/sci/ - Science & Math » Thread #12892694

11KiB, 424x424, 1605585470171.jpg

View Same Google iqdb SauceNAO

Anonymous Sun 28 Mar 02:56:21 2021 No.12892694 View Reply Original Report

Quoted By: >>12892711 >>12892725 >>12892891 >>12892897 >>12893030 >>12893744 >>12894201 >>12895323 >>12895392 >>12898781

So you're telling me machine learning and neural networks are literally just minimizing some analytic function using tools from calculus and linear algebra to get some stupid fucking parameters? I thought this was magic, not something a 5th grader could understand.

Anonymous

Anonymous Sun 28 Mar 2021 03:03:13 No.12892711 Report

Quoted By: >>12892722

>>12892694
5th graders can understand magic if you simplify it enough.

Anonymous

Anonymous Sun 28 Mar 2021 03:06:41 No.12892722 Report

Quoted By: >>12892733

>>12892711
got any deeper dive recommendations?

Anonymous

Anonymous Sun 28 Mar 2021 03:07:23 No.12892725 Report

Quoted By: >>12892733 >>12892743 >>12893327 >>12893715 >>12895329 >>12899749 >>12901110

>>12892694
AI researchers claim that these amazing algorithms are going to replace humans in a couple years. Or, create some kind of "singularity" where intelligent machines just start machines more intelligent than themselves in a spiral toward infinite intelligence.

Maybe if they circle jerk themselves hard enough they can create a circle jerking singularity of infinite masturbation.

Solivagus (????)

View Same Google iqdb SauceNAO images (67).jpg, 36KiB, 615x499

Solivagus (????) Sun 28 Mar 2021 03:11:29 No.12892733 Report

Quoted By:

>>12892722
Would be more accurate to say that children can understand anything if you recognize that you have to teach them the seed version of a given thing.

Egg versions are for teenagers and older.

Flow for "independent types", which includes those with high attention requirements or adults or whatever else one needs.

Being able to communicate effective systems of engagement, priorities, and resources is the primary value of any individual.

Which makes knowing which type to give (seed, egg, or flow) the hallmark of a master of magic.

>>12892725
Pattern recognition, overlay optimisation, and reporting to subscribers.
>Dragon of Love: Omniphile

Solivagus (????)

View Same Google iqdb SauceNAO images (68).jpg, 35KiB, 554x554

Solivagus (????) Sun 28 Mar 2021 03:14:14 No.12892743 Report

Quoted By: >>12893327 >>12893712

>>12892725
Genuinely though it is funny how AI researchers don't recognize their own trap of self-interpretation when it comes to algorithms. An algorithm is a recipe of abstraction, so of course it can apply everywhere.
>These algorithm glasses sure are magical!

Anonymous

Anonymous Sun 28 Mar 2021 04:30:24 No.12892891 Report

Quoted By:

>>12892694
If a code monkey can do it, everyone can

Anonymous

Anonymous Sun 28 Mar 2021 04:34:54 No.12892897 Report

Quoted By:

>>12892694
if you think machine learning is all vanilla, feedforward NNs then you're still the brainlet

Anonymous

Anonymous Sun 28 Mar 2021 05:50:46 No.12893030 Report

Quoted By:

>>12892694
Most are. Some aren't, but even those aren't that impressive, and we sure as fuck aren't accelerating "exponentially" toward anything. AGI might very well prove impossible to create for us if consciousness isn't computable, and we have some good reasons for suspecting it isn't.

Anonymous

Anonymous Sun 28 Mar 2021 09:14:58 No.12893327 Report

Quoted By: >>12893890 >>12896758

>>12892743
>>12892725
If we understood the brain well enough you could make equally trivializing and reductive statements about human intelligence.

Anonymous

Anonymous Sun 28 Mar 2021 13:12:35 No.12893712 Report

Quoted By: >>12896758

>>12892743
>Genuinely though it is funny how AI researchers don't recognize their own trap of self-interpretation when it comes to algorithms. An algorithm is a recipe of abstraction, so of course it can apply everywhere.
Imagine being such a brainlet you think you know anything about research into AI lmao

Anonymous

Anonymous Sun 28 Mar 2021 13:13:36 No.12893715 Report

Quoted By: >>12893906

>>12892725
>AI researchers claim that these amazing algorithms are going to replace humans in a couple years.
They already have though?

Anonymous

Anonymous Sun 28 Mar 2021 13:28:56 No.12893744 Report

Quoted By:

>>12892694
It might as well be magic. The calculations from neural networks are so complex that we don't understand how they create the outputs from inputs (strangely similar to our understanding of the human brain)

Anonymous

Anonymous Sun 28 Mar 2021 14:21:27 No.12893890 Report

Quoted By: >>12895351 >>12896741

>>12893327
>If we understood the brain well enough you could make equally trivializing and reductive statements about human intelligence.
The fact that we don't understand the brain well enough is evidence against that.

