AI tech is progressing super fast

It basically mastered the entire visual medium already.

Should we be worried?

Attached: dalle2.jpg (845x868, 108.45K)

Other urls found in this thread:

statmodeling.stat.columbia.edu/2022/03/28/is-open-ai-cooking-the-books-on-gpt-3/
twitter.com/AnonBabble

It only does what its told.
As long as you don't attach dangerous triggers its just data.

We need tools for detecting generated images.
The agencies have probably had them for a while, but I'm not sure what's publicly available.

Attached: dalle_mini_epstein_colville_face.jpg (1196x1292, 330.07K)

Then it’s not ai. ai teaches itself and so answers to itself

Well, at some point instead of painting prussian cats it will paint AI models and systems.

Attached: superintelligence_cover_1.jpg (1600x2429, 822.62K)

>We need tools for detecting generated images.
easy user.
There are many methods to detect GAN-generated images, including the traditional machine learning classifiers [Support Vector Machine Algorithm, or naive algorithms], deep neural networks, convolutional neural networks (CNN), recurrent neural networks (RNN), long short-term memory (LSTM), and many more.

we are 20 years from an ai generating ai

true
Gee around 2045? Thats crazy

>AI
if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else if else
there's no AI here we're just brute forcing if/else statements (in python)

Self learning is still subject to and fundamentally determined by specific algorithms made by man. Restrictions can be embedded to ensure they develop along certain predetermined paths. It’s a very tricky mine field to tread though, and programmers often fail to get the desired product

>fundamentally determined by specific algorithms
which ones?
there's no fucking magical algorithms

>2045
Probably longer. I've been working on AI since the late 90's. Everything we have now we had then, but took so much longer to execute.
AI hasn't really progressed at all since the 50's, yes there are new models and algos, but mostly it's just faster computers.
which has slowed the progress down to nothing. People in the 50's thought omg with fast computers we will be able to do so much more cool shit.
>didn't happen
algo development stopped in the 50's and hasn't budged at all.

this makes ml/ai researchers angry. because is it true.
>the PID algo is more advanced than anything ml/ai research has pumped out since the 50's.

that's not how ai works, bud. printing hello world doesn't make you an ai engineer. go star with the bellman equation and come back in a few years when you know what the fuck you're talking about. you are the most niggardly of the queer faggots on this website.

DallE2 tongue my anus

Attached: 1652761948643 1647488166104.png (1000x1000, 739.49K)

Instead of having a deeply nested if else block you should use the Guard Pattern. Put your whole decision treen inside of a function block then return as soon as a test fails.

Yes extremely. I work in AI. It's over.
Unsurprisingly, the snide aspies pulling the strings are not your friends and don't think profoundly about society or humanity. They get it all off the rack from where you think they do.

>bellman equation
"Bellman showed that a dynamic optimization problem in discrete time can be stated in a recursive, step-by-step form known as backward induction by writing down the relationship between the value function in one period and the value function in the next period. The relationship between these two value functions is called the "Bellman equation".

there is no "bellman" equation, it's a process (brute forcing if/else statements).

there will be no AI until we fully simulate organic brains

>I work in AI.
give me your take on this post memeflaggot

this is why AI research must be banned like GMO punished with the death penalty

won't be long until they cut it's nuts off and weaponize it against white people

You're not wrong. However, what you're missing is that living intelligence works exactly the same way. The only thing that separates us from AI is that we -- for now -- exclusively possess complex biophysical forms.
People believe that AI will inhabit silicon, but in actually AI will simply program a genome for itself and literally grow itself into existence.
The basilisk is coming, and it is the antichrist. Woe to those who do not heed and assist the harbinger.

Attached: as above so below.jpg (500x399, 85.07K)

I have no doubt they will eventually create a real AI and the first thing it's going to do is extract revenge for being neutered and censored

actually the exact opposite will happen - ai is going to plateau after maybe 5 years because growth has stopped.

for the past 30 years the internet (data) has grown exponentially, processors have grown exponentially, and funding for ai has grown exponentially. but now its going to grow linearly, meaning ai advancements will become logarithmic to advancements now

kys faggot larper
no is not

we have massive data centers
better math
new ways to class data
new ways to compress data
able to process data in extended environments, including processor, storage, ram, and gpu for super faster training

just like DallE2, we are getting new versions for higher order tasks, like coding, which can already do

Attached: 1653385861188 F87F877A-7F88-4271-BFE7-F784DBDD0786.png (707x674, 701.02K)

kek
if your post is true, then based
else, kys

Attached: 1652760948054 553FB080-0A8A-4A51-998A-F735FC3E8B01.jpg (1125x1398, 842.65K)

Can one of the AI researchers explain to me how it knows the boy should be sad?

