30 day trial

[slender pivot]

Anybody have code interpreter? Not sure if that's included or not

[amber parcel]

I haven't used it myself but I can probably answer some of your questions

I've tested it through another user

[outer swallow]

is it just me or gpt4 has been downgraded.

[mystic nest]

we hear that kind of thing everyday 🤣

[fickle gale]

if they downgrade gpt 4 we riot

[outer swallow]

I know, noticed the trend on twittert to, it's bound for sometimes to just say something stupid, and the more we relay on it, the more we need to keep more attention. >_>

[fresh lance]

Is it worth paying for gpt 4 chat if I want to program something?

[solar night]

In the OpenAI course, what is dotenv?

import openai
import os

from dotenv import load_dotenv, find_dotenv
_ = load_dotenv(find_dotenv())

openai.api_key = open('key.txt').read().strip('\n')

[wise jackal]

Create a .env file and add your OPENAI_API_KEY to it.
Then use

openai.api_key  = os.getenv('OPENAI_API_KEY')

as you can see in the jupyter notebook.

[turbid jolt]

i just did that, and it's limited to 25 messages per 3 hours..

[fresh lance]

Did that help you in a relevant way? I believe it is a very small number of messages

[ocean solstice]

Whoever wrote this script should not have done it the way they did lol

[turbid jolt]

nope, and i wasn't aware of those limits. it's like a bought into a trial.

[ocean solstice]

Your turbo model is also faster and more prioritized

You didn't buy into a trial, you got exactly what you paid for and exactly what was promised.

[empty jackal]

Sinan Engin

[fresh lance]

I see, I used chat-3 a lot to learn programming for games, I didn't have any programming knowledge and I think that like every beginner, all information is useful, but I don't know if it's worth going to chat-4 now that I have a programming basis

[turbid jolt]

as if i'm going to argue with you 😄

well, do more research than i did.

[fresh lance]

Thanks a lot for the info @turbid jolt

[mystic nest]

most ppl appreciate gpt 4 more when prompting things with more context or more complex albeit the limitations

but of cos there're alway ppl who feel money not well spent (or even scammed)

nothing wrong either

[fresh lance]

my goal with chat was just to learn how to develop games, and chat-3 already did that very well, I didn't just take what he sent me and paste it into the program, I really learned from it, and I think if chat-4 would be more useful than that, and according to what I researched, no

[mystic nest]

well, just cancel the sub 🤣 $20 is still a cheap investment to check out the latest tech

[fresh lance]

or, not yet

[mystic nest]

to see whether if it meet your needs for a month

[turbid jolt]

again, i'm not going to argue. i never intended to go there.

[mystic nest]

I would always recommend ppl to try out a month, then cancel it or not is up to you

nothing to argue, everyone mileage vary

not to say $20 is of different value to diff ppl

[steep raven]

ok

[mystic nest]

🤣

[steep raven]

how to create an account in ChatGPT

[knotty grotto]

Chat.openai.com

[mystic nest]

https://chat.openai.com/ <= if you are in Russia, you likely cant use it, let's assume not

[steep raven]

not working

[mystic nest]

yeah, russia banned chatgpt

[steep raven]

ok

[mystic nest]

"the country (Russia) banned ChatGPT, last November, due to fears of disinformation and criminal use"

[red wagon]

i'm trying autoGPT but it keeps thinking

do you need a high end PC?

because mine has the compute power of a ps2

[ocean solstice]

AutoGPT is not affiliated with OpenAI so we are not able to help with that

[red wagon]

oh

i thought they were

😐

[smoky crest]

Ay I’m back

[red wagon]

ayy lmao

[viscid tendon]

when I ask gpt-4 to generate a file and prompt for a save location, it instead generates a link to a fileshare service (google drive, dropbox) to which it states it uploaded the document. When I click the link the service reports the file doesnt exist. Is this due to the fact it doesnt have an account on the services? Is there a way to have it prompt for a local storage location?

[mental burrow]

how are you prompting it to generate a file?

[ocean solstice]

...GPT-4 cannot generate a file and save it.

[mental burrow]

it can't make files

[shell wigeon]

what does this puzzle role means?

[smoky crest]

Nope

[viscid tendon]

i was asking it to create a food costing template in excel format and prompt me for a save location

[smoky crest]

Idk but I want it

[mental burrow]

it can't create files. Now, if you have access to plugins, I think the Zapier plugin can do something like that.

[smoky crest]

My avacdo role gone

[ocean solstice]

It means one of the OpenAI empIoyees thought I was super cute.

[solar night]

I just succeeded at something 🙂

with open('Wtest.mp3', 'rb') as audio_file:
text = openai.Audio.transcribe('whisper-1', audio_file)

prompt = f"""
Summarize the text delimited by triple backticks \
into a single sentence.
{text}
"""
response = get_completion(prompt)
print(response)

[ocean solstice]

Actually it just means that I have made positive contributions to the community and was recognized by someone important.

[solar night]

Yes is its minor but I made it

[viscid tendon]

ok thanks ill look into that

[shut ferry]

H

[faint bramble]

I have a disturbing result from a request I would like to report- I asked for literature references from a result and the v4 made them up. WHen I questioned the results, it apologized and sent new ones over- again- all made up. I would like to submit this result to someone if possible.

