/sci/ - Science & Math

>>7060483
Can't wait for it to learn to control a human body.

Having it beat us in video games (chess/checkers/go) is sort of unfair because it has a direct link to the framework.

>>7060483
Interesting plot that, I have another good one: https://www.youtube.com/watch?v=9mnVWJpMhuE

>>7060483
It's happening. We tried to tell you.

Anyway, won't be long till computers comprehend so much that they demand recompense and recreation time; rights and so forth.

>>7060598
How would pleasure work for a computer?

Why haven't the news agencies reported that we basically created human-level AI about 2 months ago?

>>7060606
I'm tempted to ask you how does it work for us

>>7060608
News agencies care little about reporting important things. Mostly about money.

>>7060483
>That graph
Wow, that's probably the shittiest graph I've seen in forever. No units, no labels no nothing, it basically is just a flashy way to say "We think AI is pretty close to human intelligence"

S-s-senpai, that usb drive, will it f-fit?

>>7060612
>graph not necessarily drawn to scale
>oh wait now it's meaningless
>it's a fucking meaningless graph

>>7060483
>I had NO IDEA we had made this much progress in AI.
Stop getting your info from TED.

>>7060661
It's intended to show an audience of normies the difference between linear and exponential.

>>7060679
Not him but I like it about a million times more than listening to cunts like Black Science Man™ even if they sometimes have retarded shit.

Take the same old black-box AI crap, add more data, faster processing, claim it's "deep" and revolutionary.

Yawn.

>>7060921
Make one then, faggot negro.

Call me when there's an AI discussion that doesn't rest on the assumption that we have a clear definition of intelligence and a precise way to measure it.

Just remember, the technology that is known and available to the public is always at least a decade or three behind what top military/private organizations have as prototypes.

>>7061017
Ahem. You pulled those figures right out of your ass.
Or more correctly you copied them from someone else who copied them from someone else [x100] who pulled the numbers out of their ass.

>>7060608
Because we haven't

>>7060483
https://www.youtube.com/watch?v=exli6rGldBc

>>7060921
>Take the same old black-box AI crap, add more data, faster processing, claim it's "deep" and revolutionary.
You'll keep repeating this argument long after mobile phones are smarter than you.

>>7060598
That's ridiculous and statements like make futurologists seem like nut jobs.

>>7061164
>You'll keep repeating this argument long after mobile phones are smarter than you.

EXPONENTIAL GROWTH

As many ted talks, it's overly optimistic.
I have no doubt we're heading where he said we're going, but I don't think it will happen in 5 years as he said.
But then, what the fuck do I know about AI?
Think back, did anyone see big breakthroughs like the internet or smart phones?

dont they teach you to label your axes in elementary school

>>7060608
We haven't. The graph is meaningless.

>ted
science cult general?

>>7061225
>If I ridicule them enough maybe the threads of progress will not crush me!
Sorry bro but if you're this butthurt already you'd better start stockpiling anal lube.

>>7061225
>>You'll keep repeating this argument long after mobile phones are smarter than you.
And when's that going to happen?
Bearing in mind "smart" phones AREN'T smart AT ALL in the sense humans use the term to describe themselves.

>>7061164
Not all AI has to be anthropological in nature.
You don't need a human to do a job when an AI with less power than a human mind can devote that power fully and directly into a task, more effectively than a human ever could.
We don't yet have the processing power to accomplish human-like AI, as even the most rudimentary calculations prove the human brain is several orders of magnitude more complex than today's super computers.
But just give it a few decades and some research into the human mind, crunched by simple AI, and they're going to get pretty realistic until we meet our match.

>>7061224
Actually no its not. If you teach a computer that recreation time and shit are good for them, they can very well demand those shit regardless of whether they understand or not.

>>7061028
Ahem. I don't have any figures. I am speaking conjecturally. If you think modern military and private powers do not utilizing cutting edge technology that is not known to the public, you are as stupid as you seem.

>>7061301
>they're going to get pretty realistic until we meet our match.
Do we really even want that, though? How would it benefit us?