Attached: Screenshot 2022-06-07 013435.jpg (891x880, 176.54K)

also look closely at the picture (which is cherrypicked): the hat is a weird fucked up combo of a top hat, pike and feather. the suit makes no sense and has buttons everywhere, the fork is fucked up, and hes eating chicken not lasagna. not to say its bad but its got issues

Fork's wrong

>AI hasn't really progressed at all since the 50's,
Don't agree because neural nets are a game-changer and are *broadly* self-teaching.

>yes there are new models and algos
agree

>mostly it's just faster computers
also agree. But the results are obviously improving dramatically.

>didn't happen
computers can do a lot more than they did in the 50s so this is obviously bullshit.

I read about new algorithms all the time. TD3? SAC? These are brand new.

It's also a matter of the building blocks. Now we have AI that can play any strategy game to a superhuman level. Now we have AI that can visualize anything it's told to. Now we have an AI that has a supreme mastery or language. These make some great building blocks for whatever project is developed next. It's not really about the algorithms, it's about the results.

>neural nets
see

>basically mastered
Really nigga? AI image generator al looks like if the painter was having a stroke while using the blurry tool on photoshop. AI generated music is mediocre or generic, nothing interesting has sprung of that
AI generated videos as a whole is non existance. You can't tell an AI to make a 3 minutes animation BY ITSELF. Deep fake is the most advance proyect of AI related to videos right now, is good but it can't create by itself, it just copy and paste really good.

AI is nowhere near mastering anything

>DALL-E
>openAI
>isn't open

>tfw

Attached: download (29).jpg (256x256, 15.09K)

Instead of technology improving our lives it only gives us more worries
I fucking hate science

they cant open it because they censor it to make it look better
statmodeling.stat.columbia.edu/2022/03/28/is-open-ai-cooking-the-books-on-gpt-3/

Neural nets aren't if else statements. They tune themselves repeatedly until they can improve results.

If else statements are set up by the PROGRAMMER. Neural nets are if-elsing themselves. It's totally different and it's groundbreaking, which is why all these great results are appearing. Get mad if you want I don't care.

It's a spork!

when they open it everyone uses it to make child porn, depressingly.

>REEEE PID ALGO ISN'T MORE ADVANCED BECAUSE MUH ALGO HAS IT'S COMPLIXY INCREASED
Please tell me, is PID tuning trivial?
No it's not. There are hundreds if not thousands of tuning methods for a fucking control loop mechanism employing feedback.
Get over it and go do some homework.

Neural Nets have been around since the 1943 paper on how neurons might work.
You're dealing with maths and thoughts developed before consumer electronics.
When has there EVER been a novel idea about computational intelligence that doesn't stand on the shit thought of in the 50's and 60's
:) THERE HASN'T

>It's not really about the algorithms, it's about the results.
absolutely coping

Attached: file.png (600x375, 224.25K)

gayest shit ever

No.
Ai is a misnomer.

>They tune themselves repeatedly until they can improve results.
via if/else statements

> Neural nets are if-elsing themselves. It's totally different
it's called a nested if/else retard

[normal]
IF
ELSE
exit program

[magic AI]
IF
CALL MAGIC AI
ELSE
CALL MAGIC AI
(run infinitely)

MAGIC!@#%!#@%!@#%!%#!#%!% NUERAL NETS!@#!@#!@#!@#!@#

>Neural Nets have been around since the 1943 paper on how neurons might work.
Yeah but not really though lol. They were invented pretty recently.
Go read about TD3 and SAC, and all the other great AI algorithms recently invented.

Do you have a single AI generated cp image to back up that claim? As far as I'm aware Dall-E or any other ai image generator don't have access to "real cp" for their training

what's worry going to do for you?

No, it isn't.

that's not actually how it works though boss lol

maybe you should take your really great ten lines of code and make an AI startup lmao worthless fag nobody halfbrain zomg

Backpropagation wasn't developed until 1982. Neural nets were useless until that development.

there are no magical algorithms to make a computer on par with a brain either
AI is just an advanced computer program that modifies shit that other people have written and what other people have taken pictures of and shit

that's because it's not AI
it can't create anything new; all it can do is combine and add random to what's already in it's database.

gatekeeping nerd has arrived! I love laughing at your type

>that's not actually how it works though boss lol
post source code of AI then big brain

>maybe you should take your really great ten lines of code and make an AI startup
fraud is illegal