#gpt-models
1 messages · Page 217 of 1
Ok
having a large amount of traffic control (and good traffic control, at that) is the best way to intelligently and efficiently move traffic throughout a city
more traffic control is more expensive and harder, but seems to produce better results
Obviously we don’t know the exact amount but 300 hidden layers is a lot of hidden layers
more parameters is more expensive and harder, but seems to produce better results
So like the inputs would correctly return the right outputs as they travel through the neural network
Hey guys, it there any useful plugins could recommend?
If there’s more intermediary steps and routes
hm
whenever I think of how gpt operates, I automatically abstract it away from the neural model
in my mind, it's some fusion of computer, personality, and mind
I think of it as not being pure but idk much about actual neural models tbh
But that’s because it is the result of multiple models, right?
Limiting and constraining chatGPT and giving it its ethical framework
you don't need to know too much about neural networks
I’m probably going to major in machine learning
in fact, it might aid a more useful understanding if you avoid thinking of it as a massive neural network
Lol undergrad comp sci major
or do machine learning at uni haha
How do
why cant I use gpt 4 anymore?
all I'm saying is that you might be a little too focused on trying to understand gpt as specifically a neural network
So*
it says i was switched back to 3.5 randomly
Im not trying to understand it’s functionality as it appears to me as@neural network
But I am@fleet pumiceterested in learning the fundamental stuff with Ai since idk how many none-fundamental jobs will be left when im d
Old*
how do I switch a thread back to gpt4?
You can’t
You cannot
maybe. The reason why chatgpt, and openai's API is absolutely blowing up atm is because you don't need to know about, say, reinforcement learning and calculus to be able to leverage these models.
Yes for sure
I'm very glad haha
I agree 100% this appealed so hard to the common public and is honestly a turning point in technological progression globally
(there's an open-source storm brewing off the ChatGPT coast, but it hasn't hit yet)
so
Maybe a revolution
VoxScript seems amazing, it can summarize youtube videos for you
chatgpt is the fastest growing app/service of all time
Yeah, people don’t always get how fast this progression has been
a lot of people compare current-gen AI to blockchain, cloud, virtualisation, etc
this is generally as a rebuttal to the notion that this is the most important advancement since the internet
It’s generated a lot of unicorns way more than other economic booms
but blockchain, cloud, virtualisation, etc, did not grow anywhere near this quickly
Lol that’s a big misjudgement imo
Like very big
generally, this stems from people who have no (or very little/limited to the free chatgpt3.5 version) first-hand experience with these tools
does anyone have any idea when the code interputer extesnion will be given to plus users
The growth cannot be explained as a trend, it’s so new and has already revolutionized people’s power by enabling so many to take shortcuts
anyone else having export issues?
the only reason why someone might not yet have experience with gpt3.5 or gpt4, at this point, is pretty much purely because they don't understand the implications of this technology
If people still don’t see the potential of Ai and seriousness of it idk what to say tbh
Ai has@been crazy for a while but the sudden boom of public access alongside the power of gpt models started a new era
im mostly getting click/reading content failed with the web browsing part of ChatGPT, is this a bug? (ive checked the robots.txt file of the websites its going to, and it isnt blocked)
most likely due to massive usage atm
it's just recently* rolled out, and is very much in beta
ah ok
I've had mixed results
AI is not perfect yet
guys
Yeah but if you graphs it’s progression against our progression in ability to control it it seems kind of unsettling
using the wolfram plugin, i just calculated that it would take 343,187,316 bananas to equal the amount of calories in 1 gram of uranium
best plugin ever
lmao nice
Like maybe I’m overlooking something but at the very least we’re gonna have elites with extremely powerful AI that we never even know about
what about coal or oil?
Haha it’s such a good plug-in
lol
Will AI One Day Take All Jobs?
just got access to PLUGINS
Potentially, But Probably Not.
how can I promote an air conditioning company?
In fact they will create jobs because they are almighty and merciful 🙌
nice
Go into peoples houses with a some form of heating device when they’re asleep to create demand
Make them sweat like crazy but don’t let them find the heater
Then advertise at their most vulnerable moment
the instacart plugin is underwhelming
It’s like 20 million calories isn’t it?
How is that a rebuttal at all, the blockchain in-particular has not turned out to be revolutionary at all outside of cryptocurrencies, which haven't amounted to much outside of greater fool theory.
Yeah some of the official plugins honestly aren’t all that
🙂
They’re basically just affiliate advertising for their own products
good job
exactly correct, but this was generally in conversations from a few months ago
the browsing capability is also incredibly flawed
however, there are still definitely people who think this way

This is obviously on par with the internet revolution ? Am I just stupid
yeah, I'd say it's about as important as either the world wide web, or the internet
No it’s definitely there
Like im not overreacting when the data is so combing
sarcasm?
it's not yet there
Convincing no not sarcasm
Which is easier to use, chatgpt or bingAI?