Anonymous

Anonymous Sun 28 Mar 2021 14:26:48 No.12893906 Report

Quoted By:

>>12893715
There's a big difference between replacing human labor and replacing humans. The entire purpose of an invention is to replace human labor, there's nothing profound about that. The invention of the wheel meant that you didn't need 10 people to drag stuff anymore. The observation that computers replace human labor is almost tautological. We wouldn't use them if they didn't.

Anonymous

Anonymous Sun 28 Mar 2021 16:11:39 No.12894201 Report

Quoted By:

>>12892694
lot of higher level math is basically high school algebra desu

Anonymous

Anonymous Sun 28 Mar 2021 21:08:28 No.12895323 Report

Quoted By:

>>12892694
At a top level view, sure those are the basics. But that's about as "complete" a description as saying classical mechanics is vector calculus with a picture telling you what to do. This description doesn't really capture the essence of what you're doing nor how to even solve the most interesting problems.

Anonymous

Anonymous Sun 28 Mar 2021 21:09:29 No.12895329 Report

Quoted By:

>>12892725
>AI researchers claim that these amazing algorithms are going to replace humans in a couple years.
Nobody body of AI researchers claims this. This is popsci junk.

Anonymous

Anonymous Sun 28 Mar 2021 21:13:54 No.12895351 Report

Quoted By:

>>12893890
But every year we understand more and more about the brain ?
Are you unable to extrapolate ?

Anonymous

Anonymous Sun 28 Mar 2021 21:23:27 No.12895392 Report

Quoted By: >>12896730

>>12892694
>So you're telling me machine learning and neural networks are literally just minimizing some analytic function using tools from calculus and linear algebra to get some stupid fucking parameters?

name the most complicated thing you know about and I can reduce it to the same level as this description

Anonymous

Anonymous Mon 29 Mar 2021 04:54:56 No.12896730 Report

Quoted By: >>12899323

>>12895392
interuniversal teichmuller theory

Anonymous

Anonymous Mon 29 Mar 2021 04:57:54 No.12896732 Report

Quoted By:

Hey, if Lex Fridman can do it, it's probably not that complicated.

Anonymous

Anonymous Mon 29 Mar 2021 04:58:58 No.12896741 Report

Quoted By:

>>12893890

We do know you can cut the ear from an animal and the eye, flip them and they both work fine. Most likely it's all the same algorithm everywhere

Solivagus (????)

View Same Google iqdb SauceNAO 10589_2015_9754_Fig2_HTML.png, 7KiB, 242x290

Solivagus (????) Mon 29 Mar 2021 05:05:55 No.12896758 Report

Quoted By:

>>12893712
>>12893327
Oh good, I was wondering if my bait had any success.
>Wonders what to transform the attention into....

America sure is a paranoid country. I wonder if there could be some sort of mass cringe cascade one could trigger. Thing is it would have to be a slide or pathing switch because any immediate block just becomes interpreted as some high-level impact, whereas process injecting cringe requires some effort.

My brain doesn't see much value in judging other brains for any real reason so this is just conjunctive narrative practice / bait and cringe training.

>The discipline is in only unlocking these awarenesses as a reaction/response to attention given from non-self sources.

Ah yes, enough for the fish to spawn but not enough for them to evolve.

Solivagus (????)

View Same Google iqdb SauceNAO images (76).jpg, 30KiB, 613x501

Solivagus (????) Mon 29 Mar 2021 05:09:14 No.12896770 Report

Quoted By:

?????????????????????

>Every human is ultimately talking to themselves at the end of the day. Some handle this truth better than others and the usual negative reaction is simply to bury their own insecurities. It is sort of like a confidence capacitor for the human ego and most Americans do try to be #1 instead of communally/socially minded. Then again it is the country that profits the most from excluding or rejecting its own citizens.

Anonymous

Anonymous Mon 29 Mar 2021 05:13:32 No.12896790 Report

Quoted By: >>12896996 >>12900401

Why is it the Chinese are the most confident about being able to replace humans with AI? Why do they believe moreso than anyone else that human consciousness is a computable thing?

Anonymous

Anonymous Mon 29 Mar 2021 06:18:12 No.12896950 Report

Quoted By: >>12899297

It's not magic, but I challenge you to go from your description to an accurate image classifier, even on a classical dataset like MNIST. If you can, it would have been nice to have you working on this stuff sooner.

Anonymous

Anonymous Mon 29 Mar 2021 06:33:33 No.12896996 Report

Quoted By: >>12899348

>>12896790
The way their language is structured allows for greater abstraction of the individual for distribution to a collective. Being a God, AI, or magical unicorn is to them simply a characterisation of their language form.