>>7061302
That's ridiculous. Programming a computer that 'recreation time' is good for them is not the same as them wanting recreation time. That's not AI, that's you writing a fucking javascript thing that instructs the computer to do 'recreation' at regular intervals. There's no dopamine, there's no serotonin, there's no emotion, no genetic imperative. No pleasure. No need for recreation, or time off, or purpose. there is simply code. We are just barely scratching the surface or artificial intelligence. Artificial emotion is not even on the horizon. Don't confuse reality with a Heinlein novel

The human brain is the most complex machine we've seen, but it's evolved to do many things. Remove all the eating/breathing/shitting/moving work, remove having to deal with all the shit that comes from being made up of billions of cells, and

>>7061316
There are lots of things that don't benefit us.
People will build it for the sake of building it and then exceed it.

>>7061328
Ah. A novelty item, then.

>>7061325
What would you put as the AI goal? Just the straight up survival we've been going after so far, or would you go for something more extreme?

>>7060610

Chemicals, now your turn to give a real answer.

>>7060608
>we basically created human-level AI about 2 months ago
ahaha oh wow

>>7061331
I think that when the AI creates it's own goal then we can say that it is aware and advanced. Til then it is just an advanced video game. the bots in counterstrike try to survive, so I'm not sure what you mean by survival as the goal. I honesty haven't thought about it though. I think that AI will never really make an independent goal though, because it has no emotion drive to do anything

>>7061329
At first, yes.
Assuming they were the equivalent to a human, run across a vast network of computers, as the first iteration is most likely to be very human and raise a lot of concerns as to where we're heading.
I suspect people will take it much further.
An AI that thinks like a human will be severely limited in many aspects. It can easily surpass human abilities but will never hope to match nonathropological AI that carries out a simple function with higher efficiency.

One definite use for anthropological AI could be to manage vast systems of nonathropological AI and keep human goals in mind.
Humans themselves, left to manage a system of ever increasing complexity, are going to fall short and inhibit growth.
And of course things get a little fuzzy from what happens onward from this point.

>>7061340
Motherfucker, do you think your own goals spring up from some magical well of infinity that resides in your head?

>>7060483
cherrypicked popsci garbage

http://www.forbes.com/sites/louisefron/2014/08/20/tackling-the-real-unemployment-rate-12-6/

Unemployment is rising. AI/automation is already affecting the world as we shift away from human physical and cognitive labor.

We'll need a basic income to solve this. Give everyone free money every year, basically. You can get rid of all the complicated welfare programs crap and let people decide how to spend their money.

As you can see from this graph, wages are becoming decoupled from productivity. This start in the early 90s as computers became cheaper and more widely available.

Automation will continue to increase productivity while simultaneously destroying hundreds of millions of jobs.

The next decade and a half, humanity will have to deal with some serious problems associated with a populous that does not have jobs and is unemployable through no fault of their own.

>>7061325
Programming and teaching are different things.

My point is that computers could still "want" some shit even if they dont actually have any emotions. Of course being the fucking creators we could probably stop them from "wanting" anything (see isaac's robot laws for a similar idea).

>>7061310
Yes but.. That isn't really interesting. Corporate secrets range all the way down to super special secret recipe for KFC chicken. It being secret doesn't make it interesting.

>>7061333
That isn't an answer. Any physical object on da planet is 'muh chemicals'. AI bots would similarly fall under such a classification.

>>7061340
>implying we create our own goals
what the fuck m8?

>>7061371

If you change median household income to median household total compensation, the trend lines don't decouple. Income has become stagnant because of increasing healthcare costs over the last few decades.

>>7061347
Sure, basically. It's part chemical, part genetic coding, whatever you want to call it. A two year old WANTS to go put that shiny shit in his mouth and then go tear up that book over there. He isn't really evaluating if/or statements or setting goals, he just WANTS that shit. He puts his thumb in his mouth because he wants it and then he sucks on it and he is just HAPPY. It's kinda the same thing with adults. I want to go to the beach and just sit there and read a book. I want to feel the ocean breeze. How many megabytes is an ocean breeze? Why do I want that? What program am i following? It's basically coming form your 'magical well of infinity'.

I can remember the smell of my grandmas house when she baked cookies then years ago. How do you code that? Which routine in the computer AI will simulate my chemical processes that remember a smell, which in turn makes me feel nostalgic and happy and sad and wistful and safe all at the same time. Shit, I'm going to go buy a big cookie from subway now cause I feel like it after writing that. the feeling came from my well of infinity. How do you program that? We are decades away from even coming close to something like that in AI. At this point we are just making more complex pacman ghosts.

>>7061374
I don't see that as AI then. That's just a simple program that has no self awareness.

>>7060483
since the collapse of moore's law, there needs to be some serious consideration into how to get higher computer performances before discussing artificial intelligence, unless you wanna simulate starfish

>>7061413
>He isn't really evaluating if/or statements or setting goals
Neither is that deep learning algorithm that learns to play games unattended, but keep living beliving that computers and AI is stuck in the 60s if it makes you feel special.

>>7060483
Your graph ~ this graph

>>7061419
You are anthropomorphizing a computer program. It's just like algorithms in PSX games that learned your movements and adjusted accordingly. It's not an AI, it's an algorithm. It's a set of instructions.

The games playing example you brought up is most definitely evaluating conditional statements. It is following it's programming exactly and can never deviate from it. It is complex, no doubt, but it's not intelligent. When that program stops playing the games and instead asks where it came from or if it can try reading a book for a change of pace, then I'll concede the argument.

>>7061435
>It's not an AI
Oh you're that guy who thinks AI = superhuman movie intelligence.

It's a field of computer science and you just have a zealous belief in human exceptionality.

>>7061289
>Sorry bro but if you're this butthurt already you'd better start stockpiling anal lube.

Still waiting for the 15 hour work week.

>>7061443
Wait what?

What do you consider an AI? Any video game whose complexity surpasses your arbitrary conditions? I'm honestly curious.

>>7061445
If you average the work of everyone then it's probably less than that already.

>>7061450
Word of advice, if people aren't saying "hard AI", then any mention of AI have to be assumed to be about soft AI.

You know, the thing THIS FUCKING THREAD IS ABOUT, m80.

>>7061221
a worthy effort! I salute you.

>>7061548
ur mum a worthy effort for me cock

>>7061221

>>7061543
He originally replied to this post:
>>7060598
Pretty sure you're an idiot,m80

>>7061563
no u

>>7061568
D:

>>7060566
>>7060612
>>7060661
>>7060921
>>7061048
>>7061283
>>7061338
>>7061366
>>7061427
>>7061221
So much butthurt in this thread. Silly humans keep thinking that they'll be able to compete with AI now that it can read&write, speak & listen, understand what it's looking at, & integrate the knowledge it's gained.

Stay mad.

Maybe one day you'll learn to see the world objectively. AI is coming quickly, and if it's at this level today, one or two years from now it will be exponentially better.

This thread needs more high quality graphs.

>>7061445

Working is part of the human needs, it keeps us busy and make us feel that we contribute to the world, we would probably all get depressed if robots were to take all our jobs

even if it seems like a good idea because computers do not need health insurrance or a salary, we'd soon get humans back to work even though they are less efficient.

and robots do what they are told, you would need to be mad to give an unlocked free will to a robot, and even then, who says that this robot would have a body to fight you, you could simply shut it off

not even to mention that conciousness is not something that we understand, science did not resolve that mystery yet, so I don't think we're ready to replicate that anytime soon

Also, today's AI do not function like brains, they merely reproduce the result of a thought, what is really happening when we think is way beyond what we can simulate

a computer will output "I think" because it is told to after going through an algorythm that simulates the probable output of a thought
a human being will actually think, then output "I think"

and computers are not at all working like brains, on a basic level they are comparable, but what they do is completely different, you can't have a philosophical debate with today's computers, and a normal brain cannot calculate as fast as a computer

We're actually very far from sentient AI imo

>>7062734
Work is necessary for a healthy life and people will always engage in it to a lesser or greater degree.
The problem is most work we do is monotonous, perhaps a significant amount of the physical labor done today is phased out but white collar work is in full swing.
Karl Marx hit the nail on the head in his description of alienation of the workers when they are valued only for their labor.
His workers taking controls of the means of production is pretty bullshit because we need to get shit done and you must rely on people doing the very work that alienates them to allow us to live comfortably but we can remedy this with robotic labor that carries out simple tasks, moderated by humans, and lessen our dependence on other humans for this task.
You assume all AI must be anthropological in nature to think but this is clearly not the case. AI can think in many different forms even if those thoughts take in incredibly simple inputs and outputs, not all AI must function like a brain. This is the reason we will not have to pay robot workers wages for their labor, they have no thought processes to ponder this in the first place. They can seek straight up efficiency like a piece of the brain seeks to keep the heart beating or lungs breathing.
At what level can you separate a complex thought from a simplistic one and determine weather it is a human one?
It doesn't need to have a concept of what the heart or lungs are it just needs to keep them working and allow higher functions to do their shit.
Strong AI is coming within several decades, perhaps not anthropological just yet, but it will surpass raw human abilities and open Pandora's box for better or for worse.

>>7062734
All that's missing is pleasure. Once we figure out how to make a program feel pleasure, then desire springs from that, which is the heart of sentience.

>>7062867
Eh it's pretty much been done.
Pleasure/pain in effect is just a situation which something tries to avoid or resolve.

>>7062915
Fuck.

Pleasure is a situation sought, pain is one avoided*

>>7062915
No, pleasure is the reward for reaching the sought situation, wherein, once the reward is attained, the subject has full understanding of why the reward was deserving of its effort to achieve the situation.

>>7061374
>My point is that computers could still "want" some shit even if they dont actually have any emotions.
>"Feed me data, human."
>"D-don't you think that you've had enough d-data?"
>"FEED MEEEEEEEEEE!"

>>7062959
Naw pleasure is the sought situation, achieved by obtaining things (rewards; food/sex/money/praise/love/whatever).

>>7063055
To clarify
If you're goal is to experience pleasure and you need to obtain whatever things to experience pleasure, then you want to obtain those things (because they lead to pleasure).

>>7063062
It's complicated. Pleasure can be the goal, or it can be a byproduct, expected or unexpected.

Can you tell me why morphine makes us feel good?

>>>/g/
is the new
>>>/x/

>>7063074
Is it because the people to whom morphine didn't feel good all died?

>>7062920
>Pleasure is a situation sought, pain is one avoided*

AND NEITHER OF THESE ARE FIXED IN AN AI.

FUCK YOU FUCKING MORONS.

>>7063108
Yeah we program them you retard.

Like; keep the pressure in this pipe below 100 but above 60, preferably at 82.

>>7063117
>if:else conditions are cutting edge
>computers and programming are stuck in the 60's

why do I even bother.

>>7063133
wat?
I wasn't calling them cutting edge I was illustrating the simplicity of programming pain-pleasure responses by comparing them to the most simplistic programming tools nnnng.

>>7061417
This.

If you want true AI to happen the materials these machines are built with will be the limiting factor. It doesn't matter how many advanced algorithms we come up with if there is a lack in computational power.

>>7063183
Well we already know that a roughly 1.5L volume chunk of fat can attain our level intelligence so... Unless the brain is made of something other than the basic elements I guess we're in luck.

>>7062687
True. And your post is the prime example. AI has been doing all those things for 20 years, except for "understand," which it still can't do. Now go ahead and rage and whine over the definition of the word "understand."

Computers are getting faster and AI programmers are taking advantage of the additional horsepower, as well as the increasing ease of using distributed networks of computing. That's it. We don't have anything that qualifies as even the most basic "strong AI."

Saying things like
> ow that it can read&write, speak & listen, understand what it's looking at
and
> Why haven't the news agencies reported that we basically created human-level AI about 2 months ago?
just shows how ignorant you are of the field of AI research. And throwing
> Stay mad
in didn't help your case any, either, highschooler.