GPT for specific questions. Bing for browsing.
the current state of AI is, probably, quite similar to the state of computers in the 70s
Yeah It’s not there yet but idk how that’s really relevant since we clearly hit the point of exponential growth recently
At least in terms of this field of AI
Generative Ai
depends on a few things, mainly the alignment
Which could have the potential to consolidate all other forms in a way to create autonomous Ai
LLM for critical thinking and plan formation
thank you
it could plateau
in fact the current architecture is expected to show diminishing returns
"autonomous" is a strong word
Well not@fully autonomous
we need lots of breakthroughs for it to go right
doubt
Because fully autonomous general Ai seems like a decade or two away before it’s really dangerous (idk with the rate of progression rn)
I can't post links
Would you agree or
google site:openai.com Language models can explain neurons in language models
openai recently used gpt4 to analyse the neurons in gpt2
That’s dope haha
the models themselves are so incredibly complicated that they might as well be a blackbox
Bro@what kind of side projects are they cooking up@fleet pumice there lol
this kind of stuff is lifting it out of blackbox status
Theyve@got the most op gpt applications to themselves I imagine
the only reason why we use current training methods is because we can only see what the model does, not necessarily how it works
but if we can understand exactly how models work
I think there'd be massive efficient gains in training
Yeah that’s not Normal and it’s definitely the first time an Ai has directly provided so much convenience to such a large amount of people which indicates that it’s really powerful and also likely to be popular
It’s set up to be an industry distrupting technology in every way
or perhaps even, "training" is not required - you could just build a model to do exactly what you want
Once we start to turn the blackboxes into greayboxes into possibly whiteboxes, I think there will be new training methods we can't even really imagine right now
that’s some real science fiction stuff… for now
with results we can't even really imagine right now
It’s probably not even that far fetched
From the perspective of the most cutting edge researchers
That’s incredible tbh
This field is setup to have soo many points of explosive growth like
would strongly suggest reading this
It’s nuts. Gonna read it now
What is the term black box mean. I would assume it means you can't tell how it came to the conclusion it did because it's so complex
somewhat - technically it's not really a great term though
@wet drift are you familiar with hash functions?
I'm sorry I am not
hmmm
if we could find a way to always deconstruct behaviors in deep learning Ai no matter how complex imagine
when I say "black box", I mean that there's no feasible way of analysing it
Yeah im sure
the models are so incredibly large and complicated that any classical approach to analysing it would be essentially impossible
Well that same type of complexity may find it’s own non classical approach to doing it that we’re then not able to deconstruct lol
this means we can't really design models by understanding exactly how they work - instead, you design models and guesstimate how they work based off their outputs
Hmm it's interesting though because imagine an AI that was build for that proupose. Just to feed it large amounts of data on how the ai works
Yeah that’s what so nuts to me with Ai
if there was a method of effectively analysing the model itself, the entire process we go through to make a model would probably be overhauled
Just got access to plugins! What should I use it for?
training would become an archaic model creation method
Your just kindling a simulated petri dish of computational power designed to be very powerful at some task and you can’t know ifs@full intentions or behavioral tendencies
Right yeah for sure. When we "unlock" quantum computing do you think that will be a way into creating a sort of check on ifnthe ai is providing factual data
Lol imagine someone does this and they go down in history as one of the most@legendary scientists ever and then we go into dystopia
I feel like that would be an sccomplishment but almost certainly wouldn’t be the result of the human
To the same extent as previous discovers
Or@rather wouldn’t be the direct result of the labor of the human
honestly, this is what I think will happen soon™️:
- New NLP model (let's say GPT-5 for funzies)
- GPT-5 is a massive improvement on GPT-4.
- We use GPT-5 to analyse GPT-4, and completely understand how and why it works (very hand-wavy, I know.)
- You can then procure custom versions of GPT-4 where everything except only what you require has been stripped away. We can do this because we know exactly how the model works.
Medical and scientific research breakthroughs of the future seem likely to be heavily aided by AI
Medical ai models have been killing it long before got
say, you get a gpt-4 model where everything has been stripped away except only what is required to produce python code
Hm how do they train the existing variations of gpt models?
Like gpt4 code generator or whatever
Vs the normal model
what do you mean?
Cpuld@you compare it to find tuning a model on coding data
It's such an interesting thing to think about. Will the ai get to the point we are no longer able to make certion decisions ourselves. Let's say for example the stock exchange. "Could be this way already I'm not too fermiliar with it" but imagine a point where you wouldn't trust a human to male those decisions for you because the ai has so much more data that it would become impossible for any person to do anything better in that scenario
Gpt-4 is used for like copilot or smth right and has its own model
?
Trained by OpenAI or maybe im thinking of gpt 3
Consider that there already are decisions that we're only barely able to make - like trading info
@sullen wren you my be thinking about Codex
AI is honestly already controlling us lol
Behind the scenes as it@mines our data and hyper classifies us
The AI doesn't make us not able to make decisions - we already couldn't make these decisions; the AI, however, can
Now of course I am using that as an example but something like that leeching into things we think of now that we would tipical ask a professional for we would now turn only to ai because it is so much smarter then any human
if we create something more powerful than ourselves let’s just celebrate and call it quits
That would be the second best outcome of a civilization
if we can create perfectly aligned AI, this might become less of a problem
And species I mean that’s not bad
Hmm
Very true I bet the top companies and research labs will be the places@where that really matters
so gpt 4 on gpt plus is better than bing chat right? 'cause i keep getting these short answers on bing chat, like it feels like a nerfed version of gpt 4
Bing chat is way different and more niche
niche how?
different system prompts - microsoft made Bing try to be brief/curt
It’s more for simple research and web browsing
With language to fill in the gaps
ChatGPT is a lot more
I'm interested to see what this will do for the medical field. Specifically but not limited to the ability to diagnose things along faster and more accurate
At least in my experiences
sometimes the answers i get on bing chat are amazing. but sometimes i cant get more than a few sentences, its weird
I actually have noticed that they’re great
I’m just salty they limited it to that specific use case of searching
Like all they wanted was to boost bing search engine
Does anyone know why I can't activate the beta version even though I have a plus plan?