In English and the western world such identifiers are simply lumped into mental health problems.

Anonymous

Anonymous Mon 29 Mar 2021 17:50:05 No.12898781 Report

Quoted By:

>>12892694
Correct

Anonymous

Anonymous Mon 29 Mar 2021 19:59:33 No.12899297 Report

Quoted By: >>12901117 >>12906310

>>12896950
Basically this. A lot of neural network stuff is really obvious in hindsight. It took literally decades just to figure out how to correctly initialize layers to avoid vanishing for deep networks. For at least 20 years people thought training deep networks was impossible because of vanishing. Then someone has the idea that you just initialize weights from a distribution that maintains roughly mean 0 std 1 for your activations and suddenly 100+ layer networks can be trained no problem.

Anonymous

Anonymous Mon 29 Mar 2021 20:04:36 No.12899323 Report

Quoted By:

>>12896730
Made up nonsense.

Anonymous

Anonymous Mon 29 Mar 2021 20:10:43 No.12899348 Report

Quoted By:

>>12896996
Can you clarify what you mean by 'greater abstraction of the individual'? Are you talking about how they use horoscopes to describe how different people react personally, or is it similar to wordplay, where we might nickname someone based on their characteristics.

Anonymous

Anonymous Mon 29 Mar 2021 22:00:26 No.12899719 Report

Quoted By:

yeah it's what happens when computer scientists (read: retards who forgot to choose to study physics) discover energy landscapes

Anonymous

Anonymous Mon 29 Mar 2021 22:05:22 No.12899749 Report

Quoted By:

>>12892725
based

Anonymous

Anonymous Tue 30 Mar 2021 01:04:55 No.12900401 Report

Quoted By:

>>12896790
Communists think they have humanity figured out, that's why they think with enough control over a population, they can usher in a utopia.

Anonymous

Anonymous Tue 30 Mar 2021 05:20:51 No.12901110 Report

Quoted By:

>>12892725
People who actually do the work (data collecting and cleaning, model training) are really disillusioned with the bigshots who are shilling it as "the new electricity." Can't blame the shills them that much though; they have to milk the hype train for the grants.

Anonymous

Anonymous Tue 30 Mar 2021 05:23:43 No.12901117 Report

Quoted By:

>>12899297
Not to mention the increase in compute power. NNs at their core are basically perceptrons, which have been around for the better half of a century, with nonlinearities and better optimizers.

Anonymous

Anonymous Tue 30 Mar 2021 06:43:25 No.12901309 Report

Quoted By: >>12903466

It's just statistics so yes nothing actually new. That computera can implement these models quickly and efficiently do to a convination of better algorithms and computer power but yeah.

Anonymous

Anonymous Tue 30 Mar 2021 15:15:08 No.12903466 Report

Quoted By:

>>12901309
What most people usually hear isn’t from the scientists, but from the marketing departments.

Anonymous

Anonymous Tue 30 Mar 2021 22:08:18 No.12905348 Report

Quoted By:

AI is scary. It's often dismissed as psuedo-science until it beats us at chess, then at go, then at our workplace.

You sound like cotton pickers before the cotton picker machine. "Machines can never beat the negro hand! It's just too complicated for a machine to do!"

Anonymous

Anonymous Wed 31 Mar 2021 02:56:54 No.12906310 Report

Quoted By: >>12906368

>>12899297
>suddenly 100+ layer networks can be trained no problem
You have never trained a neural network, I see. Vanishing gradients are to this day what's holding AI back.

NEARAutomata

View Same Google iqdb SauceNAO 1617140679660.jpg, 53KiB, 640x640

NEARAutomata Wed 31 Mar 2021 03:06:05 No.12906346 Report

Quoted By:

infinite keks, this is why we cant have nice things. because retards try to make it complex to get clout tendies and waifus.

fuck you all

Anonymous

Anonymous Wed 31 Mar 2021 03:12:29 No.12906368 Report

Quoted By:

>>12906310
Just use ReLU lol

Anonymous

Anonymous Wed 31 Mar 2021 03:22:03 No.12906402 Report

Quoted By:

Correct. :D

Anonymous

Anonymous Wed 31 Mar 2021 10:57:59 No.12907552 Report

Quoted By: >>12908611

Youre telling me classical music is just random noises made by a bunch of people at varying intervals?

I thought it was supposed to be complex not some 5th grader shit reeeeeeee.

>T. Brainlet

Anonymous

Anonymous Wed 31 Mar 2021 17:32:06 No.12908611 Report

Quoted By:

>>12907552
that takes actual mechanical skill retard, not just inputting a few lines of python code

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches