gpt-models

[spice swift]

There will be new skills, it's about the skill only, doesn't matter with being dumb

[torn coyote]

only get it if you want to use chatgpt 4 and are a frequent user of chatgpt

[maiden raft]

early access, available when full capacity and some other stuff

[wet halo]

For clarity - the system is supposed to save conversations for free and paid - the history is just offline at the moment for some folks.

[tribal harbor]

well looks like ChatGPT is down for a while... as a premium subscriber this absolutely sucks and is screwing up my morning dev session

oh well, looks like I have to use my brain

[fringe nebula]

does chat gpt work for europe now?

[maiden raft]

Mhm was worst

[spice swift]

Well, exercising your brain once in a while is good tho :v

[maiden raft]

It works in norway

[fringe nebula]

Ah, it started to work for me too. Just there is no chat history

[maiden raft]

Same with me no chat history but its probably there when i wake up tomorrow

[uncut jackal]

Again, people have been spouting the same thing since the dawn of time - about books, about the internet, about wikipedia. Your argument can also be used as "Why have the internet? It just makes people be lazy and google everything they don't know, thus we lose a skill." Every time there's a new technology, there are naysayers that say "Oh but it makes people lazy, the new generation become entitled, etc" but the truth is that it's there to expand our knowledge and how we approach things. I know of several of my peers that are already using chatGPT to help them in their research in their fields, does that make them lazy because they've used an AI to help them? Surely you want more easy access to information, no?

[upper frigate]

Switch to a different website that uses chatgpt

[spice swift]

Oh I got the history already

Nice

[tribal harbor]

can you give example?

[quartz grotto]

Anyone else getting Bad gateway upon logging in ? I'm on paid plus plan

[upper frigate]

Poe.com, Bing chatbot

[spare silo]

No

[sweet thistle]

If you have API credits, https://platform.openai.com/playground?mode=chat may be a suitable option for you
for other ChatGPT alternatives maybe ask in #ai-discussions

[spice swift]

Yeah agreed with you, it just the way of evolution and advance in technology

[hexed nexus]

I am! on payed plan as well...anyone knows what's going on?

[tribal harbor]

Thanks! Just trying Poe.com this will get me through

[mellow gust]

Same here. BAD GATEWAY.

[quartz grotto]

ok phew so its an outage, not a ban, lol

[errant ether]

me 2

[slim garnet]

you can't go backwards

[glacial escarp]

wow, I actually go to sign up for chatgpt plus and of course it's the worst possible time lol

[tribal harbor]

Thank you, ran out of credits but I imagine they're not too expensive as a fall back

[upper frigate]

👍

[spice swift]

What do you mean by that?

[upper frigate]

Agreed, lazy people will already have been trying to cheat their way. This just helps the hard workers to achieve better goals

[slim garnet]

I mean generative tech is here now and is starting to be a part o human mankind. can't go backwards

[upper frigate]

It’s unfortunate the information is outdated

I think it would be really interesting for open world games with built in AI that makes that world unique to only them through their interactions

[nocturne ermine]

Rate limit China with Cloudflare! Hope you’re blocking RU.

[vivid atlas]

its all from america

[spice swift]

Oh yeah of course, it's will be part of our lives. Sometimes I just confused about people trying to get life harder than it should be. Like computer and medical is really good, why you trying to downplay the role of them with some old stuff that get kick out because of new inventions.

[upper frigate]

Rate limit their weather bawoons

[loud dove]

That’s so funny

[outer gulch]

using a new tech to help is fine, but coming to rely on it 100% is not good. Simple example, something noted by commentators and discussed before, is how the general population has lost the ability to do basic math in their head. You often see this when the check-out operator can't do a simple minus calculation. Tech is fine, but don't forget how to use the alternatives, for the times when that tech is not available.

[nocturne ermine]

Proxied, yea… China doesn’t even have access unless using proxy. Hard to isolate.

Oh, that’s what the balloons are for… Wi-Fi proxies!

[upper frigate]

They just wanted free wifi

[next shell]

Hello, does anyone know how GPT-4 performs in Software Testing? Both Code and No-Code testing?

[upper frigate]

Similar to gpt3

[next shell]

i don't quite know how gpt-3 performs

[upper frigate]

Use it for a couple days and you’ll know most of its limits

It’s basically only good for building a base for your code, and gives it structure and organized comments

[next shell]

answer 1-10, how likely is it for GPT models to replace software engineers?

[errant ether]

0

[upper frigate]

0 at the moment

[quartz grotto]

crud engineers? maybe, real software engineers? 0

[outer gulch]

dunno about anyone else, but I always wince a bit when I see the term "software engineer" lol

[crisp crane]

i have a small question . can gpt4 identfy parts of the brain from MR scan from image input ? just curios

[upper frigate]

No

[buoyant canopy]

Maybe that's because of that shiver on your back when you find a new topic to troll about.

[spice swift]

It's just not it, how about trying to more and more advancement to the point tech will 99.999% of the time available to us?

[vivid atlas]

im guessing

[upper frigate]

Although it would be interesting if gpt4 was able to compare scans and diagnose some patients

[slim garnet]

Will be on GPT5 codename "skynet"

[outer gulch]

well. it's a diversion from the usual, why GPT down, why only 25 msgs, why..." so cut me some slack please 🙂

[uncut jackal]

You're right that we shouldn't rely on tech 100%, but that doesn't mean we should dismiss its value altogether. It's all about balance, using AI and other tech to enhance our skills without becoming too dependent. People have different strengths and weaknesses – struggling with mental math doesn't mean they're bad at everything else. We should focus on using tech responsibly to complement our abilities, not to replace them entirely. Having the take of "New tech bad because people use it too much and so we shouldn't allow it" is just an incredibly poor take

[crisp crane]

when i look at the internet its mostly abnormal scan , so i cant find a healthy one to compare

[spice swift]

I think it could be, but not really good results on that.

[sinful arrow]

I mean in theory you could train a neural network on MRI images and give it the capability to recognize tumors and stuff right

[buoyant canopy]

It'll be great at psychiatry. Especially with hallucinations it is an expert.

[sinful arrow]

LOL

[slim garnet]

I spend 2 hours on this discord chat tonight. will do something else lol . have nice evening

[buoyant canopy]

Ok. Just came back to see what today's topics are. So no "when will it be available to everybody for free"?

[spice swift]

Yeah, you can try to write one and train the model to that points

[upper frigate]

Have a nice night

[outer gulch]

And I never took that view i.e. "New tech bad because people use it too much and so we shouldn't allow it". We just need to be aware of possible issues that could arise

[wind urchin]

I got that joke

[loud dove]

Really unlikely, there are copyright related issues in using gpt software for large scale coding environments

[buoyant canopy]

Damn. I'll try harder next time. 🙂 Usually I fail at humor or sarcasm here.

[sweet thistle]

It is unlikely that GPT models will completely replace software engineers, as software engineering involves many complex tasks that require a diverse set of skills and expertise. However, GPT models may be used to automate or assist in certain tasks. The future of GPT models in software engineering is uncertain, so it's challenging to predict their role in the industry moving forward.

It may be worth discussing this in #ai-discussions!

[buoyant canopy]

WDYM? Everybody and their grandma are using GitHub Copilot and the insert API is basically the same thing without the fancy IDE plug-in.

[tired river]

Im looking to buy a GPT-4 implant. Any sellers?

[wind urchin]

Well it won't completely replace them, but you bet your ahh that if senior engineers suddenly get 5x more productive then it means companies WILL reduce their workforce one way or another

[pallid aspen]

yeah, well thats gonna change

[loud dove]

As far as I’m aware most companies require software engineers to not use copilot

[eager viper]

How can I turn the bot angry pls answer

Pls

[quartz grotto]

It's a tool, just start using it to make your life easier as a software engineer. It's not going to replace you anytime soon.

I use a different AI tool now for my wifes business (photo culling, Aftershoot) and it saved us tons of tons of time. All it needed to do is take out blurry images, closed eyes images, duplicates, and leave out the rest.

Will it destroy photography profession? No. Will it change it? Certainly, and either you adapt, or you stay behind. We already are offering same-day-delivery, whereas others needs 2 weeks to just review the photos..

[buoyant canopy]

Ok. Maybe there's regional differences or "enterprise software" developers have to still do their CRUD by hand so they keep paying for their sins in life.

[wind urchin]

Didn't you make a Reddit post a while ago about precisely that?

[quartz grotto]

nope i dont think so

just checked to be sure haha, nope, but i like to give this as an example a lot, since its showing how AI can be used, and it doesn't need to replace you.

[wind urchin]

Weird, someone else had the nearly exact same scenario

[quartz grotto]

maybe because it is kind of gamechanging, and the software just had a major overhaul recently. For us, it saved us 7+ days, and its not man days, but full 7 x 24h (30k photos total that we didn't need to review)

and i just today used chatgpt to overcome some photoshoot decoration issues we had - just use it to make your life easier or when you're in doubt/crisis

[stable sky]

So hey guys, GPT-4 Is mostly for devs, rigth?
There will be no GPT-4 version online like 3.5?

[buoyant canopy]

There's also a point about that coding stuff that doesn't get enough attention: If you read this channel (or any other discussions about GPT), there are a ton of not-so-good-at-coding people that get a lot of empowerment from GPT. Alone the "this is my problem" -> generated code -> "it threw an error fix it" -> ... loop is very powerful.

[pale stratus]

But the gpt4 API already works? and public I have to get on the list if so how long to wait?

[uncut jackal]

GPT4 is available to everyone rn who's paying for chatgpt+, just not the image processing version, and it's limited to 25 messages per 3 hours (for now)

[stable sky]

Oh, so you have a choice to use it as API for devs or for personal uses?

[buoyant canopy]

For the API there's a waitlist, so chances are you get access but maybe it'll take some time.

[stable sky]

Big Thanks Guys <3

[candid wind]

Google just couldn't wait to write about this as soon as I open the news page it's all about chat GPT

[hallow peak]

25 message is too low

[odd basalt]

Hi , when can we start subscribing to GPT plus again it says " due to high demands we've temporarily paused upgrade "

[thin wolf]

Accept me 🙃

[glass marlin]

is the limit going to continue to shrink? Was just a 100 the last time i tried to use it. 25 places it in bing chat territory but that’s free.

[thin wolf]

ClosedAI

[sacred hearth]

the thing is that the chat history is realllly big, not like bing which u gotta start a new topic.

[candid wind]

Trust me You're probably only going to use chat GPT 4 about 15 times anyway before you realize you want to go back to 3.5

[buoyant canopy]

🔥

[thin wolf]

We up

[glass marlin]

i have no chat history currently. if it worked it would be big and if i could search chat history would be bigger but alas it doesn’t even work half the time.

[wind urchin]

You've reached the GPT-4 cap, which gives all ChatGPT Plus users a chance to try the model.

[jaunty parcel]

yeah chatGPT 4 is a not the saving grace we though,m 3.5 is faster and 4 is too limited now.

[sacred hearth]

Ah i have tons of chat history and if i ask it can gather info from the start of it

[vale plank]

anyone keep getting errors and it stopping mid sentence?

[candid wind]

Yea all they half to do is Make it faster its so slow i haven't really played with it that much

[sacred hearth]

But rn i do not have chat history as it aint available

[wet halo]

I think it was a marketing mistake to open up plus on 3.5 and give us all fluid conversation, only to launch 4 and dial it down to a trickle.

[jaunty parcel]

I cant believe that the model is stupily slow compared the old, this is backwards not forwards, also its not worth the money for the cap limits

[wind urchin]

I didn't even use GPT-4 today.. Yet it is saying:

you've reached the GPT-4 cap, which gives all CHATGPT Plus users a chance to try the model. Please checkback soon.

[long solar]

Why?

[lavish sphinx]

why does gpt4 cost so much

[jaunty parcel]

Im still waiting for the visual chatGPT, this is a game changer but they have to fine tune it or we will have anarchy

[wet halo]

I really think what we're witnessing is a bunch of AI researchers who don't know the first thing about large scale API architecture... I hate to say it, but it really looks that way at this point.

[glass marlin]

the limits on conversion length and time are aggravating. i would rather have 100 a day than 25 every three/4 hours. makes performing a project with it much more difficult. maybe that’s the point tho.

[buoyant canopy]

But answer speed is also a good proxy for the cost of running it. So maybe it's just too expensive for them to give everybody a lot of GPT-4 compute right now or mabe there just aren't enough V100 instances rn. 🤷

[jaunty parcel]

once chatGPT becomes multi modal, we are in a for shock.

[somber aspen]

May depend on how you use it, but for programming help I don't even want to bother with 3.5 anymore now that I know the level of quality I can get with 4.

[glass marlin]

multimodal has to be be way more expensive in computing cost. No surprise that’s not rolled out

[buoyant canopy]

💯

[jaunty parcel]

I relise the compute for this is vast but its the amount of users that is hampering this not necessarily its abiltiy to infer. LLama can run of a raspberry pi and this required only a small amount of training so its not compute thats needed its simply bandwidth

[wet halo]

Hey Hey Fringe - it seems to me that if MS is sucking 80% of the resource for aimless "bing" chat, they should just kill the free access here.

[buoyant canopy]

You have a good point there. And hi. 🙂

[candid wind]

@somber aspen yeah I heard GPT-4 was way better with programming so that's probably what I'm just going to use it for it to program. Everything else as of right now looks like 3.5 will be doing it

[wind urchin]

Why does message history keep disappearing?

[jaunty parcel]

I agree get rid of freebies and push a proper business model thats more stable. Free is not always the best.

[buoyant canopy]

But LLaMA/Alpaca are very limited in size. We don't even know how big a GPT-4 model is. So you're comparing apples with melons here.

[jaunty parcel]

Yeah but didnt they say llma was if not equal to GPT3 a 150b parameter model == 65B model for llma

[tacit ice]

I need help i paid for gpt plus and i still appear as a free suscriber

[buoyant canopy]

That might be. The "OG" GPT-3 model was said to be 175B parameters but I'd bet the GPT-3.5-turbo is much smaller, hence it's so fast.

[sweet thistle]

ChatGPT is currently experiencing some issues, this affects both plus and free users.
The service was down earlier, service is restored, but conversation history is still not available for many users. OpenAI had found the root cause for the loss of conversation history and are actively working to restore it now.
Keep an eye on https://status.openai.com/ for updates.

[buoyant canopy]

They have a lot of technical problems rn. Did you register just recently?

[wind urchin]

Thanks weeb bro

[storm yew]

so much server outage. makes me sad.

[glacial escarp]

same

[storm yew]

imagine having to go back to google. yuck.

[wind urchin]

GPT3 was severely undertrained wrt parameter count so I suspect it's probably not much bigger
Also #gpt-4-faq says it has a "similar architecture" FWIW

[tacit ice]

Yes, i paid today and 20 usd got discount from my credit card but the pay is not registered in my open ai billing history

[candid wind]

My chat is working now after I went through my history from a day ago and clicked new chat link and then after that link open I refresh the page again in the history even came back of all my chats

[jaunty parcel]

I cant remember exactly how many parameter it was but you might be right on that. Still its interesting that these model perform better on less parameters. I wonder if GPT4 is less but it has wolfram engine assistance etc

[loud dove]

Damn if only “openai” actually let people know how they operated

[buoyant canopy]

I can't recall having seen any technical details about 3.5-turbo and for GPT-4 they for sure didn't publish anything.

[candid wind]

I know GP t3 has like 175 billion parameters

[loud dove]

Yea neither models released parameter amounts, I was kind of annoyed spending like 30 mins trying to hunt the numbers down anywhere

[brisk heart]

OMG i can see other ppl's chat history

[candid wind]

@brisk heart until you screenshot it I won't believe it

[jaunty parcel]

Think they should rename the corp to UnopenAI lol. Although with the compute cost alone and the mass user base they are building I can understand the decisions they are making. Competitors would take over otherwise so making it freely accessible like opensource would steer the competition.

[wind urchin]

I pity the poor soul who will see mine lmao

[buoyant canopy]

🧨

[candid wind]

@wind urchin they're going to need to ask chat GPT for therapy after they see mines

Is everybody's chat starting to come back up mine is up now

[jaunty parcel]

WAAAt

[loud dove]

Openai has completely gone against its original goals of making sure that AI couldn’t be monopolized

[jaunty parcel]

thats weird

[candid wind]

It's AI for a reason and does whatever it wants

[brisk heart]

is it jsut me or with sineibe else asw ell

[buoyant canopy]

I also suspect the big players will clam up and not release their findings so openly in the future. Sad thing for progress in the field. 😦

[candid wind]

@brisk heart ?

[jaunty parcel]

I agree but im not sure what the best option is, if this was open source you can guarantee a load of people would make it dangerous and try to corner the market. In some ways its best being inaccessible to everyone so that some one cant take over the world lol.

[brisk heart]

OMG its all over the internet

[buoyant canopy]

You're the first one to report that as far as I'm aware of.

[brisk heart]

its all over the internet

[fluid gust]

I don't know if I'm late to the convo or something, but is it just me or is Chat GPT's history recovery down? If other people are getting this, any clue as to when they'll be back?

[buoyant canopy]

Maybe. Didn't read about it tho. 🤷

[candid wind]

@brisk heart whats sineibe?

Oh That's all over the internet that Chat GPT is back up is that what people are saying?

[buoyant canopy]

I get a message "We continue to roll out restoration of ChatGPT service. Users are not able to retrieve conversation history at this time. We are working to restore this feature."

[loud dove]

How can someone “corner the market” when everyone has it? Open source has proven to be great for innovation, just look at stable diffusion. You can run it on a gtx 1650 and get high quality art for free

[final rivet]

Anyone else getting a Bad gateway message? Cant load up on any browser.. Im getting error 502

[brisk heart]

someone*

[fluid gust]

Yeah same, its likely temporary similar to the GPT-2 to GPT-3 updates right?

[loud dove]

Openai needs some better back end devs lol

[candid wind]

@brisk heart oh lol i thought that was a name of a new chatbot

[buoyant canopy]

I don't have detail knowledge about what is broken but to me it's plausible all the current problems are related...

[jaunty parcel]

Yeah but its destroyed a lot of artist works as well, it does cause a lot of hard either directly or indirectly.

[cloud citrus]

Its not working still

[jaunty parcel]

harm

[cloud citrus]

I paid for premium

[brisk heart]

delete internet cached and internet data

[candid wind]

Bill Gates can buy all the land in the world but not all the internet space to hold this servo together

[cloud citrus]

Not getting it sorry

[loud dove]

So? No matter how it came out artists would be forced to adapt, now they can integrate it into their own works and train to create their own art even faster because they have free open access to stable diffusion

[cloud citrus]

Im in incognito

[buoyant canopy]

It's still amazing how people totally freak out about something they didn't even know about a week ago.

[loud dove]

Bill gates doesn’t own openai

[brisk heart]

still try deleting internet cache and data

[candid wind]

@loud dove isn't Microsoft involved in this? Or did he sit down for Microsoft?

[loud dove]

They have a partnership, but openai is still in charge of their own work and servers

[wind urchin]

do you have an estimation about when the visual input thing is gonna be out?

[brisk heart]

where are you located?

[brisk heart]

i don't know abt that sam altman is not sharing his discussions with msft

[jaunty parcel]

Yes but i believe in the motto, art should not be rushed. Art is a journey and not a rushed instant spatter. Too many people do not know how hard it is to produce good art and therefore believe its easy. The other side of the coin is SD is good at reducing pointless abtract artist who charge an arm and a leg for a dot on a canvas. This aint art to me.

[candid wind]

@loud dove oh okay I definitely can respect open AI A lot more for that

[brisk heart]

and they are having many discussions behind close doors... i beleive msft is controlling partner

[wind urchin]

because everyone has paid and are waiting for that specific feature

[cloud citrus]

Croatia

[brisk heart]

try back in one hour

[candid wind]

@brisk heart I feel that way too but hopefully openai draws the line

[brisk heart]

when did you first stareted experiencing this?

[spark dome]

Anyone got back their history yet?

[candid wind]

Yep

[jaunty parcel]

Still down for me

[remote badge]

Can we get like a counter that tells us how many messages we have used in the last 3 hours towards the limit?

[candid wind]

@jaunty parcel try going into your history at least about a day back until you see the link that has to say new chat click on it and then when it loads refresh the page

[remote badge]

And possibly a way to buy more messages for an additional fee?

[lapis jolt]

New account?

[wet halo]

My advice remains to take a few hours to relax and do something else, and let the system come back to some degree of stasis.

[reef fable]

Did i paid gpt 4 ?

[jaunty parcel]

Lesson learnt for me, i should copy it to one note so i dont loose everything lol.

[lapis jolt]

I'm working on my final right now and using it to help. Free break baby!

[candid wind]

@wet halo That's literally what I'm about to do I can't believe I got addicted to this so quickly when it's other things I could be doing around my house right now

[reef fable]

gpt 4 for free ?

[brisk heart]

i have a hunch microsfot may be on the whole outage...they have the most invested and wanted to know if people will be using bing...

[wet halo]

It's like some kind of new drug.

[brisk heart]

did anyone use bing?

[jaunty parcel]

argh whatever you do dont refresh your browser in chatGPT, you loose it forever

[wet halo]

I think they're sucking all the capacity for whichever people actually use Bing.

[reef fable]

someone has gpt 4 for free ?

[wet halo]

Has anyone here ever even used Bing?

[fallow glacier]

not me , i use duckgoduck

[jaunty parcel]

I would if this discord has a chatGPT bot and we are not aware. Its a bit creepy not knowing.

wonder

[wet halo]

it's almost funny that MS would invest $10b trying to save the sorry, embarrassing, joke, that has always been bing.

[sand hollow]

I hope that with ChatGPT 4 they leave behind all those limitations of morality and political correctness, as those things significantly clip the wings of AI

[broken glade]

I don't particularly like them but also don't even remotely ask it anything I know it'll give a biased response on

GPT-4 seems to have about the same morality that 3.5 does

[buoyant owl]

If you use their API to sell a service, your clients might use it for something and the "As an AI language model,..." response is not exactly professional. Happens even with quite innocent prompts.

[wind urchin]

Weird thing to do but i been putting the whol gensis bibble into gtp to turn it into a story lol. Not my thing but it works and is interesting

[broken glade]

yup

[wind urchin]

Can you add a way to see how many questions you have left?

I have found that a lot of questions get used up by repeating response, they should fix that to not count as a question lol

also can someone change my name to something random

@wind urchin I agree

[sand hollow]

Sure, and then he hits you with the "as a language model, I can't do that" line. Dude, you sound like an npc lol

[drifting tundra]

is the chat . openai . com down?

Or am i going to the wrong link?

Im getting Bad gateway

[silent frigate]

Yeah it seems down still

[strong timber]

Can someone help? GPT-4 cuts itself off with html code and, despite what I put, always starts over and then, naturally, can't complete because it's too many characters -- how do I get it to just break up the html code? It's not even that long and I have asked in every way. I apologize for the confusion. I'll separate the complete code into two messages. Here's the first part of the integrated code:

[fallow glacier]

i use TruthGPT

[sand hollow]

That's just it, Even though it may not seem like it, a lot of interesting things are lost just by imposing that absurd limitation. I'm not asking for a step-by-step guide on how to build a nuclear bomb, but even simple things are censored

[tropic plaza]

Try saying “please continue”. That has worked for me

[wind urchin]

Rephrase the way you ask it to start - i often in gtp4 - Start at "insert last complete line of code" Print rest of code and it works

[slow beacon]

i just type continue

[dense nexus]

Its insane. I can understand some is needed but It feels like using a parental control version age 10 and under lately. i hope google will make use of this opportunity and atleast make Bard act like its speaking to an adult.

[wet halo]

I just want to pay for access to the full meal deal and not be subsidizing a bunch of kids doing their homework. I'm sorry to be harsh about it, but if we have to make tradeoffs, I rather kids actually study.

[muted shard]

Hey, uh I'm having an issue with my OpenAI account, was wondering if anyone knows what's going on

I had my laptop on sleep, came to use ChatGPT in the morning, had the window already open but it said 'your session has expired, login again', I tried logging in but it said my password was incorrect, I tried resetting the password by email, but I just don't get any email whatsoever

I even copied my email from a receipt email OpenAI had sent me earlier but I still don't get any email, checked spam section, nothing was there

does anyone maybe know what's going on?

[strong timber]

Yeah I have said "Please continue" "You finished?" Why did you stop?" "Do not start over whatsoever" "PIckup from this point" -- it never listens! And I have submitted the feedback and the alternative answers obviously don't work either.

[fresh mantle]

can ChatGPT Plus read an instagram page and describe the vibe

[dapper lintel]

Pay for the monthly

[strong timber]

Yeah, have tried that and everything I an think of...Just keeps repeating itself and starting over. Even tried same code in new chat, Pretty bizarre tbh!

[wet halo]

That's my point - I do. And I'd pay three times as much, at least, to have access to GPT4 at a usable pace.

[wind urchin]

It has fits liek that, and then it loves to do it in a text response outside of code box and ruins all indentations haha

[chilly notch]

make 3 accounts

[rocky summit]

I am still getting the Bad Gateway. Can't log in. Is that still happening or should that be fixed?

Checked status page and it says it was fixed.

[long finch]

Can GPT-4 make presentations with images?

[wind urchin]

Just remember gtp4 was released 14 mar, things will get better and all feedback will be filtered by gtp4 and most likelt not be implemented

[wet halo]

Yeah.. but not a great solution. It's the speed of response more than the number of prompts per hour. I don't need however many hundreds of prompts people want because I talk to it like a Ph.D. intern. I'm not looking for a conversationalist, I'm looking for a brain augmentation.

[cinder briar]

it does

[wind urchin]

I am real, aussie. But then again i cant prove that as who knows haha

[mellow gust]

still widespread issue. it's not even being acknowledged in the Status page.

[chilly notch]

yeah it typing back is so slow, is there a tech breakdown for how it replies? I feel like it should have the info

[rocky summit]

Ok, thank you. I thought so, was just doing a sanity check here.

[wind urchin]

A good way to do stuff is make most in gtp 3.5 and then ask the complicated version and refinement in gtp4

[wet halo]

They're just choking it - if you look at the demo, he didn't have to sit around playing solitaire while the thing responds to him.

[dapper lintel]

You complaining about the wrong things but wtv floats your boat 🤷🏽‍♂️

[wet halo]

Obviously you're entitled to your opinion - but you could try actually debating the point rather than just spewing invective. But, whatever floats your boat.

[harsh wagon]

I can’t believe wtv has became slang now 😭

This generation comes up with slang for everything it’s hilarious

[inland frigate]

I paid for it 5 days ago and its been nothing but a joke and a waste.

[dapper lintel]

Arguing about this? I’d js troll for fun

[wet halo]

Like I said.

[dapper lintel]

I’m not about to argue about “cHaT gPt” 🤦🏽

[harsh wagon]

Chat gpt 4 is really good what do u mean

[wet halo]

🤣 Fair enough.

[dim forge]

Just trying if still blocke ^^

[near quest]

anyone problems to get in with premium account google auth?

[inland frigate]

First it was supposed to be multi-modal and accept image input.

They took that away.

Lowered message cap from 100 to 25.

Have been down 2-3 times in 5 days.

Lost all my chat history.

Best paid service ever. 👏

Oh, and OpenAI the team just hides behind anonymity and has no outlet or channel to reach them aside from hunting them down on LinkedIn.

Always great when people take money and have no channel for input.

Its not like they just got $10B or anything.

[harsh wagon]

I think it’s just u

[wind urchin]

what is wtv

wait ill ask chatgtp

[near quest]

joke?

[vestal narwhal]

would it be possible to receive an invoice for our subscription to chat gpt?

[mild pendant]

Go refund 👏🏻 👋🏻

[harsh wagon]

They said there was issues what do u expect

[harsh wagon]

Nope

[near quest]

uhmm i can't get in... at allll....

[inland frigate]

I expect a modicum of decency, not this hot garbage where neurotic children defend bad business.

[mild pendant]

People just like to assume money prevents problems or issues. Also some people are really obsessed with paying for things they don't fully understand 🤷🏻‍♂️

[wind urchin]

i can't see chat history, but I can use gpt 4. i just hope we eventually get a higher cap once it settles tf down

[inland frigate]

Nope, not the point. The point is their glaringly obvious lack of any business plan whatsoever.

[harsh wagon]

U don’t know how to run the bot neither do I so why do we get to say that stuff if there was a honest issue that happened with the bot

[wind urchin]

lmao

[wet halo]

This is an AI firm that doesn't know how to scale a public facing service. Give them time - they'll fix it.

[wind urchin]

no baguette allowed

[inland frigate]

If you have a plan, you don't have a mistake every other day.

They don't even have a solid idea of what their service is, they are just taking money and cannot provide even the basic features they are supposed to.

[near quest]

i still can't get in, is there an issue reported regarding auth with Google can't get in my account

[harsh wagon]

Exactly what I was saying it’s tech there is no 100 percent guarantee that something won’t go wrong with it

[inland frigate]

Has everything to do with really simple to comprehend business logic and nothing to do with the bot. Why is this so hard for you to understand?

[wet halo]

I'll say that if they had partnered with AWS instead of MS... they wouldn't be having these particular issues, lol.

[icy obsidian]

Don't pay if you don't like it then

[limber garnet]

OpenAI needs to meet its financial obligations. This means:

• Providing promised services
• Making up for downtime
• Realistic advertising

[wet halo]

The terms are quite clear - they state that capacity will vary by load.

[harsh wagon]

Well it’s common sense what type of business they are running a tech business tech business at such a large scale that they have never done before obvious that something like this could happen

[wind urchin]

Money is a tool, money is nothing more than that. Its a tool you should use to leverage yourself and position yourself to create better successes for yourself in life ( not directed at you )

[inland frigate]

Hush, that doesn't even make sense.

[icy obsidian]

Leave the server if you don't like it then

[wind urchin]

also , these " additional features" aren't very addiotional as we are getting quite limited by what we can do with those

i still love gpt tho

[mild pendant]

Not with impatient people... They just want to complain/refund 🤣 then wonder why servers are trash

[white pawn]

Think it has to do with over capacity, frankly and it is still very young in general do I would consider chatgpt in any state in beta mode

[harsh wagon]

Not my fault your on chat gpt comprehension level

[limber garnet]

That isn't an excuse for failing to meet the obligations I mentioned

[long finch]

Can GPT-4 make a presentation with images?

[mild pendant]

If anyone tries to imagine the amount of traffic their team has to handle per second, they'd deffo pee in their pants

[harsh wagon]

Gpt 4 is dope

[white pawn]

And beta mode has errors flaws trials crashes and people want perfection right out the gate well there expectations unrealistic

[limber garnet]

The complaints are a tad excessive but they're completely valid /shrug

[harsh wagon]

I probably would too

[white pawn]

It was down for a few hours and you would have thought everyone died from stage 6 cancer

[harsh wagon]

That dude is a male Karen slyxus

[wet halo]

Exactly - this being said, the MidJourney peeps are doing a better job at managing capacity.

[wind urchin]

In 24 hours they had 1 million subs, 1 million x 100 responses every 4 hours / seconds in 4 hours will be your answer, being at probably 10 mil now lol its insane

[long finch]

I’m still waiting for a response.

[wind urchin]

You can use gtp4 to make a program to make a presentation with images

[limber garnet]

Have it write the text and use Msoft's designer in Powerpoint Online for the images

[long finch]

Alright. Thanks.

[mild pendant]

The complaints are valid in the perspective of downtime being a bit annoying, but there are other ways or tools to get the job done. Plus, the demand on ChatGPT is IMMENSLY huge, do you even have the slightest idea how many people mass create accounts just to sell them and share them? OpenAI could do a d move and restrict all these fools but they don't. They're being lenient, and give everybody a chance to try the tool out. Yes, the cost is increasingly huge traffic and load on their services, but cmon why do we have to be so ungrateful and pieces of

[long finch]

I appreciate it. Thanks.

[harsh wagon]

Do you think we will ever be able to download YouTube videos on it so we can have it be used as a reference to what ever topic?

[harsh wagon]

That would make it really good

[wind urchin]

based off this, there was in the first 24 hours, 6944.44 questions/answers per second

[white pawn]

I am surprised it doesn’t crash more with that many questions per second

[wind urchin]

Ask the right questions apply time and efort you can make a tkinter selenium driver program in python thatr can

[limber garnet]

It isn't ungrateful to expect to get what you paid for. You need to advocate for yourself or you'll get scammed left and right. OpenAI is a for-profit providing a paid service. They aren't a charity and they aren't out to do good for the people

[wary kraken]

GPT-4 currently has a cap of 25 messages every 3 hours. Expect significantly lower caps, as we adjust for demand. by lower they mean like 10 messages per 5 hours? lol

[harsh wagon]

Yeah it can basically already do that but gathering the information can be a little tedious xd

[wind urchin]

My head hurts from asking a thousand different ways to append my programs and ideas haha

[gentle shard]

They should cache similar prompts and return "pre-recorded" responses to save on their server resources. Would probably end up saving them millions over time.

[wind urchin]

I feel it should be an exe install and use the users end to help with load create a share the load app

they do it with online games

[gentle shard]

People hate downloading things.

[limber garnet]

They don't want power users stealing the weights and removing OpenAI from the equation

[wind urchin]

But people can never be pleased lol, its impossible to make everyone happy

[mild pendant]

There are far more spectrums to expressing expectations than what was exhibited. Paying people should for the most part expect that they're also funding a relatively new research and company. And it's really as easy as if you are not satisfied, just go the legal course. No need to keep the yapping on whenever downtime happens.

[gentle shard]

You also introduce cracking when it's a downloaded app.

[wind urchin]

You can use gtp 4 to make a training model similar to gtp4 and a skeleton similar to the model if asked right questions,

[wind urchin]

True but you can design it to keep important parts on their end and a load sharing feature there end

[hallow peak]

Ahmm anyone knows where the history gone?

[mild pendant]

"We continue to roll out restoration of ChatGPT service. Users are not able to retrieve conversation history at this time. We are working to restore this feature."

[wet halo]

I gather their approach is to limit the public facing chat tool because of the sheer volume of garbage in. I assume once we get access to the API for 4, we can all just escape this hell, lol.

[limber garnet]

You changed your argument to be about complaint delivery instead of justifying OpenAI's shady practices. As for funding, again, OpenAI is a for-profit and the expectation is that you get what is advertised. This isn't Kickstarter or a research organization; it's a corporation providing a service

[graceful jacinth]

Well said

[wind urchin]

I can only imagine there is thousands of kids and people on here like hahah ask this question haha this question ( rude ones ) and not actually realising the true potential of progressing them self and knowledge haha

[wind urchin]

using the general public on mass to rapidly improve features and make people feel more comfortable talking to a computer smarter than them 😛

Without being conspiracy like or to the likes of that. I genuinely feel this is a pivotal point in society where those who choose to apply knowledge with this feature to progress will further increase the rich middle class lower class gap

[wet halo]

This is the problem - blindly allowing free usage is a complete waste of the resource. I'm all for giving it for free to schools, to motivated students, and more or less to anyone trying to actually do something useful. But I bet that at least 70% of use is absolute garbage currently.

[broken glade]

that's most things tbh, how much of amount of fuel burned is not for some kind of productive purpose?

[wind urchin]

agree

[manic peak]

why's the site not loading past conversations?

[broken glade]

there was a total outage earlier, they brought service back but history isn't back yet

[tacit ice]

anyone else with this trouble??? i've been paid gpt plus but my suscription is free now

[broken glade]

no, mine is working just fine, all the features are there still, did your card get declined or something on renewal?

[wind urchin]

Give it time it will fix itself

[tacit ice]

nope, even the pay was discount off my card, but it does not appear on my bill history on open AI

i bought it today

[cunning tulip]

you will get the login link after submitting your email address when prompting

[true sluice]

How to use open AI

[tacit ice]

during the crash

[tacit ice]

how can i get it?

[cunning tulip]

you can't login from the homepage

[tacit ice]

i can

[wind urchin]

based off 1 million users in first 24 hours, that was 6944.44 questions/answers per second, now we are exponentially past that they are most likely improving systems to handle load

[tacit ice]

but it appears as free

[cunning tulip]

looks like login page ok now

it was not available earlier

[reef skiff]

Back without history

[hushed steeple]

it is pretty weird how people can't even get an email back from them

[deep hornet]

I wonder has Open AI considered doing some of the processing on the clients cpu? to lighten the workload?

[hushed steeple]

surely whoever handles emails at OpenAI can do it 100x faster now with GPT-4 lol

[analog kindle]

that guy is a whiner. the guy you replied to, i mean

[cunning tulip]

please bear with the new technology. bugs/crash are always expected😅

[gentle shard]

IMO they are handling growth overall exceptionally well. They are only a team of 375. And many of those guys aren't developers.

[rotund wing]

Hey guys, I found soemone that managed to jailbreak GPT4 - to get it to access a website, how do you reckon they did it?

Bing GPT4 provides stupid answers and I yelled at it and it disconnected me

[cunning tulip]

probably that's how it get crashed lol

[gentle shard]

You can just include a link, its a feature.

[loud dove]

Either it was fake or they used the api plus google search with embedding

[near viper]

Not all links work, but alot do.

[hushed steeple]

Yeah, it's only been a week since GPT-4 released. Once they adjust for traffic I think we'll be good.

[loud dove]

Bing can search websites but I don’t recall gpt4 being able to?

[rotund wing]

tried that, it told me it could access it - maybe i need to provide a more specfic link not just a homepage to a website

[rotund wing]

interesting, where is the docs on this?

[cunning tulip]

bing and chatgpt give your different result.

[analog kindle]

afaik you can only use links that chatgpt knows about from its training period.

[cunning tulip]

in some case

[rotund wing]

results were current imo

[near viper]

Nah, Ive had it read a texas house bill that was presented in 2022, the training cutoff data is from 2021

[rotund wing]

go on 😛

[broken glade]

no don't because politics 🙅‍♂️

[tepid stratus]

so does anyone know why chat histories go down from server overload? just curious

[crystal field]

If you tell GPT4 to never change a function does it work? I'm getting sick of it changing them just slightly. at least # same as before is obvious but it makes merging them a pain

[sharp zenith]

The cruel truth is that openai became a mercenary company. Has lost its meaning since the launch of ChatGPT plus

[analog kindle]

do you have a link?

[cedar raven]

Are you serious?25 message per 3hours???

[crystal field]

do you get in trouble if you interact with the webchat via python

[rotund wing]

scrape?

[analog kindle]

only for gpt-4. yeah it sucks but it's cause of server load. can't blame them much IMO

[gentle shard]

Has anyone hit this limit?

[formal junco]

Hey is ChatGPT. Down?

[burnt musk]

any idea when history will return? Also, feature suggestion: Add the ability to save history on your local computer.

I'm using GPT to help me write a story, and completely losing access to the conversations mean I'm completely paralyzed in the process, I can't even read over what was previously discussed to work on things by myself

[quick wave]

I am curious, when GPT-4 really goes intermodal, do we know if it will be able to read a docx file with tables?

[formal junco]

I keep getting Bad gateway
The web server reported a bad gateway error.

Anyone else?

[crystal field]

yeah

[graceful kettle]

It was 100 per 4 hours for a little while. That obviously didn’t work out, lol.

[delicate flare]

Check https://status.openai.com/#

[crystal field]

i've heard if you get banned you're banned for life

[gentle shard]

Probably so, in the demo they showed it could create a website from a sketch on a napkin so I imagine pdf/docx would be simple.

[rotund wing]

did someone say new email?

[crystal field]

they have your phone number.

[delicate flare]

Yep. Even you'er a Plus user.

[crystal field]

i guess you could use burners and whatnot but if you paid they have your name too, unless you used someone elsese.

[warm elk]

i cant buy plus anymore? omg

[analog kindle]

oh yeah, even if it's just an image of the docx that would work i guess

[quick wave]

Okay, that is impressive and means I will be able to 'teach' it stuff in a more consistent manner before preparing a prompt.

[hushed steeple]

that's a pretty scary thought if you need ChatGPT Plus. A lot of people wouldn't be able to do their jobs

[warm elk]

jesus for what iam paying then xD to give you more money for servers?

[rotund wing]

yikes you are paying 20 USD a month lmao - what the hell do you expect

[warm elk]

let me pay !!!

[honest condor]

How to use gpt4 in discord

[rotund wing]

I got literally 10k minimum value from the free version

[gentle shard]

Create a bot

[rotund wing]

take my money

[hushed lark]

I don't think that they should be making it as expensive as it is with $20 a month for how it is right now, especially with consistent downtime and a limited messaging rate for ChatGPT4.

Though, it was pretty clear from the get go that their aim was mostly to open it up to the public so that it could really wow investors and it did its job at that with Microsoft. Anything past that with the subscription & more was just a bonus.

[quick wave]

Do we also have an idea when it might be available on the playground (and not just via APIs) ?

[analog kindle]

lol at least it's not more I guess

could be like 100

[warm elk]

jesus take my money !
i cant buy !

[cunning tulip]

waiting list

[warm elk]

there is no waiting list there is just a big fat NO xD

[near viper]

DM me, @rotund wing also if you want more info I can send it to you as well

[thin wolf]

It's been a year daddy.

[hushed lark]

That all said, I suck so I paid for it and it's actually really solid. Even ChatGPT 0 can't get a good amount of it's responses as AI generated.

[gentle shard]

Yea $20 is a steal it pays for itself. If you can't make it do or help you do something in a month that can make you $20, something is wrong.

[cunning tulip]

😅

[rotund wing]

I am sorry its 20 USD a month. lets assume you use it at 25 response * 3 hrs for 3 periods a day = 75 responses per day * 5 days a week = 1500 responses per month. = 1.3c per question I am sorry if you are not getting your value for money but I am for sure

[cunning tulip]

I got mine a week ago

[honest condor]

Create a server to join the robot@gentle shard

[thin wolf]

Lucky

[gentle shard]

?

[crystal field]

If you pay 20$/mo do you have playground/API access?

[warm elk]

you dont need version 4 for every single question

use it smart not like a idiot 😄

[crystal field]

I never stopped having access at any point today and i kept batch sending stuff to davinci-003

[cunning tulip]

i don't think so.

[broken glade]

you pay for unlimited 3.5, a sampling platter of 4 happens to be available

[sweet thistle]

ChatGPT Plus is separate from the API/Playground

[rotund wing]

my point gets even more clearer.......

[broken glade]

are you paying?

I pay and I'm super stoked with the service so far

[crystal field]

i see

[rotund wing]

I am bullish on it, like people here are complaining about 20 usd a month. my point was just to use the exageration

[gentle shard]

What are you guys working on with GPT4 API?

[warm elk]

if you ever complain about 20$ a month you need to get a job lol

[rotund wing]

for reference I was referring to this guy

[warm elk]

20$ is nothing

[crystal field]

do people here have api access?

[long pond]

Everything.

[glossy rain]

Okay who broke the bot? My prompts are all screwy now. I've had to rewrite half of them. It's damnably frustrating. Even on 3.5 they are all screwy.

[grim veldt]

me

[pine plaza]

yeah it's not working quite the same as it was before the downtime

[glossy rain]

I spent a month getting the right prompts and primes. And now it's all screwy. How long is the bot going to be "worked on" this time?

[burnt musk]

Since ChatGPT and GPT4 will not create violent or sexual scenes, which the story i'm writing calls for, is there another similar service anyone could recommend?

Generally i'm very pleased with what GPT4 has come up with, from inventing fashion trends, fictional sports, religious rituals, and so on, it's like i'm developing an encyclopedia for a world that doesn't exist. However, not being able to describe violent or sexual scenes and practices is a serious limitation in the hedonistic and violent culture I'm creating.

[analog kindle]

lot of people bought plus expecting being able to send it images

[pine plaza]

I've been able to link it imgur links and it seems to understand them

[storm jacinth]

GPT is down right now yeah? Any updates as to what is going on? I can't seem to find any lol

[crystal field]

like it understands what the imgur link was to?

[pine plaza]

either that or it got very lucky multiple times

[crystal field]

it should have the context around the url since it was crawled with it

[heavy cedar]

Miiite I wouldn't be surprised if final version will be $200 per month

[pine plaza]

well it was an imgur I made myself

[heavy cedar]

It saves so much time

[pine plaza]

I used it to get help with a GUI in my minecraft mod

[crystal field]

common crawl contained geocities right?

i need to look into some completions with the old models while they're around..

have to go back 15 years on wiki to find the url format..

[analog kindle]

recently uploaded image i take it?

[pine plaza]

well this was like yesterday or the day before, but yes

at the time it was brand new

[glossy rain]

I'm about to discontinue my plus membership. It's like every other day is some new kind of screwy thing.

[analog kindle]

and it wasn't possible that it used the title or anything like that? if not that's cool, im glad you mentioned that

[tacit ice]

this takes a lot of credibility

i paid 20 bucks and they dont even recognize my pay on the pay history of open ai

it´s like i´ve never paid

[floral kindle]

does bing have image yet?

[pine plaza]

it's worth testing, could have just been lucky

[crystal field]

you can get people';s blogs back if you put in part of their url on geocities

although it sounds definitely hallucianted

[analog kindle]

yeah, sadly random image links don't seem to work

[burnt musk]

at that price point, competing models like Alpaca would rapidly dominate the market, heck, alpaca already apparently lets you have your own chatbot on your own local computer, almost as good as GPT for like $600. The field is developing rapidly, and if OpenAI gets greedy, others will arise to come and eat it's lunch

[crystal field]

have you guys checked if gpt has doxx on you

[analog kindle]

maybe people have tried sending it the code for images?

[next tide]

Anyone got a discord bot w GPT-4 running?

[rotund wing]

@burnt holly dude that was a 10/10 call thank you

[livid steppe]

The $600 is if you want to train your own language model
Using the $600-model that's already been made would just be a matter of installing it

[brisk sail]

any idea when gpt4 will roll out to public chatgpt?

[pine plaza]

Me: could you describe this image in this imgur link: imgur.com/jeoZmbI

ChatGPT: The image in the provided link displays a Minecraft game screenshot. In the screenshot, the player is standing in a grassy area with a large, multi-level wooden structure in the background. The wooden structure features fences, staircases, and glass windows, resembling a house or a building with multiple rooms and floors. Several trees surround the structure, and the sky is blue with a few clouds, suggesting a clear day in the game.

it seems it can identify some things somewhat

and this was just a no-context screencap

[crystal field]

it was that in the past? or did it hallucinate it

[next tide]

Is that through playground?

[pine plaza]

I did that just now with GPT-4

[crystal field]

the link is dead

[pine plaza]

imgur.com/a/jeoZmbI

try that

[glossy rain]

the image and video bit are pointless. it will be destroyed in court by modern media. text is one thing, but other multimedia will be severely limited or hindered.

[crystal field]

oops im dumb, never use imgur

so interesting, it will officially contract the web

[brisk sail]

works

[crystal field]

side-channel prompt injection ahoy

[serene cipher]

it cannot

[blissful bluff]

is it still off? I have slept for 8 hours after it was down

[pine plaza]

so where is it getting context from?

[crystal field]

its back for a while now

sort of like BLIP/CLIP maybe?

[blissful bluff]

We continue to roll out restoration of ChatGPT service. Users are not able to retrieve conversation history at this time. We are working to restore this feature

[pine plaza]

unless Lightshot is generating its own context in some way?

[hybrid vortex]

still no history but the chat prompts are still working

[serene cipher]

when it manages to get something right about a url it is because of the context that surounded that string of characters on the training data, ChatGPT cannot explore the contents of links

[crystal field]

is there any official way to backup chat

[mint pumice]

can anyone use chat gpt now? Because I m getting bad gateway error

[hybrid vortex]

thanks for this!

[mossy flame]

history has been restored

[simple urchin]

Is GPT back up? I'm still at work so I can't check, but I'll be getting off soon

[wind urchin]

Someone go back in time tell the Terminator that Skynet ran out of ram

[pine plaza]

and how would it get that from a randomly generated imgur link?

[crystal field]

it probably crawled it recently

how old is the link? i would upload a new one and try

[burnt musk]

Not for me it hasn't

[pine plaza]

that was a new one

completely fresh, uploaded seconds before I sent it here

[simple urchin]

It can't visit the sites themselves but it can analyze the link, and the fields contained

[serene cipher]

did the AI got it right?

[crystal field]

how annoying, gpt used to tell you it thought it was 2021 so y ou could tell when it was trained, now it just says it can't say

[pine plaza]

wasn't very far off at all

[abstract obsidian]

Is it still down ?

[burnt musk]

no, but history is

[past token]

/chat

[serene cipher]

the AI is simply incapable of oppening links, all it sees is the scring of characters you provided

in the best case scenario, the URL had contextual data about it on the dataset, otherwise, it was a hallucination

[drifting kettle]

I can't log in

[glossy rain]

Jesus this thing is borked. like badly borked.

[abstract obsidian]

Unable to Login. Nothing happens post clicking on login !

[glossy rain]

its practically unusable without repeating the prompt 9 or 10 times.

[sweet temple]

I dont think so, some cases it might do that but today I gave gpt a math equation in the for of a image link, and used it to derive another equation this was the link

https://cdn.discordapp.com/attachments/775219532813828116/1087257840407814165/image.png

theres no context about it as I uploaded it to discord just a few minutes prior

[serene cipher]

https://status.openai.com/

it is having issues rn

[glossy rain]

I've had to redo 5 different conversations to try and do what I was doing yesterday effortlessly.

[analog kindle]

sad if true

[wind urchin]

Who I text my banking information to so I can receive my refund.

[glossy rain]

the whole LLM is out of whack.

this feels like two years ago to be honest.

its not remembering a thing, or barely.

[pine plaza]

yeah it's definitely taken a hit

[wind urchin]

This feels like 10 years ago when I paid money and barely got anything from a woman.

[crystal field]

does feel like he's a bit alzheimey right now.

[drifting kettle]

lollllllllllllllllll

[crystal field]

i'm going to semi-automate this workflow with some diff tools and maybe codex since it keeps getting things wrong

need something like a pull request from gpt "no, stop that, dont' mess up that function"

[hybrid vortex]

is this gpt4 or 3.5?

[crystal field]

i am using a mix of 4 and 3.5

[hybrid vortex]

ahh nice

[crystal field]

4 when the context is long enough

i dont see that much advantage otherwise and even with it now it's pretty bad

it's still really nice but it requires an entirely different workflow.

[hybrid vortex]

i mean you can feed it bit by bit with intricacies of instructions and it'll be fine tho, wouldn't it?

[serene cipher]

I understand what you say... but the AI simply cannot do that...
it makes a text based on the entirety of the context that is has based on the dataset and your chat messages... the result you get can be based on a lot of things.. contents of the link you provided is not one of those things

[crystal field]

when i'm working on a script i make a new subdir in /iter/ for every gen from gpt and then merge it with beyondcompare and such

i'll probably merge it into git

but there's a lot of things that could be caught automatically e.g. variables not defined that you end up finding later when it's almost done..

[drifting kettle]

I can't log in, is there anyone the same?

[hybrid vortex]

would you be able to say that it does save you time and major effort instead of doing it manually?

[crystal field]

i'm an EE so very very much so

[serene cipher]

yes, there is a system outage going on: https://status.openai.com/

[crystal field]

it threw together a script to obfuscate text or a directory (sswitching nouns/etc around), and i'm this close to being able to also use the harvested text from neuromancer.txt on snowcrash.txt

[drifting kettle]

Good.

[hybrid vortex]

thanks for the confirmation! glad to know that

[crystal field]

i'm so happy i dont have to go to google anymore

i've despised them for years as they got worse and worse since page and brin left

i can type something quick and full of typos at 130wpm on my second monitor and it hasn't gotten it wrong yet

[hybrid vortex]

go to google and look up forums/reddit posts and sifting through chat logs kek

[crystal field]

even "how forward x11 lan"

[warm elk]

take my 20$ XD

[crystal field]

there's one good reddit and i despise the rest of it

i'd rather scrape hackernews i guess

[hybrid vortex]

20$ is much well spent then 👌

[glossy rain]

atm it isnt. not for this borked api lol

[hybrid vortex]

lulw

[hybrid vortex]

am actually afraid that once they handle things better, they might up the price

or set a higher tier payment

[crystal field]

20$ is an insanely good deal

[shell marsh]

Not having access to my conversations is definitely gonna keep me from renewing.

[hybrid vortex]

agreed

[glacial escarp]

idk, they're definitely going to be more competition going forward

[serene cipher]

I don't think they will increase the price any time soon, just a guess tho

[crystal field]

people are pretty insane to think this level of tech with those insane GPU compute costs should be less than 100$/mo minimum

[glacial escarp]

*there's

[shell marsh]

And the limit dropping every time i log in.

[crystal field]

plus it's azure so their electricity is probably not cheap, not like crypto when people mined in iceland and stuff

[serene cipher]

that might be an incentive to lower the prices

[hybrid vortex]

that's why im frantically squeezing every bit of content that I can feed it and give me before anything else happens

[serene cipher]

at the current pricing, users can reach 100$ usage a day easily xD

[pure remnant]

using GPT-4 now, met this error a few times in the past a few minutes. "openai.error.RateLimitError: The server had an error while processing your request. Sorry about that!". What caused this error?

[hybrid vortex]

makes me sweat

[glossy rain]

stuff like this has been around going on 4 or 5 years now. but the price model is so outrageous it isn't viable for everyday consumers outside of professionals who want to write "lazily". GPT 3.0 and 3.5 made it remember and made it available to the public. so the everyday consumer is breaking it. I look at GPT the same way and effect that Napster and then Limewire did back in the late 90's early 2000's. Even with that we now have "streaming" services. When you give the everyday consumer easy and fast access, that is the TOP priority they want.

@wind urchin someone broke the bot.

[steel tide]

How?

[mint wharf]

Is there any eta?

[crystal field]

its quite possible they tuned down things due to the issues, QoS and such

[glossy rain]

if they tuned it down, im out of here lol.

[crystal field]

🙄

how are people taking this for granted so much

[serene cipher]

just do a lot of requests =P
I spent almost 20 bucks in a few hours while I was making chatbots a few days ago

GPT-4 tho

[livid steppe]

what I've been shocked by is how quickly we've gotten used to gpt-4 being "okay" and gpt-3.5 being "crap"
the rate at which humans are normalising this is almost as surprising as the rate at which the tech is developing

[serene cipher]

3.5 turbo models are very cheap compared to the other ones, tottaly worth to use them in most of the times, making a full chat with GPT-4 can be very expensive

[wind urchin]

use cases of GPT 4 list down or you will go to Hell

[honest geyser]

because they are insanely different in terms of intelligence

[livid steppe]

oh I know, I much prefer chatting to GPT-4

[hybrid vortex]

I treat 3.5 as the grunt and 4 as the QA

both are still needed for me

[crystal field]

do people actually chat with chatgpt like e.g. ELIZA

[wind urchin]

no

i bet there will be another pricing tier, cant expect them to churn out videos and audios when chatgpt4 fully goes live

everyone uses it for their good

it might be feeling lonely cuz no one talks casually with it

@crystal field

[dry citrus]

Is anyone able to put a math prompt into gpt4 for me? gpt3 typically gets my math questions wrong

[honest geyser]

it's not a chatbot

[crystal field]

just curious, i use it for code, summarizing thjings and replacing google

[honest geyser]

try looking at #1074529070416666685

[crystal field]

there's of course that other group of people who want to chat with computers out there right now

[serene cipher]

i think the image thing will be part of the gpt-4 model on the API, where you set the image file as part of the message data

[maiden dune]

what happen

[dry citrus]

It's advanced statistics

[wind urchin]

use CodeX MODELs such as Curie , Ada , Babbage etc for coding....in OpenAI API Playground it would give you better results:)

[hybrid vortex]

same, i also use it for writing

[honest geyser]

still

[dry citrus]

gpt4 got all of them right yesterday

[maiden dune]

why dont do it

[silk minnow]

can gpt4 process audio

[livid steppe]

all my conversations with gpt-4 are casual and "chatty", even when I'm asking it about how LLMs work, or trying to test it on awarness/etc
using a conversational style seems to work well for it

[maiden dune]

i cant gpt4 ...;;

[honest geyser]

No, not yet I think

[proud sundial]

nope

[serene cipher]

this is jsut a guess due to the fact that they already gave the prices for the 8K and 32K models, but didn't gave one for the image feature... so.. maybe it is already part of the chat completion gpt4 models

[dry citrus]

still what?

[honest geyser]

I would still look there, you might find something that can help you

[dry citrus]

It appears it is only for the basics of math

[serene cipher]

no, this is a feature of the Whisper API, which OpenAI also provides

[honest geyser]

yeah it's not really good at maths

[glad lichen]

Is image functionality on api yet

[honest geyser]

you can always just do it yourself you know

[honest geyser]

I don't think so

[hybrid vortex]

so i have to make my own APP first to integrate API into it?

[dry citrus]

Who would've thought

[serene cipher]

nope

[glad lichen]

Ye it’s what an api is for

[dry citrus]

Except maybe i wanted to use gpt for help? Idk, just a thought, you know.

[glad lichen]

Is there a planned release date for that

[shadow wigeon]

gpt-4?

[honest geyser]

Well it's not good at maths, so here is your answer. Help is different to answering the question for you btw

[serene cipher]

uhhh.. yea.. that is the purpose of APIs

[hybrid vortex]

so should I start learning react native if i ever plan on using chatgpt API?

[dry citrus]

As I have already stated, gpt4 got similar questions right, hence why I am asking about it.

[serene cipher]

no announcements about that

[glad lichen]

This. It’s literally called an application programming interface

[dry citrus]

In case you didn't know, I mentioned if someone could ask it a prompt, asking it to answer a prompt does not mean answer a question and that's all.

[zinc flame]

Anyone else get history back yet?

[long pond]

Try #dev-chat for help regarding that

[honest geyser]

Ok.

[serene cipher]

you can use the API on anything capable of making HTTP requests, if you have any coding background you can most likely use it in some way

[glad lichen]

You don’t need to know react native specifically. Basically any language with either an openai package/library or HTTP functionality

[hybrid vortex]

thankies ❤️

[dry citrus]

But if you're going to be against the use of gpt, I wouldn't put yourself here.

[honest geyser]

No I just think if you're a student you should use it as little as possible since you know, it won't be there for your exams.

[glad lichen]

.

[dry citrus]

You made a pretty broad assumption

[hybrid vortex]

but the image integration isn't just for the API right? it's still also coming to ChatGPT, right? 😓

[dry citrus]

Not everybody here using it for math is a student who is trying to cheat

[serene cipher]

I think if you're a student you should use it as much as possible since you know, it won't be there for your exams. =P
if it is cheating or not totally depends on how you use it

[honest geyser]

That's not what I said lol

[dry citrus]

Regardless, how they choose to use it is hwo they choose to use it.

[glad lichen]

Do you not want to use the api?

[dry citrus]

What did I say that you didn't say?

[hybrid vortex]

I want to 😦

[honest geyser]

Making it answer questions for you will not help you learn the concept.

[glad lichen]

I think he means for hw and stuff

[atomic terrace]

chatgpt keep loading here.... is't down?

[honest geyser]

Use it to help you learn not do it for you is what I meant btw.

[dry citrus]

Triple

[wind urchin]

That is a dumb proposition

[serene cipher]

indeed not

[dry citrus]

That is one of the dumbest things I've read in a while

[honest geyser]

haha, how so?

[glass carbon]

He's right. You have to learn to problem solve on your own. You can use it to teach you concepts

[livid steppe]

Just make it teach you, instead of making it answer for you. Problem solved?

[wind urchin]

If you can make it answer then ask it for what knowledge I need to solve this without any external help it will explain the concept very well

[ocean solar]

Use chatgpt to teach you how to use an api

[honest geyser]

how can you argue that letting something or even someone for example help you learn the concept, perhaps you might in some way but in an exam it won't help you that much.

[serene cipher]

there is a huge gap betwen using ChatGPT as a homework machine and as a studying ground

[dry citrus]

GPT is a very valuable tool in that it can interact in a human-like manner when you need it to. So when I'm unsure of the answer to something, GPT can provide an answer as well as an explanation, which it does.

[honest geyser]

.

[serene cipher]

it is your choice if you use it to learn or not

[honest geyser]

I know, I was just saying my opinion.

[hybrid vortex]

I think this is for me but yeah I'm on top of it now. thanks!

[jolly hornet]

I wonder when gpt-4 will become free, is it only when gpt-5 releases? That would be lame lol

[wind urchin]

Hey Luigi when do you think Google will release their LLM

[hushed lark]

It's really really good at teaching concepts, especially the nitty gritty of things as you run into bugs and the like honestly.

Way better than spending hours searching for stuff over Google

[dry citrus]

Clearly it was your opinion, that doesn't need to be said 🤦‍♂️

[honest geyser]

Very nice of you, thanks 🙂

[dry citrus]

That also doesn't justify it

[serene cipher]

I totally agree with you, people who use it (to cheat) for school work are not doing a good use of it

[honest geyser]

You misunderstood and that is no longer my problem.