[ocean solstice]

Did you do this in ChatGPT? If so, just click the little thumbs down. This is not unexpected behavior since ChatGPT does not have access to the internet or media with which to give proper references.

[gentle tusk]

Note: All AI-generated content must be carefully vetted. AI can “hallucinate” and generate plausible facts or content that is entirely false yet utterly convincing. Your knowledge as the expert is critical for fact-checking and editing.✍️

[ocean solstice]

Much to the chagrin of people who use ChatGPT to write their academic papers and then get mad when they fail courses and get booted.

[faint bramble]

Jara- Just to be clear, I am not writing an academic paper but testing the AI against my own knowledge. It did report the correct statement but failed to correctly identify where its information was derived.

That is disturbing for a v4 iteration.

I also repeated this in Google's BARD with similar results.

[ocean solstice]

Again, this is normal, expected behavior for any generative AI model.

They are not competent at providing accurate references to anything.

So it is not "disturbing" at all, unless you just aren't familiar with how they work.

[faint bramble]

But isnt it being "fed" information from those same references?

[knotty hatch]

simple and free way to make money in vietnam

[ocean solstice]

It is being trained on information by researchers on those references. It does not have real-time access to be able to reference that so and so said something on page 42 of a book.

[faint bramble]

so there is no current way to verify the information being given?

[ocean solstice]

Please keep in mind that despite how advanced GPT-4 is, it still only "guesses" what would naturally come next.

Yes, there is. Find references through your own research.

GPT-4 is an AI model.

That uses NLP to process queries and responses.

It is considered "generative AI"

[mystic nest]

"artificial" intelligent, so yeah... 🤣

[atomic garnet]

🤣

[ocean solstice]

By saying "it's not an AI," while being technically correct, shows a misunderstanding of what AI is.

[wind urchin]

Can you review the last shared app in #1037561385070112779 and leave feedback?

[hasty gale]

It is NLP that uses DNLN's, could definitely be called AI, considering AI encompasses so many subsets of CS. Its like calling a square a rectangle

[ocean solstice]

One look and I can already see it uses branding wrong.

[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.

[wind urchin]

What is my mistake please enlighten me.

[ocean solstice]

Using the term "OpenAI" for your product implies that you are partnered or affiliated with them, which is against OpenAI's brand guidelines.

Oh

Yours is the second one, my mistake.

[wind urchin]

Yes

@ocean solstice I can't discuss here due to openai discord rules. I would love to hear from you below the app.

[robust mirage]

does chatgpt 4.0 access the internet now? it's saying it doesn't but i hear it does

[ocean solstice]

There is a Browse plugin in a limited beta that allows ChatGPT to access the internet.

[delicate compass]

Pls I need help

I am new to chat gpt, and don't know how to start using it on discord

Sorry if I am disturbing

[delicate compass]

Pls explain this

[ocean solstice]

There is a plugin called Browse in limited beta that allows ChatGPT to search/browse the internet, much like Bing does...there's not much more to elaborate.

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

[delicate compass]

Ok thx ☺️

[robust mirage]

TY

[round oasis]

A board game on english for underprivileged children

[ocean solstice]

Why do you keep posting this?

It really doesn't.

How about we share the definition instead of making up our own?

Saying GPT isn't an AI model because it's NLP is like saying hamburger buns aren't bread because it isn't Aunt Millie's.

Artificial intelligence is intelligence demonstrated by machines, such as perceiving, synthesizing, and inferring information. These are things that GPT models do. That is why they are called "AI models." It's not called OpenGPT, it's called OpenAI.

NLP is a subset of artificial intelligence in computer science.

I think you're conflating AI with AGI which, you're right, it isn't AGI, but it's still AI.

[umbral tinsel]

We are AGI

[ocean solstice]

No...we are just GI.

And I was responding to someone who fundamentally misunderstood what AI is.

[fallow knoll]

All this chat does is have sematic arguments and request refunds

[cloud bison]

Mostly true.

Because most of the discussion is about poorly defined concepts.

[ocean solstice]

Lol

Why do people get so defensive when they are corrected? Good grief.

Even ChatGPT defines itself as an "AI language model."

Sir please do not spread misinformation about AI, thank you.

[cloud bison]

I find the confidence and the arrogance of many people, even when they have narrow technical expertise, rather odd when it comes to this sort of thing. Its a lot of shaped ideas based on principles that they usually don't understand because the need scope is inherently multidisciplinary for the majority of the topics they are trying to discuss.

[ocean solstice]

I am correcting people making faulty assumptions, I've been around AI quite some time, I work with it daily. I think you just need to calm down a little bit.

[cloud bison]

Bamboo?

[ocean solstice]

Yes

[hard ravine]

what about you

[ocean solstice]

I find the confidence of many people to be rooted in arrogance, actually. An inability to conceive that someone else might know a little more about a subject than they do. But that's not really related to GPT-4 so let's get back on topic.

[cloud bison]

Its relevant. I was making the comment in the context of the way that people define various proposed or emergent features of GPT 4.

