Is there any meme worse than artificial neural network...

>Your brain took millions of years to get to this point
Bullshit, video games haven't been around for millions of years

Not only that, but you'll only see them attempting level 1 in video games because it's literally the tutorial level with the simplest concepts, took it 24 hours to learn to press right as long as there's a platform underneath it, you think it can ever solve the maze or even simple jumping puzzles? heck it can't even differentiate between a long jump and a short jump.

MarI/O could be improved a lot. The information the network gets about the level is too basic and even wrong (powerups are seen as harmful, for instance).

>took it 24 hours to learn to press right as long as there's a platform underneath it
Even niggers are smarter than that

>Is there any meme worse than artificial neural network?
Natural neural network.

Not all neural networks are created equal.

There are neural networks that can play Atari games after watching someone play them for 30 seconds. There are neural networks that can change any style into a different style, like take a Van Gaugh painting and turn it into a da Vinci painting. There are neural networks that can change a person's apparent gender in photos.

Just because some hack made a shitty neural network for something that sucks doesn't mean neural networks are stupid, it means the hack is stupid.

Found neural network racist.

try manually programming a 90%+ accuracy video object detection and classification for 10+ categories without a CNN

and be sure to collect your nobel if you can do it

a 6 year old kid is better at learning a language than you fucking idiot

>a modern neural net works similarly to your own brain
Please do not spread this disinformation. We have no working model for how the human brain works, so you can't say this.

Which is, sort of, an amazing testament to the general problem solving ability of the human brain.

Great, but that's not the topic, idiot

So neural networks are shit at learning unless you program them to learn exactly the shit you want them to learn.Nice intelligence you got there faggot.

Attached: 1534998889912.jpg (306x306, 14K)

There's also a neural network that will create a 2d waifu for you, yet anything produced by a a neural network will fall apart upon a closer inspection, it takes a lot of resources and care to make something that will produce a highly specific thing when you really can do a better work handcrafting a program to do the same task, it is only a problem for image/video processing because it has too much data to take account for, even then it's not impossible.
Kinda sad how people are solving problems by throwing more money/resources/time at it, that's exactly why technology became stale, we get more powerful machines only to do things sub-optimally to nullify any kind of advancement.

Yeah you have to know how to train them you dumb faggot, they're not magic.

people talk about them like they're magic

Isn't there a way to play any SMB where the graphics are not scaled to fit the original screen size, and the graphics are native 1:1 so you can see like 6x of the map on screen at a time?

because even the people at the top of the industry don't have the slightest fucking clue what they're doing

there are 2 type of learning, supervised and unsupervised (i.e. re-enforcement learning). So far supervised is the only method that gives real results

Probably took you around 2 years to learn to walk so given the same amount of time this Chad Neural Network would be shitting all over your life and enslaving you to its bidding.

Neural networks do not start learning like a child does even when given the same amount of time as a child had.

>Comparing something that's basically brute-forced through billions of years to the state that it is today with something that's intelligently designed.

It works on the same basic component: the neuron.
It's just that human neurons are way more complex.

You're looking for a remake like Mari0 where you can edit the viewport.

so it works on the same basic component except that thing that's the same is actually completely different?

Neural networks are networks of neurons
The brain is a network of neurons
The brain is a neural network

Why is Jow Forums so fucking stupid? Jeezes fucking Christ the ignorance is painfully to even read in this dumpster tier trash thread.

It is not about applying the logic to one shitty example you dumb shits. Is trying to find a loose enough logic that when given another similar challenge it will also be able to complete it.

AI is amazing because it is a type of specialized general intelligence. No human is needed anymore to make a program for every specific situation .

You could make an AI to create websites for example.

> Hi I just too my first machine learning course.

modern AI is literally in pajeet stages

the "neurons" in neural networks are not the same thing as the actual neurons in the human brain they are named after

Are you fucking serious nigger?

Yeah but they principally work the same.
You have inputs to the neurons and an output. Signals come in, output sometimes activates.

Prove me wrong.

>le robot apocalypse is coming soon xD

The "net" in neural networks is a collection of filters applying mathematical functions on the input. Saying that neural nets are like the human brain is like saying linear algebra is neuroscience

Ok?

>2018
>After years, the Google AI still needs you to learn how to make the difference between a car and a fucking store front

People's reaction to NNs is always "it's just statistics it's just tree search + reinforcement, it's just..."

Brainlets lack metacognition. Next time you learn a new task, observe yourself, and you will see the way you learn is not that different.

Look at little children, they're basically brute-forcing everything. How do you think they learn to walk? They perform random movements enough time until it works.

It's human narcissism to think we have some magic special sauce that no machine can ever imitate. MUH SOUL. You'll be taught a lesson of a lifetime once we get to AGI in 10-20 years.

Luddites will be BTFO, as always.

>Yeah but they principally work the same.
no they dont
all that's the same is it's a giant connective web

It was very clever to call them neural networks. If they were called something like connected logistic regressors noone in the general public would care.

>"it's just statistics it's just tree search + reinforcement, it's just..."
That's because other peoples reactions to NNs is unironically "we've created intelligence, singularity soon guys"
It's human narcissism to assume that because they've made something that acts a bit like a brain that they've created intelligence - we don't even know how brains work

>they're basically brute-forcing everything
No, mirror neurons.

Only laymen think like this, actual researchers know it's just a hype.

The brain is a complex conglomerate of many different types of intelligences, you brainlet.

We have discovered only ONE aspect of it. one function. One module, if you will. That which applies to recognizing patters via images and sound.

We need more breakthroughs but DeepMind already is on the way to finalize them (they got short term memory now, and some transfer learning)...

We're still very early when it comes to natural language processing AND symbolic/conceptual thinking and abstraction. That will be the toughest nut to crack. But based Kurzweil is working on it at Google with infinite billions of $ and talent at (((his))) disposal.

>all that's the same is it's a giant connective web
Yeah exactly.
Of neurons.

The last time they posed the question to top researches I believe the average answer was we are 20 to 30 years away from AGI.

>We have discovered only ONE aspect of it. one function. One module, if you will.
No, we've invented (not discovered) something analogous to a function of the brain
there's a huge difference, assuming anything otherwise is arrogance

>Is there any meme worse than the greatest technological breakthrough of the decade

nice meme, kiddo

What's AGI?

>greatest technological breakthrough of the decade
You misspelled the 90's

NNs weren't even invented in this decade.

OK, I admit you are right. I was simplyifing things.
Anyway the point is, we can imitate one of the functions. We don't need to re-make the same thing, only find something that acts and functions similarly. I know the brain doesn't use backpropagation.

Biological neuron is not comperable to a NN "neuron". They just named it after a neuron because the graph structure looks similiar. It's like skeuomorphism. A "window" in a computer program is not a real window like you have in your house, it's just named that because it's similiar and for lack of any other term

>invention is the same as implementation

>invention
Actually the invention is even older than that

Except there isn't a lack of terms for a node in a graph, you can just call it a fucking node. I'm sure there's tons of terms you could use.
They called it a neuron for a reason. They wouldn't just make an arbitrary reference to the brain for no reason.

>you can just call it a fucking node
but they do, lol
a node in a neural network is no more a neuron than a window in windows is something you open to let some air in
well, maybe a bit closer

Fuck you. Deepfaking as a hobby has revolutionised my fapping habits

>Muh AI
>Muh singularity within 15 years
Meanwhile n the real world the AI can't properly translate a single page of text without fucking up, fails to auto-correct properly, crashes your car in a giant truck or runs over somebody crossing the street, can't extract the edges of a simple drawing, live translation is a joke, live captioning is a joke, vocal diction is still a pipe dream, ...

Based user standing up for his hobby.

...but it had to take millions of years of brain development to develop the AI. The brain wins.

>b-but muh alpha go

>AI can't properly translate a single page of text without fucking up
Most humans can't either.

Bullshit, my brain took 25 years to get to this point.

In the 90s I used to tell my elementary school mates the Internet will change everything, I was so excited as a kid... I got it when I was 10 years old. I was obsessed with it. People mocked me for it... now all those people spend their days on Facebook.

Back then some visionaries said the Internet will change everything and people like you pointed to the shortcomings and imperfections of the technology. Krugman even said the Internet will be "as revolutionary as the fax machine".

People like you are always left behind. The eternal beta, the eternal NPC, the eternal non-visonary wagecuck. If you were a visionary, you'd not be working right now. Those of us who bought btc in 2011 don't have to work.

AI will absolutely revolutionize everything in the next 10-15 years. The world will be unrecognizable. Only a myopic brainlet points to the current state of applied narrow AI and doesn't instead look at the exponential trend of its development and evolution.

Except imitating is superficial, imitating the obvious doesn't mean remaking the whole behavior, a simple hidden behavior can have domino effect on the whole thing, it's the same with scientific theories, they can describe one simple thing almost correctly but then a new one pops up and changes everything we know.

>AI will absolutely revolutionize everything in the next 10-15 years
You'd have fit right in with the "visionaries" of the 70's.
Congratulations.

>Because the visionairies in the 70s were wrong about the dates it means the visionaries of today must be equally wrong

That's not a valid logical inference

The difference is you were talking up the internet when it already existed, but there is nothing remotely close to artificial general intelligence yet so you're merely speculating on something even experts in the field don't agree on.