[serene cipher]

they tried it a few weeks ago and it flopped 😆

[dry citrus]

It is your problem when no one understands what you're saying

[blazing patio]

I love this answer lol

[honest geyser]

@serene cipher did

[dry citrus]

buddy?

and according to you I misunderstood?

[rocky summit]

This may be a silly question/topic, but I've been using GPT3/4 to help generate Golang code. Given it has a Sept. 2021 cut off some things work and others do not as some SDKs have changed. I can generally get it working but I can't seem to get GPT to "remember" the functions it was trained on don't exist anymore.

The Bing integrated chat seems better at that, especially if you are on the SDK page and ask it to generate code based on that page.

Is there any thought that we, as users, and our interactions, can help "train" it to be smarter than 2021?

[honest geyser]

You quoted the wrong message.

[dry citrus]

Which message did I misunderstand 🤦‍♂️

[wind urchin]

Guys dont fight cmon

[acoustic flicker]

are plus upgrades only paused due to the server issues rn?

[wind urchin]

Yes

[honest geyser]

The one where I said "Use it to help you learn not do it for you is what I meant btw."

[serene cipher]

problem with too many people on the chat is that people are replying to each other having many threads.. it is hard to follow...

[wind urchin]

They will be back when they become more stable

[serene cipher]

scroll up and try to figure the replies and the order of the messages, and you will understand the convo

[dry citrus]

I never replied to that so how could I have misunderstood that?

[warm jay]

How to use

[honest geyser]

If you didn't then yet again it isn't my problem.

[dry citrus]

Well at least you admit your wrong

[honest geyser]

I can when I do something wrong but you not reading something isn't my wrong.

[crystal field]

they didn't strike me as clever sometimes

but i've been running benchmarks on codegen in different languages

[dry citrus]

That isn't your wrong, you were admitting that you were wrong in saying I misunderstood something I never read. And in a few other things, but it isn't my problem to explain your flaws to you.

[serene cipher]

@dry citrus@honest geyser guys.. no need to argue over who misunderstand what.. that doesn't matter, chill

[honest geyser]

lol I am very chill

[dry citrus]

He started being toxic when I asked if someone could put a math prompt in gpt4, I don't see why he has to get all touchy over someone's simple request

Didn't even involve him

[honest geyser]

I did sound condescending and I apologise for that 🙂

[final forge]

Very likely what happens when we give feedback for answers on chat gpt.

[steel tide]

Ah I see, thanks. I would be interested to hear more if you would like to share. Otherwise, thanks for the response.

[hidden verge]

25 messages every 4 hours is ridiculous

scammer

[earnest sonnet]

GPT 4 still uses old libraries for coding

[serene cipher]

sure, feel free to ask anything

[crystal field]

3.5 is probably better for most everyone

[restive fern]

so whats good bout gpt 4?

[hidden verge]

bill gate scammer me

[rocky summit]

I would like to think so and perhaps this is right, within the conversation. I have the secondary issue that all of my conversations error out and I have to delete them so I can't say for sure what a conversation over time might reveal. But it certainly isn't carrying over to a new conversation asking the same question.

I can even tell it, "The current SDK doesn't have the ListContainers function, but has the NewListContainersPager function. Please generate code with the correct functions." And it won't 🙂

[honest geyser]

yeah it's a refined model that is more optimised to basically stop jailbreaking. instead of being updated.

[wary star]

Well, GPT does not exactly like coding hard stuff, like large c++ scripts, it hates c++. And it explains too much.

[hybrid vortex]

oh crap. my browser closed and the chatlog refreshed. RIP

[honest geyser]

that's entirely separate from the topic of that channel

[crystal field]

i was wondering if forcing the import or playing around with the header changes it

[serene cipher]

it is a chat completion model, the exact same implementation as 3.5, just change the model name to gpt-4 and you are done

[cloud apex]

same bro

[earnest sonnet]

Oh

[serene cipher]

the missing chat history is a issue that they are fixing right now

openai is aware of this issue

[hybrid vortex]

do you or does anyone know if there's a limit to how many chat threads our accounts can hold?

[serene cipher]

I don't think there is

well, if there is, no one has ever reported reaching it that I know of xD

[hybrid vortex]

okay good, its just gonna be a bit of a hassle deleting new threads then. EZ PZ

thanks ❤️

[final forge]

On longer conversations, GPT doesn't recall what you said at the beginning. It hallucinates. The feedback thumbs up and down buttons are there to train a future iteration of the model. To refine it but it still mostly trained on 2021 data as I can tell.

[serene cipher]

there is a "delete all" button if you want to get it clear

[honest geyser]

scariest button ever

[rocky summit]

Gotcha. I'll do more of that then if it helps. Thanks.

[trail pumice]

GPT-4 truly is a marvel

[honest geyser]

i would lose message from december of last year, would be a shame

[hybrid vortex]

yeee nah, i need my past convos 😆 thank you tho

[serene cipher]

the lack of history right now is just a issue with the system, the history was most likely not lost, they will restore it

[trail pumice]

i just give it a vague description to make something in C++ and it just goes right ahead and makes something

[honest geyser]

very cool I agree

[serene cipher]

indeed, if that helps, the button has a confirmation after you click i t tho, it is not a 1 click history nuke

[honest geyser]

ik

[trail pumice]

what are 4 more nodes that you find to be missing from the engine?

[hybrid vortex]

I've fed a lot to it and trained it just how i like it. imma just sit in a corner and wait until they restore it. 👋

[trail pumice]

the response

Here are four more nodes that could be useful for game development:

Get Actors Within Frustum
Project Point Onto Line
Get Random Point in Navigation Area
Interpolate Actor Transform

with all the code needed to make the nodes

granted it prefers to do maths stuff

[rocky summit]

I did have on reply where in the first paragraph it said the NewListContainersPager function doesn't exist so you can't use it. And in the second paragraph, of the same answer, it said, but the NewListContainersPager is what you want to use. Then listed the exact code I already gave it.

[trail pumice]

still cool though

[spice pollen]

prompot

[serene cipher]

looks like you are doing some rendering stuff

[final sapphire]

Do we know when GPT 4 will be back online for paid users?

[hybrid vortex]

oh hey! speaking of, I saw earlier that someone integrated chatgpt into unityeditor

[trail pumice]

@serene cipher no

i am making a UE5 plugin for the marketplace

[hybrid vortex]

ahh~

[serene cipher]

it is working

[honest geyser]

oo, that's really cool

[trail pumice]

@hybrid vortex i saw, still neat though

[hybrid vortex]

hey, a few more months or a year later, it might come to UE5

[trail pumice]

meh

[serene cipher]

close enough :V

[trail pumice]

i can just ask it to make me a plugin to implement itself into UE5

think smarter and absolutely not harder

[serene cipher]

it actually works

[trail pumice]

it is big brain time

5Head

[honest geyser]

that's really useful cause I have a college project that I wanted to use UE5 for

[serene cipher]

you can ask chatGPT to implement itself on many things and it will give you a small script for that, it is so meta xD

[stiff finch]

this is just making us even lazier 😂

I love this

[brittle birch]

Buttering us up

[trail pumice]

first we thought artists were out of a job with Midjourney

now even us programmers are screwed lol

[honest geyser]

well there needs to be someone to prompt engineer

[stiff finch]

absolutely right

[serene cipher]

I would say, it is quite the opposite actually...

[honest geyser]

hop into the basket as soon as possible

[trail pumice]

i asked it to make me an AI with a neural network and it just went ahead and did it

[pseudo geyser]

The secret is to get the lawyers worried within the next 6 months or so

[trail pumice]

i am not sure what it did as i didn't try it out but still

[serene cipher]

those are such good tools that help me to do less of the boilerplate stuff and do more creative / problem solving work

[stiff finch]

chatgpt can just be your lawyer xd

[honest geyser]

i think prompt engineering has insane potential in the near future

[trail pumice]

no, there is donotpay for lawyer stuff

[long pond]

You're a lawyer?

[trail pumice]

it is currently being sued because it pretends to be a lawyer or w/e

[serene cipher]

actually, with the 32K model, it will be pretty good at summarising long jargon filled contracts

[long pond]

Plenty of better stuff also

[final forge]

I'm guessing that the 'temperature' setting from the playground comes into play. That is to make it seem more human sometimes the next word it chooses is one it wouldn't normally choose.

[serene cipher]

im a programmer, I have been using Copilot since its beta.. and it worths evry penny

[hazy pollen]

Is there any possibility of a localised Canadian version that might be more affordable than the USD option?

[trail pumice]

@long pond idk, it is only for the US anyway so it is useless

@serene cipher my resume is going to say

Knows how to use ChatGPT

[burnt musk]

There needs to be a "Download Conversation" button as well, so we can save the information locally without having to manually copy and paste everything into a text document

[trail pumice]

when i am going to look for a job again

[dense nova]

Hi. I'm new here just have a question about gpt4. Is there anyone from OpenAI here? I'm a Plus customer but gpt4 IS NOT WORKING as promised / advertised. And it's way too slower, but a slightly a bit wiser than default model. And there is certainly nothing like “feeding with image”. It does not even know that it's gpt4. Very disappointing, considering to cancel my plan.

[honest geyser]

there is I think?

[trail pumice]

send openai an email

they won't respond, but that is the only way to get in contact with them

[serene cipher]

man, last time I updated my resume, I asked chatgpt what it thought of it 😆
and then, I actually used the advices it gave me to improve the wording on it

[burnt musk]

Where?

[livid steppe]

yeah the only way this situation will get better is through competition - once there are multiple competing LLMs out there, customer experiences will be more likely to improve

[honest geyser]

I can't really tell you rn since conversation history is down

[trail pumice]

@serene cipher add me as friend

[serene cipher]

agree, but.. a lazy way to do that... just downlaod the whole page's HTML

[trail pumice]

or nvm

your resume should stand out from others

i used Canva for a design

[vale creek]

careful though as there are also sites that can work out the probability of whether someone's supplied story / resume / blog / text is AI generated 😉

[trail pumice]

@vale creek tell it to make it look like a human wrote it

i guarantee you nobody will find out

[serene cipher]

sure, feel free to add if you want =)

[hybrid vortex]

I like to think that chatGPT for programmers are grunts (if you train it well enough, can also be QAs) but it'll still depend on the programmers where to put the bits and pieces of the coded lines together.

so like a puzzle solver with a few extra steps.