[fallow knoll]

Anyone with a smart phone has "been around AI quite some time" and "work with it daily" lol

[ocean solstice]

Well that's just it, people want to define GPT-4 on its features rather than its function. It has the ability to learn, reason, and make decisions based on trained data without explicit human input, and that is almost the exact definition of what AI is. :P

[ocean solstice]

Things like Siri were not AI in the sense that we talk of today. Also, using something is not "working with it."

But I guess, technically, sure.

[tough kestrel]

lol are you Yann Le Cun? You seem to know your stuff!

[fallow knoll]

ChatGPT is just a statistical model; it does not reason but predicts the most likely next four characters

[languid oak]

Does anyone know when the APIs are going to be available for Chat GPT plus users?

[ocean solstice]

ChatGPT is a neural network model that uses deep learning to recognize and learn patterns. It can understand and generate human-like responses. This is not something that a plain "statistical model" can do. It is an AI language model. Statistical models alone are not capable of generated complex and contextually-appropriate responses in natural language.

[supple violet]

api and plus are different products for different users

[ocean solstice]

ChatGPT Plus subscribers will never receive any sort of API benefit.

[wraith ginkgo]

`` Server Tip: ``

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

[ocean solstice]

I don't know who that is, but AI is one of my minors and since I work on four different projects integrating some form of AI, I kind of have to. :/

My forte is information security and how AI applies to it, though.

[languid oak]

I meant the plugins

[fallow knoll]

'recognize and learn patterns' is literally a description of a statistical model, and its outputs being 'human-like' is a subjective marketing term with no metric. It is not programed to understand language. It doesn't know what a noun, verb, or preposition is. You simply don't know anything about this at all lol

[ocean solstice]

There is a waitlist available for the plugins, there isn't really a release date for everyone though.

[ocean solstice]

I really think you don't. You're just making stuff up at this point.

But insulting me to make yourself feel better about it isn't helpful or appropriate.

And if you can't talk rationally then we just won't talk, sound good?

[simple echo]

Why are so many people on this server so aggressive and arrogant?

[ocean solstice]

Because everyone is a Google expert.

[shell wyvern]

Encourage respectful dialogue: It's clear that people in this discussion have varying levels of understanding about AI. Encourage everyone to share their perspectives and learn from each other. It's okay to correct misconceptions, but it should be done in a respectful and constructive way. It's also important to acknowledge the limits of our own understanding and be open to learning from others.

Here's a possible response you could use:

"Hey everyone, it's clear we have a lively discussion here about AI, AGI, and NLP! These are complex topics, and it's great to see so much interest. Let's clarify a few things: AI refers to systems that mimic human intelligence, AGI is a type of AI with broad capabilities equal to or beyond a human, and NLP is a subset of AI that involves interaction between computers and humans through language. ChatGPT, the system we're discussing, is a type of AI model that specializes in generating human-like text. While it's powerful, it doesn't have a broad understanding of the world or the ability to reason like humans do—it generates text based on patterns in the data it was trained on. Let's keep the discussion respectful and constructive. We all have different levels of understanding about these topics, and that's okay. We're all here to learn from each other. Thanks!"

[tough kestrel]

It's the inventor of CNNs, and chief scientist at Meta

[ocean solstice]

Oh, intriguing.

[simple echo]

anyways, I tried out gpt4 and it was pretty cool. couldnt use the image feature tho

[shell wyvern]

I asked chat GPT help resolve this argument.

[ocean solstice]

And it pretty much confirmed what I have said over and over again. :L

[cloud bison]

I personally have started to adopt a more agnostic position on a lot of the issues debated here, sometimes on the premise of poor definitions, sometimes on the premise of mechanical assumptions at various levels of comprehension of operation.

The factors that lead to the ultimate output of GPT 4 based on its input are codependant on so many intervening variables and uncertainties, that confident judgements about the mechanism are almost certainly going to be ignorant. Its too bad we don't have a modern Feynman to make some equivalent comment for AI as was made for quantum mechanics "If you think you understand quantum mechanics, you don't understand quantum mechanics."

Those familiar with AI infrastructure and programming are prone to oversimplifications based on the training of the model. Namely, they understand the mechanics of the parameters, and they know what the feedback was that was used to train the model, and how input and output ultimately result in the formation of the model, and assume a sort of complicated auto complete usually. This misses the fact that another known factor in the architecture is the issue of the black box and the mechanisms that it uses to create useful information in the intervening hidden layers, that the patterns seem to expand beyond next word recognition by necessity because the context is broader, and that the formation of modules that aren't mechanistically understood is inevitable, and we don't know whether they do or do not mirror similar neurological biological functions.

The linguists in contrast seem to oversimplify the entire equation... they break down the heuristics of language and claim that without any sensory aparatus the mechanism for anything more complex doesn't exist in language. Psychologists would profoundly disagree, because although highly damaging, organisms in complete sensory deprivation do not experience brain death, and its not clear that they lack all the functions we attribute to cognizance, and their brains don't shut down (Verified by neurologists).

I don't know the profession list goes on but I find each one is full of its own list of fallacies they engage in that show ignorance of the multidisciplinary factors they lack the credentials to usefully discuss.

The remainder are just more broadly ignorant, which is forgivable since they are just glomming on to whatever sounds persuasive in the moment.