On top of the recorded collective knowledge of our species, which is literally the only thing that separates us from the stone age at this point.

underrated comment, the only person in this thread who knows what they're talking about

>Back then some visionaries said the Internet will change everything and people like you pointed to the shortcomings and imperfections of the technology
I dont think so
Nobody said "this internet shit will never catch on"
Nobody was hyping it up, and people who didn't use it were largely indifferent to it
Meme technologies on the other hand have everyone hyping it up while the reality falls way short of what they claim
AI isn't developing. What developed was parallel computing and big data, the technology and ideas behind AI are the same as they've always been

Considering the experts can't even agree on the when, that VERY optimistic 10 to 15 years seems very unlikely.
Reminds me of the quantum cucks claiming quantum supremacy would be achieved in 2017.

>Back then some visionaries said the Internet will change everything and people like you pointed to the shortcomings and imperfections of the technology
The internet turned into utter garbage.

I don't get it, are you upset because AI will be the future or because you don't know when it will be useful?

DeepMind is developing fragments of what will eventually lead to AGI. Transfer learning, short-term memory, IMPALA, PopArt, check their research after AlphaZero. - It's not merely hypothetical. Some level of generalization is possible; ie. transferring learning from one simple task to another.

I just told you that Krugman said Internet will not be a big thing. Look it up. Why do you deny a stated fact? AI is developing it's just that you're too much of a brainlet to follow the research. One of the great breakthroughs in this decade was GANs.

PhD in CS who does NN and ML research. This thread is retarded.

Mar-IO is a genetic algorithm, it's not actually a neural network.

Yes, Machine Learning requires enormous amounts of data to properly run, but the entire point is that computers can eat that data in orders of magnitude less time than a human can process a much smaller set of data.

Back then some visionaries dreamed about flying cares and living in space
Keep dreaming user, humans can only be efficient at things they can already do, we became efficient at building, farming, traveling, communicating, killing each other... our lives is more or less the same, only faster.
People who think we can create something intelligent even of the level of an animal are living in a pipe dream, we can only automate, what we could've automated mechanically is now automated electronically, and that's it.
It's 2018 and we can't even understand what the fuck gravity is, how many neurons are needed to makeup intelligence, nor how do we even think at all.

>Yes, Machine Learning requires enormous amounts of data to properly run, but the entire point is that computers can eat that data in orders of magnitude less time than a human can process a much smaller set of data.
But there's also the possibility of a system to GENERATE the data, such as AlphaZero did with self-play. Sure, that's a simple situation with Go, not many parameters go into it, but we don't know yet where this line can lead us.

Because we don't know when it will be useful.
I would fucking love to have an actual general purpose AI with me right now. And a real one, not some Alexa garbage whose sole purpose is to spy on me

And do you see normies caring about AI or saying it will be "next big thing"?
Admit it, most people are apathetic
Also, I'd say the internet parallel is more accurate in the sense that it will develop in a very different way than people think it will.
Internet was never thought to develop into the surfing cesspool of social media it is now, but rather primarily a database of human knowledge.

Why does that anger you?

>People who think we can create something intelligent
Sometimes I wonder if AI more intelligent than Humans isn't a ridiculous.
I mean, you can only create something as intelligent as you are, creating something more intelligent would require you to be more intelligent too

>I just told you that Krugman said Internet will not be a big thing
Who cares what one person says?
"Generative adversarial network" is not a great breakthrough. It's a low hanging fruit. All AI research at this point is low hanging fruit - implementing things we've known about for decades but didn't have the computing power to implement. None of them point to exponential growth in anything

This, take a look at the dota 2 openAI for a good example of a proper videogame AI.

>And do you see normies caring about AI or saying it will be "next big thing"?
Yes
Everyone is so fucking delusional about AI it's unreal
Nobody knows how it works but everyone knows it's the future, and everyone is attributing human qualities to machines like they did in the fucking industrial revolution where crude factory robots were putting them out of jobs

Well, agreed, most likely we will develop technology to improve our own intelligence first before we can develop a GAI.
With a bit more research into genetic engineering and neuroscience this is quite possible, propably more than GAI.

Because too much people are shaking things in front of the masses and telling them how bright their future is.
And I fear it may turn into a parody of what it should have been.

Biological computing is the future.
Silicone AI is shit.
Prove me wrong.

General Artificial Intelligence, also known as Strong AI. It's the term for a type of AI that can think and rationalize like a human to solve multiple problems across a variety of domains rather than just being proficient at a single task. For example, MarI/O is not an AGI because it's only good at playing Mario

>"Generative adversarial network" is not a great breakthrough. It's a low hanging fruit. All AI research at this point is low hanging fruit - implementing things we've known about for decades but didn't have the computing power to implement. None of them point to exponential growth in anything
Moving the goalposts aye?