There is an AI robot who's goal is to pick up a ball and put it in a basket. There exists a red STOP button that you can press to completely stop it, just in case.
Well-aligned AI: My goal is to put the ball in the basket. *does so*
Badly-aligned AI: My goal is to put the ball in the basket. The stop button, and by extension, the human operator (and by extension the entire human race) is a threat to this goal. I must neutralise this threat, *skynet time*
Maybe it worked idk
That’s a concerning thought experiment though…
On the other hand
“Does so” makes no sense
What does that mean what did the ai do differently, in the scenario where the human was actually in the way
alignment is about how truthfully and correctly the AI will complete the task
If they truly gain power over us we will not@have leverage in that situation like the wording implies
I do wish on chatgpt they would allow you to remove a message from memory in a single session. That way code wise if it generated a bad batch of code but the previous code was good you can remove it
Because theoretically they’d make the decision with the least resistance toward@their goal and once we lose control of them, an incident with us may not result in that much opposition
Lol i don’t actually worry about that
But its funny to think about
Because its hard to know if what I’m saying is completely off or not
Provably@somewhere in the middle
Eh nvm
Ohh@I see
Chaggpt is@badly aligned
Because it won’t tell me how to rob stores and stuff
no, it's very well aligned
my second example is actually not great
Hey everyone
Very well aligned ? It seems impressively aligned for our knowledge but
What does proper alignment truly mean
What are all of your thoughts on the "segment everything"
so, chatgpt's goal is more than just "help" - it's also trying to navigate ethics, morals and politics
Thr implications of that are neat
Is it an actual measurable aspect of the Ai or is it a more subjective thing
its goal is to not provide information like "how to rob a store"
you could say that when you ask chatgpt to write python code, and it fails at that task, that's an example of the flaws in its alignment, though take that with a grain of salt
Ok but ifs@ethical framework seems like a completely less advanced but hard coded competing model trying to keep it in check
But not being very consistent or good
After a little bit of prompting , speaking from the perspective of the average user probsbly
Since you can jailbreak it super easily
How to mit god
it being "jailbreakable" is also a good example of the flaws in its alignment
To be fair gpt 4 is a lot more consistent with producing good outputs and censoring bad ones, if gpt 5 is near flawless i@may change my mind about something’s
Is it though? Isn’t there utility to be found in an Ai that doesn’t constrain its behaviors to predefined ones
That’s how novel content is created and new insights are produced
"jailbreaking" it means you're getting it to do things it shouldn't
I think@chstGPT is doing it correctly right now don’t get me wrong
Like idk what else they’d do and obviously the model needs some ethical framework
it's very much a multi-domain problem
how could you even go about doing something like this with a person - instilling an ethical framework and moral code, but also allowing "free will" and intelligence
that's a hugely complicated task
Im just saying “jailbreaking” is essentially the same as guiding in typical prompting.. no? No actual difference from a behavior perspective so to attempt to constrain the bad behavior may interfere with the rest of its functionality
in practice I'd say you're right
Which I think is the smartest approach easily rn
but my point stands
To slow down the development
Yeah lol it’s very hard to be the people who have to@do@that I bet
But it’s also so@important
Incredibly important
OpenAI is definitely giving me totalitarian vibes ngl lol I want to do a deep dive research on them one day
potentially
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
consider
what
I looked away and when I looked back I saw a spanish error message disappear
That illustrated the point I was about to make so well
hahaha
more reading material
or at least read the executive summary
(can't post links) google: "We Have No Moat, And Neither Does OpenAI"
I can’t tell if OpenAI is trying to try to become like the next dictators of morality buttt it seems like it might be their goal 😭
And@they’re positioned well to do it and that type of publicity / social profile is known to benefit tech companies
idk this is why I want to do a deep dive
this technology might be one that has no "owner"
like computers, or the internet
simply various companies that all do things a bit differently
The technology@that has no owner will be revealed to us shockingly fast I think
Maybe not but these elite labs are almost definitely keeping the most powerful tech under the reader
mmmmaybe
this is an emerging technology - most powerful is generally also most cutting-edge
I do like how much effort openai puts in to making these models ethically and morally capable
To an extent easily because at the least they have access to basically infijite@cpu power, infinite rate limits, etc etc any amount of programmagical interference
Using@their private Ai
there are still the limitations on the models themselves - tokan I/O
Yeah, but many models spend months before@being published
Hard to know how long the first working product existed before getting to the public
Do you think guys think it's really a scrambling race between Google and openai right now or do you think they are just doing they're own things not caring about the other
this takes a lot of time and effort
(can't post links) google: "We Have No Moat, And Neither Does OpenAI"
Well they’re def very relevant@to eachother
Wait OpenAI is prob legit just warring with these other companies
this is a paper by a google employee talking about how open-source NLP is poised to outrun both openai and google
Even with this discord and that’s why they’re so strict
inb4 ban for discussing the illuminati :p
Im just discussing common human behaviors
That always emerge when people are@fleet pumice this situations
Im down for like
I’ll just immortalize myself and then dip
Once I’m uploaded I can just terminate this instance of myself
@sullen wren you should play Soma
Enhancement is our only hope. Augmentation, lest we succumb to our own creations.
🙏 🙌 🙏
the moment I realised the weakness of my flesh
🦾
🦆
🦵
Look if we can like yolo clutch comeback by fully converting to ai with traces of our memories but not really us
mmmmmmmmmmm
But still overcome the Ai rules then I’m so@down@that would be so epic make it@happen OpenAI
have you seen Red vs Blue seasons 9 and 10?