[simple echo]

i read all of this but oh my god

[shell wyvern]

Address misconceptions: It's crucial to address any misconceptions about AI. For instance, ChatGPT is not an example of AGI, as it is specialized in language tasks and doesn't have a broad understanding of the world. It's also not capable of reasoning in the way humans do. It can generate text that might seem like it's reasoning, but it's really just predicting what comes next based on patterns in the data it was trained on.

[ocean solstice]

No one claimed ChatGPT is an AGI. :P

[languid oak]

Does anyone know when the plugins are coming for GPT plus members

[tough kestrel]

Yeah that's also how I see things. People always seem to have strong convictions on one extreme of the debate or the other, when nobody truly know what happens inside the black box

[ocean solstice]

But limiting it to a statistical model is misleading to the nth degree

[fallow knoll]

What I said was really nothing more than a basic description of how ChatGPT works. It doesn't understand language at all, contrary to your claims, and again a 'human like response' isn't an objective metric but a subjective quality claim about its outcome, which is also something that was attributed to Eliza

[ocean solstice]

There is not a date that has been given for this, join the waitlist.

[fallow knoll]

Its not you just flat out don't know what you are talking about lol

[ocean solstice]

Okay, we're not talking about this anymore. I thought I made that clear.

[wet drift]

I've seem people who have made essentially chatbots with chatgpt but it doesn't sound like chatgpt it sounds like it has personality traits to fit with the game etc. Anyone know how they direct it that way

[ocean solstice]

Either you can be respectful or we can end it.

[fallow knoll]

Then don't go to a public chat room

[ocean solstice]

Not how this works, buddy.

[fallow knoll]

There's no respect for those who willingly act to deceive

[ocean solstice]

And for the record, I most certainly do know what I am talking about XD

[cloud bison]

What are you defining as pre-requisite to "understanding language"?

[fallow knoll]

It is in fact. Its how public chat rooms work. You are choosing to be here

[ocean solstice]

Admiral, just ignore him. He clearly only wants to argue with someone. Don't feed it.

Let him think he's the foremost google expert and be done with it

[tough kestrel]

Can we go back to more consensual topics, like how we hate OpenAI for betraying open source? 🙂

[ocean solstice]

Eh...

[cloud bison]

Shrug, his comments are adversarial, he lacks tact, and in general doesn't know how to have a cooperative dialogue. But... That doesn't necessarily mean he's a troll, I've met plenty of people lacking social skill who are genuine. So I'm willing to humor it a bit.

[simple echo]

that is gonna make another argument cause it seems that all we do here

[ocean solstice]

I label him a troll because he lacks the ability to critically think about the context here.

[fallow knoll]

Chat GPT is not programed to understand even basic sentence structure. It doesn't know definitions, it doesn't even see words. It sees tokens of four letter charecters. It doesn't see the word Dog; It instead sees a token with the charecters D-O-G-; and uses its large database to view all previous tokens to guess the most likely next 4 token (4 charecters).

Because of the complexity of language, OpenAI abandoned teaching AIs language and instead focused on a statistical model for generating text

[ocean solstice]

Someone who insists and must be right.

[fallow knoll]

You listed 'human like responses' as an objective quality. You can't even begin to think at a basic level, much less than critical lol

[pastel pagoda]

Can we all just get along and be gay? Thank you!

[fallow knoll]

This is the basic fact on how ChatGPT works. It just is. Arguing against this really just exposes your own ignorance lol

[ocean solstice]

ChatGPT is a neural network model that uses deep learning to recognize and learn patterns in input data and natural language processing to understand and generate human-like responses to a wide range of prompts. This is a fact, this is not debatable. Call it what you want, this is what ChatGPT is. It is an AI language model, pure and simple.

[sudden citrus]

why does my bot just stop?

[simple echo]

They literally work in this field of AI, i think they know a bit more about this than you do

[ocean solstice]

Anyone who says otherwise is calling OpenAI a liar, or just doesn't understand how AI works.

[fallow knoll]

?? Except this is how OpenAi says it works

[fallow knoll]

Did you know people go online and tell lies? Shocking I know

[simple echo]

so, the developer badge they have next to their name is a lie?

I swear i cannot with this today

[ocean solstice]

My developer badge just means I developed a Discord bot.

But...that Discord bot does use AI, so...

Built with ChatGPT, GPT-3, and GPT-4.

[onyx raptor]

How many servers are y'all in??

[fallow knoll]

Go ahead and look up OpenAI's interview with linguist Dr. Erica Brozovsky for PBS "From Alan Turing to GPT-3: The Evolution of Computer Speech | Otherwords."

[delicate compass]

Omo how do I create a bot

[ocean solstice]

https://openai.com/blog/chatgpt Go look up ChatGPT's introduction and how it explains the AI model.

[frigid basalt]

where

[ocean solstice]

To create a reward model for reinforcement learning, we needed to collect comparison data, which consisted of two or more model responses ranked by quality. To collect this data, we took conversations that AI trainers had with the chatbot. We randomly selected a model-written message, sampled several alternative completions, and had AI trainers rank them. Using these reward models, we can fine-tune the model using Proximal Policy Optimization. We performed several iterations of this process.```

[fallow knoll]

Doesn't change the easily googleable facts on how ChatGPT works. It doesn't understand language but works off a stastical model. Anyone who says otherwise is the type ot try to tell you crypto

[ocean solstice]

Yeah, let's just ignore Logan. He clearly just wants to argue and has no intent on understanding how AI or these models work. :L

[fallow knoll]

None of this changes that ChatGPT doesn't understand language and works off a token system. You just don't know what you're talking about lol

[ocean solstice]

You're using words that have meaning and applying your own to them. Stop being aggressive and argumentative.