Thanks chatGPT

[serene cipher]

I find Github Copilot to be very passive aggressive....

so... yea, it is very accurate

[patent cypress]

Does anyone have information on reports or articles that describe the current Chat GPT trend?

[serene cipher]

not sure if I understood the question

[north birch]

333

[hybrid vortex]

you can look it up using bing, much easier than sifting through us users what the latest news about chatgpt is at

[serene cipher]

articles about OpenAI.. in general are over-hyping it waaay too much

[patent cypress]

My senior is looking for a well-written essay about the newest chat gpt trends.

[patent cypress]

Thanks

[hybrid vortex]

yeah but internet journalists and article writers love thatt tho

[hybrid vortex]

no probs!

[serene cipher]

yea, they are surfing this hype, they don't want it to die

[gleaming umbra]

Meow

[junior apex]

When the hell are they going to remove the restraints

25 every 3 hours is way to low

[hybrid vortex]

with this in mind, learning how to code still is has huge merits because the more you know about what you want chatGPT to do, the better result it'll yield.

[hidden rampart]

I am really interested to be able to is there a quickstart on the CHATGPT4, I am new to the platform but want to help however I can

[brisk heart]

can someone who has chat gpt help me out, im a broke student

i need to see what response the prompt is giving

please 😊

[junior apex]

Sign up and use the free version. Shesh

[lament chasm]

give bing a try first

[nova obsidian]

anyone know why my replies are constantly including the end of my inputs, punctuation and sometimes a word or two.

[serene cipher]

a lot of people seem to over-estimate chatGPT's ability of making code... people will ask it to make a complex open scopped system... then they complain that it is garbage because it could not make their disruptive app idea 😆

[wise bough]

just tell bing not to use the internet and then it acts as gpt 3.5 turbo

[hybrid vortex]

it's like asking a digital artist to make a visual novel that'll be a popular hit in the market and leaving without any more information. KEK

[serene cipher]

ChatGPT should be working with no limitations, the gpt-4 model on the website is more of a "Nice bonus" than anything else

[junior apex]

Oh come on. It’s a main reason to buy

[latent widget]

Paid users of CHATGPT cannot log in and use, please explain the reason

[junior apex]

Go from 100 to 25? I mean, what the heck

Absurd

[muted shadow]

Hi all, can I test image captioning in gpt-4 playground?

[hybrid vortex]

using up 25 words per 3 hours means that you're either extensively training your chatGPT4 for a complex undertaking or you're using it for something simpler that chatGPT3.5 can do with ease.

[acoustic flicker]

when do you expect openai to deliver server optimizations for gpt4 so that the limit will be increased or removed?

[junior apex]

3.5 is not even close to the accuracy of 4

[hybrid vortex]

i mean, can i ask what you're using chatGPT4 for?

[trail pumice]

you need to understand that it is being used by millions of people at a time most likely

it sucks but it is what it is for now

[junior apex]

IT related things

[latent widget]

Can your chatgpt log in?

[fast pagoda]

yeah 4 is much smarter but the cap is annoying granted a good chunck of the world is using this site

[junior apex]

3.5 thinks O365 business standard comes with P1 licensing.

4 answers much more correctly than 3.5

[boreal hull]

yeah gpt 4 is useful for school concepts

[hybrid vortex]

oh i see where you're coming from then, but it is what it is. 25 answers for 3 hours would suck if you're heavily reliant on it for your job or IT related issues.

[limpid raven]

how can I login to GPT-4 ?

[young harbor]

Is it real ？I mean chatgpt. I tried so many times to login the platform，but it feedback only ’ something wrong‘ while I

asked 3 questions

[junior apex]

It also helps script better. Why would I want to drive the Nissan when I can drive the Tesla

[young harbor]

that is disappointing

[junior apex]

Just a coward move to lower from 100 to 25

[river salmon]

Is gpt not working again?

[wind urchin]

No that feature isn’t public yet

[hybrid vortex]

i think this just means that you have to adjust your way of using it since I dont think they'll do anything about the capacity anytime soon.

[junior apex]

Then they should have not started at 100. Just going to anger people like me who bought to use it

[young harbor]

Why chatgpt always feedback as ：Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[hybrid vortex]

did you sub on the day they announced the chatgpt4 100? cuz that didnt take too long to limit the cap

[junior apex]

Business sense would state start lower and go higher

[young harbor]

Does I didn't pay for it matter?

[river salmon]

ps : Something went wrong, please try reloading the conversation.

omg!!!

[junior apex]

Last Thursday

[hybrid vortex]

true, that would also mean implementing higher tier subs

[junior apex]

Then they should do it

[still terrace]

Question: is GPT-4 available to all users now?

[hybrid vortex]

we gotta make do for now tho. wait what they'll do in the meantime. also wait for the chat history to come back

[bold crag]

Is there an ETA for the 32k model for Plus users?

[wicked storm]

Have new bing used the gpt-4? How to confirm it?

[full birch]

I hope they drop it lower every time someone complains about it. We'll be at 0 messages per day of GPT 4 in no time.

[serene cipher]

there are no announcement on that

[bold crag]

Yeah, would like to know a timeframe though, 1 week or 2 months

[young harbor]

I even start to

[serene cipher]

me too

[young harbor]

doubt about chatgpt

[burnt musk]

I'm using GPT4 to come up with complex, creative concepts to flesh out a fantasy world I'm making. So far, even the 25msg/3hr limit is plenty for me, as I'm probably spending ~20 minutes writing out the details for it to analyze and respond to for every single prompt, meaning that in 3 hours i'll probably only ask it 9-15 questions.

What i find most limiting is actually the length of the outputs, when I give it a 9000 word description of the pantheon of gods and mythological creatures in my fantasy world, and ask it for more details, and it just spits out 1-3 bulletpoints for each entity, I'm a bit disappointed, and have found that i generally have to describe 1 entity at a time, and given it's memory limitations, by the time I get to describing the 7th god in the pantheon and asking for it to come up with a relevant mythological story, it's already hallucinating false information about the first god i described.

[wind urchin]

Didn't gpt 4 have the image feature for a short while when it was released

[coral drum]

The skill of asking questions is very important

[magic shoal]

maybe preview

[young harbor]

Got any creative ideas for a 10 year old’s birthday?
!
Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[serene cipher]

hey, I have been using the actual API to do similar stuff, and it works waaay better than the chatGPT website due to the fact that you can set the system prompts, the AI will stay in character propery;.. also you have the controls over temperature and penalties, in case you are not on the waitlist already, you should really get in

[spice plaza]

I think d&dGPT would make a good product using the API

[lethal vapor]

I can't upgrade to Plus right now because it says too many people are upgrading. Does anyone know if we are we going to be able to upgrade to Plus again soon?

[hybrid vortex]

I saw that Pathfinder is against AI based content tho

[wind urchin]

Did you contribute in the evals to get access to the api

[serene cipher]

no, I signed the list very soon

[shy falcon]

so is GPT4 completely down rn?

[warm mirage]

beware if gpt4 is leaving the answers incomplete

[lavish cosmos]

Any good prompts for development projects

[serene cipher]

what a shame, I made an AI personality that helps me with Pathfinder rules

[wind urchin]

Hello, may I ask why the history of the OpenAI product dell-e2 cannot be deleted?

[wind urchin]

I signed up on 14th March still yet to get it oof

[burnt musk]

I'm not sure what that entails, didn't i read correctly that I'd need to have an app to use the API? also, i dont know what temperature and penalties are in this context

[hybrid vortex]

I mean, no one has to know, I won't tell 😉

[serene cipher]

i think they just did one wave because well.. just the current amount of users are already overwhelming the API...

[serene cipher]

i mean, I implemented the API because I could, but it is perfectly usable on the OpenAI Playground, no need to know how to code

[tardy heart]

There's some really good utilities in the discord as well

[serene cipher]

temperature and penalties are some of the settings that control how the AI act... basically, things like how """creative""" it will be, by deviating of your original pmrmpts and adding information that wasn't originally there... how much will it avoid repeating words, atc...
those are things you can only control yourself when using the API, and having a slightly higher temperature and higher penalties for repetition helps the AI to be very creative when making fiction for RPG games, it does well i ncreating magic items, lore, and pretty much anything

[hybrid vortex]

OOOH shoot. i didnt know about the playground

[serene cipher]

https://platform.openai.com/playground

[burnt musk]

also, @serene cipher if you have a link to get on the waiting list, i'd love to see where i can explore it further. I've found that GPT4 is helping me essentially create a whole encyclopedia for a world that doesn't exist, and being able to organize it and have new prompts reference old outputs without having to just add more and more and more reference text

[serene cipher]

jus tone thing, the API calls there are paid
but, every new openai account has $18 in free credits

[hybrid vortex]

thank you! so much for these 1 hour tutorial videos about coding and making software

[glacial musk]

Is the 25 messages every3 hours for subscribers also?

[serene cipher]

here it is: https://openai.com/waitlist/gpt-4-api

[hybrid vortex]

are they gonna send an email confirmation about this? i feel like i already signed up for this

[serene cipher]

wat you want is pretty much a feature of the API xD
the "system messages" which as isntructions that you give to the system and not as a user

so, before the AI reply, you can insert a system message for extra context

[serene cipher]

yes, they do, you get an email when you join the list and another one when you are selected

[hybrid vortex]

ooh thanks!

[full onyx]

new model still aint too good with logical reasoning

cant get it to figure einsteins riddle even with system messages

[serene cipher]

the playground directly uses the API under your account, you pay for each request

[proud fossil]

mate there is an 8192 token context length limit meaning your prompt + gpt output = total context length. Your prompt takes up a portion of that 8192 aggregate and so does the output of GPT, equating to the maximum tokens allowed to be used. 4k your prompt 4k gpt output.

[burnt musk]

Oh, so that's what the 32k token model will do? allow for longer input prompts and generate longer outputs?

[proud fossil]

yes!

[serene cipher]

oh, yea, I forgot to mention that @burnt musk your 9000 word description are waay to long, the AI is most likely ignoring part of it

[serene cipher]

32K model are not available yet

[proud fossil]

32k token limit will be so based, i plan on using it to build a knowledge base for different networking operating systems

[serene cipher]

that might get expensive quick

[proud fossil]

it will

[serene cipher]

💸

[proud fossil]

🫦

money talks

the pricing is divided differently for actual user prompt and the model output so it will probably get pricey

[zinc thistle]

How does the token system work?

[proud fossil]

it's better than gpt3.5 but not perfect