I have not
I pictured some matrix stuff though lol
create a website
did you see that snapchat influencer made 70k in a week with dollar a minute GPT version of herself
Who lol this sounds fake ngl
is there a plugin where it can find errors in a code u or chatgpt made and find the error and fix it?
Ohh@wait
👁️ 👃 👁️
Can you use google's AI?
Have y’all seen the rise@fleet pumice onlyfans Ai managers lol
google's Bard AI sucks
so whats the answer?
if you google "snapchat influencer 70k" the top 2 results are forbes and business insider
from may 11
oh...
AI version of herself that her followers could talk to
Oh god
I find it very interesting that bard is behind because google had been working with ai for years.
wait who is a Snapchat influencer lol what
1 USD per minute
That can’t be a thing
Bit that's not to say they have been working towards the same goal as openai
Google has been crushing it
But*
Like fr they dominated all the methodical games / esports
Their alpha go Ai or whatever it was called
User: hey chat gpt 4 if i give you some code, can you fix it plz?
ChatGPT-4: Of course! Please provide the code you're having trouble with, and I'll do my best to help you fix it.
If that doesn’t work:
you dont think ive tried this 😂 it still gets it wrong
Identify potential spots that could cause errors in this code as I develop my project and implement debugging statements at these locations
Well
ask it to try to fix it -> try running the code -> paste error message into chatgpt window, ask for fix
@narrow fiber try cranking the handle, turning the wheel, chooching the machine
Won’t work always especially with longer code vases
It will forget the last error it identified and start repeating the same@ones
split into functions
When is gpt 4 coming to free users
work on one problem at a time
It’s been so long
ive tried this too, up to some point it will give up
probably a few years, potentially never
Ye@but the errors almost@always some stupid thing on the backend with the state management or smth.. point is it’s just faster to debug right after@if gpt doesn’t find it immediately
At least with got 3.5
don't program with gpt3.5
I didn’t trst@it@much on 4 haha
ooft
gpt4 is extremely superior for all code tasks
But yeah
you will 100% save time writing good code slower
Its still just boilerplate
rather than writing "meh" code faster
But a lot of boilerplate
I mean I personally think it’s like 2-4 years away from being almost fully@capable of project development like that with guidance from the human potentially in solely English
it's potentially currently capable of that
For what is currently@thought of as like a lower level software engineer job or entry level
via a program utilising the API - not on its own
Ye for@sure, but I’m talking about widespread implementation that would distupt entire economies and stuff
Haha yeah I’m trying to get on this asap I feel like it’s kind of a rush rn
Like maybe modt@people caught the wave on the way up but its@still going up
part of my university work requires extracting data from websites into JSON files
I opted to use the openai API with strict, long system prompts
so I've been working across multiple gpt threads doing different work, which then coalesces back together
It’s like if you were to quit trying to be an entrepreneur after the internet was invented because you felt you couldn’t reinvent google lol… meanwhile a ridiculous amount of children companies would go on to profit from the internet
I like@doing@that sort of thing lol
General purpose data scraping and classification is powerful
Something the very elite have mastered like crazy but that most normal people don’t even care about at all
a single python/gpt thread that can determine:
A. if the given problem should be split apart into smaller problems; or,
B. if the given problem is small enough to work on now.
If B, it simply writes the program. If A, it spawns more forks of itself, and passes the smaller problems to each one
Ohhh yes threading gpt via api is so nice
this recursively should be able to write code of any length
limiting it to python code should provide enough limitations to make it possible for myself to create
Im very into long form indefinite content generation right now@sctually
this is a recursive problem, hence, the solution must be recursive too
Really want to get on the got 4 api white list so that I can begin implementing unlimited@context memory
Check out the@prompt I made
unlimited™️
Unlimited until it@is
hahaha
whats the new update is extensions available to use?
is there an area for system prompts?
I'm definitely a Proompter™️, but I'm also a System Proompter®️©️
This type of prompt would be so powerful with context memory and threading. And even@more, once you start simulating scenarios in a highly intentional and controlled but autonomous way in chatGPT (taking principles from low level Ai development) the potential outputs are nuts
It's a very expensive model. I'd assume not for some time. They are already limiting those who pay to 25 prompts per 3 hours.
Just start writing a novel chapter by chapter page by page hopping back and forth in the novel making additions and changes as@you constantly reevaluate how to write the best novel so that everything is optimal
And any type of use care remotely similar to that
That benefits from constant reevaluation , iteration, addition, and long form content
The limitations of chstGPT in terms of none-regressive self improvement lie within it’s memory limit
Hm is there a limit to how fast got models can@access their memory cache if it’s like
Huge but through a vector database
I wounder how the model is that isn't restrictive. The in house model
@sullen wren thoughts? https://discord.com/channels/974519864045756446/1108257759201333289
Looked at it just now
what about GPT-4?
Hey gus, I have found that my CHATGPT4 is not working. After asking a question, the cursor just keeps blinking without providing any response. Does anyone know what the problem might be?
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
think gpt4 is like down or somethin
same for me. and plugins don't load
i think gpt-4 server is down
believe so
but my friend's account still working.
Mine is working
Yo they literally have a priority flag attached to us btw
For service allocation I think
I saw@it in the packet data
It was attached to a version of the models with higher bandwidth allocations lol
Maybe that’s pro vs free idk but there were three
They also flagged me as unlikely to seek refund
where are you seeing this?
how do I use pdfs into chatgpt. i got the plug in for it but not sure how to input a pdf
The packets from during a chatGPT session
Where I logged out and back in and switched models and stuff
Triggering most of the data transfers
Mostly I just learned that everyone knows exactly who I am because of cookies which tons of random services were@just pulling @celest plinth me constantly
Down to my exact device specs and MAC address
But that was just the data over the network lol
Internet and plug-in mode not working
hey the dp you are using is of Mai-San from bunny girl senpai ? 😭
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
Can someone type in gpt4 web plugin mode or whatever
“Sources to help understand my thesis of Nz’s 84 year delay in acknowledging samoas influenza pandemic and the reasons for it”
Wdym
Oh lol
From 3 days ago:
We’re rolling out web browsing and Plugins to all ChatGPT Plus users over the next week! Moving from alpha to beta, they allow ChatGPT to access the internet and to use 70+ third-party plugins.
If possible
does that answer your question @wind urchin?
Does bings beta do the same thing
what am I, bill gates?
Well could u do a prompt for me please

Nah I’m being serious I can’t find any sources for this assignment
Oh I misread lol
do you mean before the weekend all the plus users could access the plugins? that is we don't need to do anymore but just waitting?
What do you need
@sullen wren
“Sources to help understand my thesis of Nz’s 84 year delay in acknowledging samoas influenza pandemic and the reasons for it”
Here's an idea @wind urchin - put this into chatgpt-4:
Jinger — Today at 15:20
there is not gpt plugin even if I am a plus user?
does anyone know how to do ?
Zephyrous — Today at 15:21
From 3 days ago:
We’re rolling out web browsing and Plugins to all ChatGPT Plus users over the next week! Moving from alpha to beta, they allow ChatGPT to access the internet and to use 70+ third-party plugins.
intersting
is it?
Can chatgpt-4 make me a Google slide
@sullen wren
@sullen wren did u put my prompt in?
What
can use GPT to do business?
No clue what you said amandhark
Kind of a broad question
public discord moment
the profile picture you are using is of Mai-San from bunny girl senpai ? 😭
Please put my prompt in I need it for my assignment
I have no clue I was catfishing someone
Yes it’s the girl from bunny girl senpai
😂😂I need yo@change it asap people always comment on it
Any ideas?
I don’t have web browsing plugin
Use bing
I have a formal answer. "How can I get access to plugins?
If you are a ChatGPT Plus user, you should be able to use plugins. We started granting access on May 12th, 2023 and we plan to gradually roll it out to 100% of Plus users within a week (thanks for your patience!).
To enable plugins, open ChatGPT, navigate to the settings by selecting your name in the bottom left, then choose settings. Inside the settings menu, if plugins have been rolled out to you, there should be a "Beta features" section which you can select and then enable plugins or browsing. If you do not see "Beta features", please be patient, it should appear by the May 20th, 2023."
maybe ask chatgpt-4 how long it thinks a week is
hhh
Y not
Why don’t u
May 20th, 2023. the ddl
We in the trenches out here
@young shoal are@you able to@explain your prompt a bit?
What was that syntax you used ?
The dollar symbols
wow
just linux variables
Was that variable declaration and was it within the LLM orr
Oh lol😂
I forgot@about that 💀
nah that was just me identifying variables
Wait was that in the command line?
Like you input into the command line
Or was@the just logs
this was all in text editors
I put variable names there to help me standardise the system prompt structure
I'm trying to create responsible, excellent, system prompts
https://discord.com/channels/974519864045756446/1108257759201333289 this is the best I've got
That’s good we need to establish some best practices for sure
someone explain me what are tokens, and i saw that you can lose tokens in chagpt plus, using the gpt4? is it true?
Because there’s def some unsolved common agitations when it comes to@promot engineering
I built that from examples from openai and microsoft
I'm imagining some kind of attack analogous to SQLi that will be developed in the near future, specifically designed to snipe AI models that have been integrated in systems, and to get them to produce outputs outside of what should occur
having a safety section in the system prompt is very important
And not lit at all
Nice
And uni as well?
maybe I should actually try developing those AI targeting methods
final year
Cool
Pls the prompt injection attack niche is brand new
You could pioneer it if companies mass adopt vultenezble systems like@that
At the very least people have the be exposing@themselves to some significant ddos and data leak threats with all this gpt implantation in their sites without proper security measures
Actually I’ve seen that happen already lol
A site currently taken down that had many active users of the gpt api on it
I want to learn about how to find@those@vulrenabilities
as you should
My impression is that we’re all way more vulnerable than we realize like way more
Because people are so trusting it’s crazy
Typing passwords into acquaintances devices
But in the case of chat gpt open ai does warn users to not prompt credentials
Though it doesn’t really stop them
As well as them possibly leaking out business logic
Yeah I’m pretty sure those packets aren’t completely secure in all cases lol
Data could certainly be leaked especially as third party extensions and plugins are added
I'm interested in seeing web browsing NLPs leveraged to collect creds online
I have all the beta features EXCEPT Code interpreter, are they rolling that out still?
code interpreter is still in alpha phase
Does anyone know how to do semantic search across namespaces in Pinecone? For example, I have a quote from a pdf I know I read, but I am not sure what it is. I query the index and the search is done over every namespace (pdf) until the quote is found.
Just got access to all plugins
code interpreter as well?
Oh yeah and the new one code Master too
Why does it not let me change my server name for this server?
Btw there is this Lexi Shopper plugin which apparently finds recommendations from your local Amazon store
what does the code master do
Gpt will give you a good answer about what tokens are so ask that first. In short, token count is the AI's contextual understanding.
how to use chatgpt4?
i would tell you
but sadly i just started
trying to figure out
oh nvm we just ask whatever we want
and IA handles it
u have to buy it
how to use chatgpt4?
ok thanks
I still don’t have access to plugins at all
How do I get access ?
I only have access to the web browsing
you just have to wait
Spell a lot of e-commerce platform to learn a few products Dongyang wood pendant commodity title
jo
what do you mean
how to use chatgpt4
👍

NPC
Is the web browsing option on chatgpt available to everyone?
not really
Can someone pls help me!!!!
Thanks 🙏
Why is everyone ignoring me?
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
?
Hello, I was wondering how to use the plugins
are you a plus user?
yes I am
then what ?
And he is gone
It's enabled but when I'm asking him a question that requires to search the web it says it can't answer
anyone else getting 429s for GPT4 requests?
it's currently on beta, and it seems a number of website blocks bot's web crawling
not a plugin actually
Check Dm's
I haven’t gotten access yet so sad
I wait ofr 1 hour on Chat GPT and it still says try again later.
I help AI startups find investors. Contact me for more information.
try firefox
or delete your cache
I need the cache and I have some like memory I need them to remember, sometimes they allow me someitmes they don't
When did you join the waitlist?
Like a month or two ago
Hey 👋 what happens when one goes beyond the 8k context window? is the model still able to keep track of past key information points to maintain some kind of solid context?
rip
I only have web browser and it’s the only one I didn’t want haha
this isnt related with waitlist anymore. waitlist was for alpha access
Is there a chatgpt bot, instead
Still strange that he didn't got access to it
so I don't have limited responses
I keep getting this error
"Something went wrong. If this issue persists please contact us through our help center at help.openai.com."
Is chat gpt down now?
try autogpt
My error is different: "too many requests in 1 hour try again later." and I wait for 1 hr or more it's still the same.
what is autogpt
plugins
Its working fine on mobile browser.. wth
is autogpt better than chatgpt
i cant post link here. search up
I know, sometimes it works on one browser, sometimes it doesn't
is there a chatgpt bot
thanks
the browsing works fine
that I can use
the clicks sometimes fail
sometines you mean? not every fkin time?
please tell me how I can download a chatgpt bot
wish i knew any other options at times like this..
if gpt-5 improves anymore then white collar jobs are obsolete
what does this mean?
the only obvious limit is gpu and cpu
you need to prompt accurately and specifically so it can query efficiently
how? do i need to add smth in my prompt?
To everyone asking the same?:
IMO, unless they start lifting restrictions on live web browsing, you’re not missing much.
Plug-in’s (up to 3 at a time) are the best feature but no one will be dramatically changing the world except the 🏠.
Be patient, you can add your data or play around within a week 🙊
nonsense
I'll try to restart my laptop
you dont know how to use it
ive already done insane stuff with this
you need to use your creativity
i can finish my book in 5x the time
it can literally query relevant citations on any subject i want
this would cost me hundreds of dollars if i hired a human
plugins=get rich
I know
I’m referring to ppl that don’t have web browsing or plugins.
the code interpreter is stupid op
I try to, but like the problem is that I have trouble, cuz I end up going into too many requests in 1 hr
i can send spreadsheets and csv files and extrapolate plot points and analyze the data
honestly its almost too powerful
is there a way to do so without going into too many requests in 1 hour mode, or just send 1 response per hour.
still waiting for the plugin
how do I complete a book without using chatgpt prenium
and not running into too many requests in 1 hour mode
Better prompts.
I can still go into chatgpt, but I waited very long and still have too many request in 1 hr
Practice Practice Practice
but I waited for very long and still have too many requests in 1 hr
Plus
Ik, but I don't have money for that
I don't have my own income
Somehow like if I send responses at home, I have more responses, then if I send in sch
or prob wait one day..
Invited fishing at a new lake 👋
how do you get the code interpreter model?
Is severyone getting the browsing? I only got access to the plugins
I think there is something wrong with ChatGPT
there's no possible way atm
I mean they need to keep to their word, like if they say wait 1 hr, then after. 1 hr they shld allow u to message after 1 hr, and sometimes I don't get enough to respond, I think it's broken or smth at the moment.
but how come space kangaroo could use it?
im using plus
there was a waitlist link to it. but it's not available anymore
oh
I'll wait for tmmrw
Sports car picture
The systems say that it's okay though
ss
6
HELOO
how do you do this
without running into too many responses in 1 hr
hello?
figure it out bro
you have god to answer your questions
gpt-4 is basically god
just ask it the right questions
I know I try
it's not, I have waited for like more than 1 hr, and it still says try again later.
na you just dont understand tech
if your a software developer you'd know it's god
goodluck buddy
I use chatgpt to write my books
okay? and I'm dieing
so can u tell me how u did it
I think like it's down or smth
Still the same error
I'm gonna wait till I go home. IDK What is going on
or prob I'll stop using chat gpt.
Or prob 1 hour 1 response, wait a few days then see how.
Another really cool use-case is building hundreds of flashcards in minutes
how
are you using plugins?
im thinking of writing a book every month
and then uploading it to amazon
with the links from gpt-4
you can make an index 100x speed
with credible sources
wow nice original idea
It isn't
i know
I just can't paste the yt link here
do it then
else I'll get timed out
itll still take effort to write
a 300 page book
they will ban AI generated books because itll have AI generated tags
I can't type anything rn
otherwise book prices would become meaningless
on chatgpt idk what happened
because you can generate books very fast
if you made a book once every 3 months you'd get 2000$ in passive income
Like I just tried sending a message to it a few minutes ago and it said too many requests in 1 hr
if it was good
@teal mica Like I just tried sending a message to it a few minutes ago and it said too many requests in 1 hr
im not talking to you
let's move the book discussion to #off-topic please
Yass. Please stop lifting having restrictions
Is GPT4 smarter than we think
Will we reach AGI within 20 years
What is AGI
Artificial General Intelligence
Please lift the restrictions on CHatGPT
It's annoying.
Are you Australian
No
I'm not, it's just the name of a character I like from Countryballs
I think next time I use
Be mindful of what other users in a channel might find helpful or interesting when posting. Stay on topic in order to keep conversations focused and productive.
I'm gonna do 1 response per hou
per hour
for chatgpt]
is the server like down or something
Beauty Room Freedom
I'm going to do like 1 hr alternate wait.
Cuz IDK what happened to ChatGPT, will try again tonight, hope everything returns to normal
hello anyone here to hel pme
what do you need?
See my last message, I waited for more than an hour
and still they say too many responses in 1 hr
prompts for human-like blog writing?
I'm writing fan-fic, then tdy, idk what happened but even after more than 1 hr, they say too many responses in 1 hr
what web browser are you using?
Opera
can you please try with a fresh install of Firefox?
sometimes croms is the same thing.
? why
to test if it's a browser issue or an account issue
I'm using macbook tho
ok, then try on Safari
is real web time searching availalbe to all +users?
let me try again on opera
yes, but it's still rolling out this week
how to have gpt examine different folders of files
what do you mean?
when will they finish this?
i am waiting for official info to buy +sub for testing purposes
the rollout finishes this week
Like component folder, etc for web app
ChatGPT can't examine folders of files
that means there is a Chromium browser issue going on, it will affect Chrome and Opera, I recommend you use Safari
it will be good to increase limit from 25messages/3h. at least on chat.openai.com page
Yes
Then we can send more responses, normally what i do is if it's like this I wait till after sch, then after sch it works for some stupid reason.
limit is not in one chat, this limit appears whole account. and this limit is too small. + added plugins and response quality little bit decreesed
Yes
can we just find a way to remove the limit, prob like riot then see if they remove?
<@&1007805617555124274> please help with checking if there is any problem with the server.
pplease
someone check open ai see if they are actually working.
I wait for more than 1 hr and sometimes they are killing me, why today,
I don't speak Chinese that well can use English?
it is working for sure
ik, but I've waited for more than an hour, but they still are like blocking me
what is happening
rn they can't extend limit of gtp4 its costing too lot. we need to wait for optimalizd gtp4 to get limit off
IP address?
I'm going to stop using chat gpt for 2 days..
when I'm at home,the limit is more leniant.
I'm using a secure up, like I'm using my sch ip rn
or maybe is ur school ip is on a block list
maybe openai happeded to use that blacklist
what browser you using ?
Opera
Sorry, I don't know you don't understand Chinese
I can speak but not write or read Chinese okay
that happens a lot actually when youre using gpt in a public network
It's a Chromium browser issue @worn current use Safari
but all my history and cahce I need is in opera
good good
So I have to wait for like extra long. and wait till I get home?
I wish I could like be an employer there, and remove the limit
maybe you could try using some VPN or proxy services
but idk if that would work
are you a plus member?
no fair..
I can't get plus cuz I don't have money to do so
plus give u unlimited responses?
idk if that will work, but probably
or you can wait for like 40min or so and check it out again
actually I've used 2 accounts alternatively to avoid this back in the time
I found that chatgpt has continue button
?
but after i purchased the plus service, i dont need to do that any longer
yes, totally viable
you can use 2 accounts alternatively, yet I have no idea if it is a violation of the T&C
Prob I'll wait until I get home. I tried using two accounts, but I think like ChatGPT is like blockign me for some reason
i think there's nothing wrong with that if you're paying for both accounts
I'll wait until I get home, normally that works.
I'm jealous
I'll wait until tommorow and see how.
I hate this
I think I'll try typing out the responses by myself, and like use Chat GPT like only when I'm at home, when I'm in sch I'll just create all of the lines by myself.
you could maybe try GPT4 for a month, and you will never live with out it
But I don't have money
20 bucks was like starving for maybe a day
they redirect me to ChatGPT
the wait 1 hr thing used to work, and it suddenly stoped working todya
idk why
go to chat dot forefront dot ai
idk man, I prefer the turbo for faster responses 😄 I haven't yet found a use for gpt-4 on the chat, only for integrating it into different services
I'll just wait till I get home.
Hopefully it works..
If not I'll quit Chat GPT and work on it myself, or do 1 run on ChatGPT aday
use bard ai or bing
It cannot generate sotry response, cuz I do fanfic
so I need them to like give me character reply adn response
Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!
I can't afford it currently
GPT4 now is like obviously faster than before
like the 3.5 default back then
:/subscribe
I don't have access to any plugins or browsing yet, is that normal?
Thought they'd roll it out this week
the rollout is taking place over the current week, so be patient and you will see the Beta option soon
Are plugins location-based? Saw some people say they didn't have access to certain plugins
no, it's just a progressive roll out
Oh yeah deffo! But I myself at least still haven't yet gotten to use the chat to it's full potential, full potential being the 8k context it offers 😄
Cheers, I'll check back. Hopefully I'll get it sometime soon
Beware of possible scams or fraudulent activities that you may receive through direct messages. OpenAI staff will never DM you for any transactions.