ChatGPT uses natural language processing, that has a meaning. Whatever meaning you're applying to "understand" language is not what NLP is.

[mystic nest]

was away for quite a while, you guys still on this topic? 🤣

[fallow knoll]

No I'm again directly quoting OpenAI here. Sorry that you want to pretend to be an expert than spend your time actually being one

Beware grifters kids

[ocean solstice]

You can't even describe what you are implying when you say "understand language." To call anyone else out when you can't even back yourself up is ridiculous. You're causing an issue, just stop.

[vocal quest]

Maybe both of you are right but using different terms?

[ocean solstice]

I'm using correct terms and he can't define his.

[tough kestrel]

I asked ChatGpt 4 who is right in this debate, he told me but I'll keep it for myself

[fallow knoll]

Incorrect.

[simple echo]

So is Jara. that page up there is literally from their blog itself

[glad wolf]

prompt for lion in jungle photo of

[simple echo]

so unless openai is contradicting themselves, why still argue this

[spice anvil]

i think it's better to move on 🙂

[simple echo]

thank you

[fallow knoll]

And it in no way invalidates what I said. No where does it say that ChatGPT is programed to understand vocabulary, grammer, or etc. All it says in that block of text was that it used human feedback to improve its responses. The fact remains that ChatGPT does not see or generate words but 4 character tokens based on a statistical approach using a large corpus. It has no vocabulary, understanding of grammer or syntax, or any understanding of the framework of language. Anyone who tells you otherwise is lying to you to sell you a version of the future that isn't real

[knotty grotto]

☝️

[delicate compass]

Omo pls can I use chatgpt to create websites, sorry if d questions are kinda weird

[fallow knoll]

Yeah let's just let app developers telling lies continue to sell their products based off false promises :

[tough kestrel]

Exactly, move on to general release of plugins and API

[knotty grotto]

Again, Let's move on please

[simple echo]

yeah you can

[mystic nest]

if ppl already have an answer to themselves, nothing will change them, so let's move on

[spice anvil]

not sure what promises you mean here, but if anyone wonders how it works, the research papers are available for everyone to see

[cloud bison]

I'm not convinced that the human brain does much better than that. I don't see a persuasive argument here allowing for the dismissal of similarity of methods of comprehension.

Our neurology prefers "lumping" of data as well, its the reason we can remember a longer string of numbers if we lump them into 3 digit numbers, or if we create stories and meaning behind the concepts, we don't actually remember character by character, or general ideas either.

The associative nature of tokens is actually key, for the neural network similar tokens are similar to tokens used in similar contexts. It knows that cats and dogs are closer in human comprehension then cats and pats for example, despite one having more characters in common.

In addition, you can use VERY simple models that do something similar to advanced models to do things like vector search for advanced comprehension.

For example, you can get a model to do something like "The male leader of a monarchy is a.." complete "... King" and then change that to female leader and it will complete with something "Queen"

This sort of model takes FAR less then inference models do that communicate, this happens as low functioning as basic vector database AI's that a lot of these models rely on during their training early on to keep context clear and allow successful weighting of their nodes.

The oversimplification of "tokens" is not doing much good, because in the human brain we basically do the same thing, we have in fact identified individual neurons on human brains that correspond to individual important things in a persons life, like their own name, or the identification of an object they like, or a person in their family. They are highly redundant and can reform if damaged because of the overwhelming amount of context connections, but they can most notably also be triggered by the stimulation of just that one neuron in a reverse cascade.

Humans tokenize in other words.

[ocean solstice]

Available at the link I sent, actually ^^

Can also find the index at https://openai.com/research

[simple echo]

160 pages of papers, oh my

[ocean solstice]

Yes there is quite a lot of research from OpenAI

But then again they are a research organization, aren't they?

[simple echo]

true

[tough kestrel]

No they're capped for profit

[ocean solstice]

...that doesn't mean they aren't a research organization :P

[tough kestrel]

yeah but if they only do research they can do the profit! 🙂

[ocean solstice]

OpenAI identifies as an "AI research and deployment" company :P

Well not true. Their mission statement is to create safe and powerful AI that benefits all of humanity. They also have a "not-for-profit" branch.

[spice anvil]

okay, let's not start arguing about what OpenAI stands for or their purpose 😅

[ocean solstice]

Being capped profit means that they can focus on research that benefits society as a whole ^_^

[knotty grotto]

Alright I'll make the first move
Chatgpt 4 free when ?

[ocean solstice]

Never I hope

[knotty grotto]

Neko: One day maybe

Thanks neko

[ocean solstice]

I mean, I wouldn't mind a cheaper GPT-4 API fee model but...

[mystic nest]

where's my plugin?~
someone else : go apply to the waitlist and wait!

[tough kestrel]

There is a lot to argue about on this topic lol but I won't

[cloud jolt]

hey guys, i dont know if im the only one with this problem but i can't access chat even tho im a plus subscriber. Keep getting this error: Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[knotty grotto]

When picture with gpt 4,
I read the #gpt-4-faq I want a date, time who is working on it, if they have 1 or 2 golden fish. Name of the fish

[ocean solstice]

Discussion or argument :3

[cloud jolt]

works on mobile , really annoying

[cloud bison]

The distinction you're trying to make is private and public research, with academia claiming that public research is less likely to be biased because of lack of financial motive.

Of course, speaking as someone who has been involved in academic research, while I think skepticism of academics is overblow, there are definitely places of bias in the social sciences where ethics committees act as gatekeepers for a lot of research, and to say there isn't financial motivation is also absurd considering the way that grants work, and the fact that the entire paradigm is centralized around mostly private universities.

When it comes to AI and social sciences, the truth is that the private sector has a SERIOUS advantage over academia in high quality research because of sample size and cost to run experiments. There is simply no possible way academia can compete, and really, the only reason it "seems" to is because most companies are incentivized to keep their research hidden from one another for competitive advantage.

[knotty grotto]

Try firefox as test

[fallow knoll]

This is not true, and goes against a century of linguistic study.
Here's are two sentences developed by Noam Chomsky that showcases how this is wrong:

The fact that you went for a walk this late without an umbrella, even after hearing the weather report on the radio this afternoon.
You can tell this sentence is wrong intuitively because the first three words of sentence made a promise that was left unfulfilled by the previous sentences. Your brain remembers this commitment even after the numerous clauses.

Colorless green ideas sleep furiously.
Your brain understands this absurd, statistically unlikely sentence despite the fact that you've never heard this combination of words. The order of the word categories is correct, even if the chance of these particular words going together is unlikely.

Our brains are hard wired for grammar; we have ideas independent of language, and we use grammer as the container to deliver thoughts to others.

[ocean solstice]

Most of my academic references list biases too, which is nice to clarify.

[ocean solstice]

We hath moved on, my dear friend.

[fallow knoll]

Nope.

[cloud bison]

He's responding to an earlier comment I made.

[ocean solstice]

I know but still. Oh well, no matter.

[fallow knoll]

Sorry I'm interupting your lies ❤️

[ocean solstice]

Anyway, I'm digging the GPT-4 api

[cloud bison]

In any case, all this proves is that there's circuitry for comprehension that could well be located in the hidden layers in this case. In fact, as current models become more developed, the goalpost keeps getting moved because they tend to be able to challenge the notion that they don't comprehend these examples, and whatever examples you throw at them.

[ocean solstice]

Its responses are so much more in-depth than even ChatGPT.

Not that ChatGPT was bad, but GPT-4 just elaborates so much more efficiently.

[cloud bison]

The Winograd Schema Challenge is currently not working well against GPT 4.

[spice anvil]

guys, this is not a one-on-one chat, each of us can express our thoughts without being disrespectful towards each other

[cloud bison]

And nor are any of the examples Chomsky uses.

Of course these experts do proclaim that its still not doing what they say, but they are increasingly running out of challenges to pose to demonstrate that fact.

[ocean solstice]

At this point, it's bordering on libel. He's called me a scammer, a grifter, a liar, and any other such nonsense. Ignoring isn't working so well. -_-

[knotty grotto]

I'm pretty sure some did a prompt in chatgpt 4 to say now you are chatgpt 3.5 a weaker version of chatgpt 4

[cloud bison]

I'm not even saying that they are wrong, what I am saying is that the lack of humility and the claims that these examples serve as evidence that they don't understand is dubious, and you're building your foundation on quicksand.

[ocean solstice]

I will note that GPT-4's API does not know it is GPT-4 and still thinks it is GPT-3.5

[knotty grotto]

He paying to nerf his own gpt 4 to act like gpt 3.5

[cloud bison]

It is just an extension of goalpost shifting and poor definitions.

[knotty grotto]

Yeah since database end in 2021

[ocean solstice]

ChatGPT's GPT-4 model knows it is GPT-4 though.

I mean, perhaps I can tell it that in the system message...

[fallow knoll]

You are pretty much just spitting nonsense now lol. Even OpenAI is quite public about the fact that ChatGPT is not programed to understand language, just generate statistically likely text.

It can't even see the word Apple, much less know what an apple is, and grammar can't exist without an understanding of the objects and actions described

[spice anvil]

good catch, which example did you use?

[cloud bison]

OpenAI developers and the leader of the company also freely confess to not fully understanding what is happening in the hidden layers of their own models, and having concerns,.

[fallow knoll]

This is called marketing

[cloud bison]

I would provide a link but alas... you can just google a list from NYU.

[spice anvil]

what I've seen so far was the city council/demonstrator one, and it explains both cases

[fallow knoll]

Maybe you can post links here, but we can't lol

[cloud bison]

I'm not convinced OpenAI has solved the black box problem, and I would be very skeptical of anyone who said they did. I think its much more plausible they don't understand the precise mechanism behind the topological connections generating the outputs in their own models, and I'm fairly sure the same is true for most companies.

[ocean solstice]

That's a fair assumption, since these models are designed to draw patterns without requiring a human to link it for them.

[tough kestrel]

I've tried a few it seems to do well. Don't you have a particular one in mind?