Please report any incident by sending a DM to @open crescent immediately.
Guys, I selected the web browsing before asking the question. How can I know if the answer it gave me was from the browser? Nothing changed
It pauses for a second then a small popup in the response says it is searching the web before it responds
Weird, this didn't happen, maybe I did something wrong, I'll retry
In the topbar it says "Model: Web browsing"
try a question like "what are the top news stories today" and that will trigger web browsing
has anyone faced issues with GPT4 after getting the plugin option ?
I asked something about code, I thought it would have searched on the web anyway
Gpt4 with wolfram just lags out
This
Okay now i dont even have access to normal gpt4
Everything lags out
What's the point of signing in the waiting list when you got access when everyone gets access
It's definitely beta! Also check that you have it enabled in the main screen where you select GPT-4. I actually turned it off there and when asking it to edit some JS it decided to check something on the web anyway!
are the plugins only for gpt4 or do we get plugins for gpt3.5 too?
ah alright, its kind of annoying because of the 25 messages cap
Do you also have gpt4 with plugins like to repeat the same code in the same message a couple of times?
he writes one message and then the same message again
is anyone else having problems with the api but only for gpt4?
Yesterday morning's sports news
ChatGPT
Finished browsing
Searched: “sports news May 16, 2023”
Click failed
Click failed
Click failed
Click failed
Click failed
Finished browsing
I'm sorry, but I seem to be experiencing difficulties accessing specific pages at the moment. However, based on the search results, here are some sports news highlights from May 16, 2023:
The San Antonio Spurs won the 2023 NBA Draft Lottery, earning the right to select Victor Wembanyama.
NFL Peacock announced it will exclusively carry an NFL playoff game.
The second major of the 2023 golf season has arrived.
Collin McHugh and the Atlanta Braves were scheduled to face the Rangers.
The West Finals 2023, Game 1, was held with the Nuggets playing against the Lakers.
I'm sorry I couldn't provide more detailed information. Please let me know if there's something else you're interested in or if you'd like me to try again.
Why always click failed?
My gpt-4 has problems on chat
Gpt4 every chat is something went wrong
do you have a product called NordVPN installed on your computer?
No
Should I vpn?
can you please try with a fresh install of Firefox?
robert is the api for gpt4 working for you tonight/this morning
yes it's working for me
hm. darn. now I don't know whats up
what is happening, are you getting error messages?
I don't see why the browser would be a problem
trust me, we see issues with Chromium based browsers all the time
It doesn't work with firefox private tab either
I did get one or two, the way I am using this stream for some reason I don't get errors usually, just kind of sitting there with no output.. but I tested with a different thing and I think it said something like, server overloaded
keep an eye on this in your region https://status.openai.com
I'm in Firefox, can't start any gpt4 conversations, yesterday everything was fine
No incidents on chat, I guess I'll have to make another account??
My account seems broken
can you access it with your mobile phone?
Only gpt3.5 works
what do you mean region..yeah it says there was an incident with dalle that was resolved, now everything is supposedly green
yeah, it's all working for me, my API apps are humming away normally
web browsing is working, but what about plugins?
finished 25 conversations in three hours?
are you saying you can't see Beta features now?
Gpt3.5 works on mobile gpt4 doesn't this isn't a browser issue lmao
No, I'm saying I can't create new chats, I don't need beta, I need gpt4
what happens when you try?
how do you get access to plugins
It waits, then it says something went wrong
and this is happening on all browsers and mobile?
ok then I suggest emailing support at support@openai.com and detailing the problem
Try to activate computer gui on your browser app
People, stop suggesting user-side things, it's a server issue.
Ah, so I don't know
With browser acess can chatgpt red paper from pubmed ?
did everybody get this access?
I got plugins, but then gpt4 stopped working, wish I didnt get plugins, tbh
did you see my message about emailing support@openai.com ?
gpt3.5 is writing it
the only thing I have
Has anyone with plus access been able to use code interpreter plugin?
Code Interpreter is still waitlisted, it wasn't made available with the release of plugins and web browsing
@hallow owl Ah OK, Thanks
I get browser but not plugins and you ?
I only got plugins
does anyone know if the AskYourPdf plugin allow the AI to analyse pictures?
no it doesn't, it only handles text
ok ty
The OpenAI Discord is an actively moderated server.
• Refrain from sharing inappropriate content on the server. This includes but is not limited to messages, media, or other topics of graphically violent, sexual nature, and drug-related content.
• Report all sensitive and offensive content in the feedback reporting tool in the ChatGPT web UI instead of here on Discord.
I gess its the same for web browser ?
yes
Damn. A friend of mine got premium like yesterday and have Access to web and plugins. I only got plugins and been a member for 3 months
Some of my friend buy lottery ticket 1 time and win 1000$ and max I did it was 20$
I really just meant that it really does seem to be totally random who get it
okay, I got gpt-4 back, finally, then I went to see if plugins worked, and it borked again :p, I learned my lesson
dont activate plugins, if you like having gpt-4
hen will plugins be available in France?
must be nice 😭
been a plus member since the dawn of time, got none of those
its better that way, as lons as your gpt4 doesnt break
plugin access breaks gpt-4?
I guess it depends on your server
I think they group users by server or something
by region
so thats why there are people having totally different versions of chatgpt
it is really opaque, worse than instagram tbh
Also depend what plug in people install
hmm ic ic
Some people don't know what they doing
I live in France, and when I open the settings window, I don't have Beta features available. Therefore, I am a GPTplus user.
Like I said #announcements
anyone could help with my account crash?