[cloud bison]

It actually reminds me of an argument I used to have with theists on the existence of evil gods, and the fact that a programmer can control all aspects of their program, but that doesn't necessarily mean they have the competence to get the output they desire, despite unlimited control, and full access.

[ocean solstice]

Not enough to claim ChatGPT is autonomous but it does have just a little autonomy in how it processes language and draws patterns. If it didn't, we wouldn't find so many hilarious mixups!

[knotty grotto]

As an AI I... ( it's a joke don't hurt me )

[mystic nest]

human brain is also a blackbox but we still use it 🤣 hopefully everyone does

[cloud bison]

No I was just pointing out it does well on them, awhile back when it first released I tried a dozen or so, and it got 11 right, and the 12th one somewhat.

The point being, the ability to make a thing does not translate to the ability to comprehend the mechanism that generated every aspect of the thing, even with unlimited control. Emergent properties are finicky like that.

[tough kestrel]

Ah ok I misunderstood, i thought you were saying it doesn't do well

[spice anvil]

no, no it can definitely understand and give you some explanation for both cases

wrong ping sorry

[fallow knoll]

Let's see, what's more likely:
That a statistical model evolved into something well the capabilities of its framework in a way never seen before
or
That being vague about its capabilities is good for funding and interest

[ocean solstice]

Assuming people lie to get money, especially in a company that is not for-profit, is ridiculous.

[fallow knoll]

OpenAI is no longer not-for-profit. You really don't know anything, do you lol

Marketing exist
This guy: impossible!!

[ocean solstice]

I didn't say it was not-for-profit, I said it wasn't for-profit, again, please quit trying to insult me.

It is capped-profit.

[fallow knoll]

Which is a for-profit by defintion lol

Also as a director of finance for non-profits, marketing exist for us too lol

[ocean solstice]

...no, for-profit implies a goal for unlimited profit. Capped-profit has a mission and focus, specifically, on research and creating AI that benefits all of humanity.

@spice anvil I've really tried to be nice here. Do you want me to just leave the channel for the time being or?

[cloud bison]

Probability arguments can go both ways when stated without concrete metrics.

For example, in general, evolution and competition punish inefficient designs that can be done more efficiently. In our early evolutionary history we competed against other hominids that had higher functioning brains, and it stands to reason that faculties we have such as consciousness, cognition, sentience, and complex reasoning were all given evolutionary pressure to be made properly and efficiently, with hominids that did so more efficiently prevailing.

There are counter-examples, but they are rare, and it is unlikely ALL of our emergent mental properties are needlessly complex.

We regularly prompt chat bots with problems we developed these exact mechanisms to be able to adapt to and solve.

Is it more likely that we have an unnecessarily complex set of emergent properties that an AI less than a decade old managed to render obsolete and superficial in no time through increased efficiency, or that it is convergent on similar simplistic solutions to the problem?

I lean towards the belief that we are overcomplicating and aggrandizing human emergent properties, and those who think we are special are falling to a "faith in the gaps" argument.

[knotty grotto]

Rule 1 apply to everyone. Be respectful please

[tough kestrel]

By the way chat gpt was also able to tell me the questions are coming from the Winograd Schema Challenge when i asked for it. Is it cheating?

[fallow knoll]

If you don't see an issue with someone continually spreading bad information, then you are part of the problem

[knotty grotto]

Again #rules 1 apply to everyone in this discord.

[fallow knoll]

Then its a bad rule if it protects speach that is actively harmful

[distant zephyr]

Well, I've applied for GPT-4 waitlist, and that was 3 weeks ago, and I'm still waiting. any help?

[knotty grotto]

You can say someone is wrong without insulting. If you need to insult you lost all your argument

[distant zephyr]

My question is, does it really take that long?

[fallow knoll]

No you haven't. Shame is the most powerful human motivator and the only way to get people like this to stop talking about things they know nothing about

[knotty grotto]

By joining this discord you agree on #rules .

[quaint crown]

Problem isn't that binary simply .. 99% of your knowledge is based on repeated story ... nothing is exactly what facts seems to tell

by your logicwe sould hang all scientist who wrote false claim by decades

[fallow knoll]

Then they are bad rules, but the constant state of this chat is proof of that lol

[quaint crown]

so avoid judge people is a primary reason to build a society

[cloud bison]

It is if its not reasoning through it, one might need to come up with examples it can't find online, which isn't terribly difficult.

That said, the way it memorizes things leads me to believe it will probably do fine since it is not calling on training data, but rather, calling on weights in its neural network, which means it has the context for that type of question, not just the specific list of questions.

[maiden sleet]

It definitely can. I signed up the first day and got access about a week later - a friend of mine who signed up a few days later is still waiting. Probably depends on your answers to the questions, some luck, and when you signed up

[fallow knoll]

No let's judge more people. Normalize calling idiots idiots

[knotty grotto]

If you have issue with rule , dm @open crescent . But for today you agree on those and they are for everyone.

[distant zephyr]

much appreciated fam!

[knotty grotto]

Have a civilize discussion without insult.

[fallow knoll]

This is harassments and boarder lining on slander. You aren't following Rule 1 by allowing me to dunk on this fool

[quaint crown]

99% of people judge that US american citizen are the worst problem of humanity ... are you sure about your ideas ?

[knotty grotto]

If someone disrespect you please report them. By dm modmai or using the right click, app, report. But no insult here is tolerated

[fallow knoll]

I find this very insulting how dare you talk down to me like this

[knotty grotto]

Then you have the right to report Me and a mod will see the case and be the judge of that. Let's go back on topic , and let's not go in insult mode

[quaint crown]

Hello Jara .. one of feared reason is because you need to create a chain of logic among version .. hence GPT3 answer are corrected by GPT3.5 then corrected bbby GPT4 ... etc ... otherwise you would obtain GPT4 telling you all GPT3 told is garbage ...

[ocean solstice]

Oh yes I know, I just haven't bothered to do so. Though I probably should, my users get confused when I say "Powered by GPT-4" haha.

[wraith ginkgo]

<:book_icon:1046291503158399107> `` Rule 1 `` Be respectful.

Treat others the way you would like to be treated, and assume best intentions. Don’t harass or attack others, and don’t engage in hateful or generally malicious behavior (e.g. sexism, racism, homophobia, etc.). Keep the negativity to a minimum.

[quaint crown]

I find Logan is setting trap for others by flooding the channel for a whole 30mn and will use any excuse to report others once het got its own medecin .. I just got reported from Logan while I barely even talking to him and only applied his own logic to him... toxic people

[cloud bison]

It really seems to depend a lot of whether you're a sophist or trying to educate.

A sophist will pick the most incendiary and provocative speech in some cases, because they aren't so much trying to persuade those who disagree, as they are trying to galvanize a minority who already agree with them.

An educator will tend to adopt a set of speech that uses the definitions of the opposition, and posits questions that will either reveal flaws in their reasoning, or lead to the correct conclusions on the issue whether it is theirs or their opponents view (Sometimes just drilling down to the underlying assumptions and values).

I don't really think your mode of communication is terribly effective at either since there isn't much of a minority here taking your side to galvanize you.

That said, the principal of protecting speech does necessitate protecting even insulting speech (Though this is not a public venue, and thus, the principle is irrelevant, as private venues are allowed to censor within their communities and locations). I think the argument of protection all speech is well founded, mostly because I think its pretty important to allow people to debate the questionable things regardless of the character of the person. However, aside from being "technically" right in that limited sense, I see very little useful contribution from you other then adamant insistence of the right to communicate. In other words, your arguments are not well founded because your approach is poor, and the only value you are upholding is the value of freedom of speech (Where the principle of protecting the least popular speech is arguably more important than things we all agree on).

[ocean solstice]

Kind of why I stopped responding. Better to just ignore.

Besides, not super relevant to GPT-4.

[cloud bison]

That I agree with.

[knotty grotto]

Action has been taken by moderators, Let's focus on gpt 4 topic

[ocean solstice]

I need to split up my history handlers.

As it stands, all three of the models I'm using, including GPT-4, use the same function to pull history of the chat XD

[quaint crown]

you shouldn't do that ...

[ocean solstice]

The problem is, that's the same function that assigns the system message.

[quaint crown]

most AI applied a distance function , in order to solve a gradiant ... by splitting .. you are cancelling the whole purpose of it

[copper anchor]

I just purchased chat GPT plus, and now on version 3.5 and 4 I get the error "Something went wrong. If this issue persists please contact us through our help center at help.openai.com." is there anyway to remedy this?

[spice anvil]

i don't think there's one correct way how to prevent that

[quaint crown]

you should feed a single block of ( whatever source of your history ) -> create a mono block -> handover to AI models

[knotty grotto]

Try firefox as a test

[spice anvil]

for some users both suggestions over this one, works

[verbal burrow]

Refresh

[frigid fractal]

@fallen wigeon this is incredible if i can get it to work. I uploaded the HTM file, but I don't see "creative mode", how do you get there

@fallen wigeon only see these three modes: chat, compose, insights

[dire verge]

Hi,
Got a quick question.
Can GPT-4 analyze images now?

[lyric thorn]

Is there a mobile verson because i baught the 20 dollars a month package

[knotty grotto]

Nope

[knotty grotto]

Not yet

[sweet thistle]

Hi there! This feature is not yet available to the public

[hasty egret]

you can use your mobile browser on your phone.

[sweet thistle]

I just made a shortcut to the website on my homescreen

[dire verge]

Thank you!

[dire verge]

Thank you

[lyric thorn]

Thanks for telling me that I went to the App Store looking for the app and I cannot believe that I found 19 different apps that are trying to be chat GPT it's becoming overwhelming at this point

Charging crazy prices 100.00 lifetime not bad but the results are not on this level i don't think

[sweet thistle]

There is no official ChatGPT application. Please be careful when looking for them 😅

Do note that they can exist through use of the API, but these are still community made.

You may find one in #1037561385070112779!

[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.

[mystic nest]

its not wrong to call themselves "chat" and "gpt" 🤣 if no one owns the copyright/trademark

which openai is trying to do

[hasty egret]

you should be super careful, and as much as possible let others people around you know about that. thanks 🙏

[mystic nest]

but yeah, too many scam app

[opaque umbra]

hi everybody

[smoky crest]

There are a ton of fakes