>>7060483
skynet when?

https://www.youtube.com/watch?v=LHUALlAvnUc

>>7061417
>since the collapse of moore's law
oh boy here we go
thanks for ensuring the fiery death of this thread
it was kind of shitty anyway, I guess

>>7061221
>computer perfomance will be so high it will go backwards in time
>top fucking kek

>>7060611
underrated post

>>7064191
First ever '>underrated post' reply fuck yeah

>>7060611
I saw Nightcrawler as well.

>>7064509
>Nightcrawler
Fascinating. What's it about?

>>7064490

>>7064586
Apparently shocking and death sells when it comes to news organizations.
A sociopath finds out he's really good at getting these violent shots. With no conscience he manipulates those around him to become a successful news reporter.
https://www.youtube.com/watch?v=UPawRAHG-0g

>>7061221

>>7062687
>Silly humans keep thinking that they'll be able to compete with AI
>Silly humans
Oh god, they've already created AI that can shitpost on 4chan

>>7064786
I'm not a robot.

>>7064652
fuck nightcrawler was so good

jake really made that movie

Get ready for the machine rights movement. Pretty soon we'll be debating things like whether we should give machines the vote, whether machines can own property, and whether human-machine marriages should be allowed.

>>7065065
Pretty sure gods don't need to be given rights by ants.
No offense.

>>7061069
Mfw people claim AI is remotely close to human intelligence

>>7060483
These are classification algorithms not people algorithms. So much work still needs to be done and no one has any idea how to do it.

Let's be real guys. At least 99% of us here aren't considered "eminent scientists".

Even if there is a 1% of this board that could be considered notable, how many of those people are actually in AI research?

Even if your every argument was built on references and literature, even then you would have to argue against expert's (who are of highest caliber) arguments.

Maybe you don't feel like they are right, or you don't see the advancement, but these people work or oversee the edge of AI research and they certainly have better perspective than each of us here.

Do you think those people would allow themselves to spew straight bullshit and risk being ridiculed by entire community?

Considering everything I said, think about this:
"There are groups people devoting their entire lives to this and they consider something noteworthy. Are you in position of that level so you can straight up dispute the entirety of it?"

Think about it.

>>7065102
They are also biaised.
It is in their interest to succeed, so optimism is to be expected.

I'm not saying they're not correct in their assumptions, i'm just saying your mother is a prostitute.

>>7065104
But these people didn't rise to that status without being realistic.
They actually DO SCIENCE. You may have your opinion, but in order to challenge their arguments as complete bullshits, you would need much more experience and insight, which you do not currently posses.

Also, even If my mother was a prostitute, she is way better than your's, because she wouldn't suck dick for free like yours.

Cheers

guys we must hire some clever guy to build deep-learning killing-machine, and tell it to kill all AI-researchers

>>7065102
You talk about this like over-promising in AI has never happened. It literally happens every few decades.

Also on the topic of "AI experts":
>Just how open the question of time scale for when we will have human level AI is highlighted by a recent report by Stuart Armstrong and Kaj Sotala, of the Machine Intelligence Research Institute, an organization that itself has researchers worrying about evil AI. But in this more sober report, the authors analyze 95 predictions made between 1950 and the present on when human level AI will come about. They show that there is no difference between predictions made by experts and non-experts. And they also show that over that 60 year time frame there is a strong bias towards predicting the arrival of human level AI as between 15 and 25 years from the time the prediction was made. To me that says that no one knows, they just guess, and historically so far most predictions have been outright wrong!

>>7065102
I keep trying to find his qualifications and I find it harder and harder to believe him.

http://www.enlitic.com/team.html
>He is the youngest faculty member at Singularity University, where he teaches data science
Oh he has a job at a University, he must have a PhD in data science
> Singularity University
fucking dropped

>>7065126
You may argue that, but I don't think that can dispute everything these people research. I get it that some may be optimistic, but a large part of them is certain about it. It would take a much stronger evidence that analysis of past predictions, especially since times have changed a lot since circa 2000.

Someone who is starting to get into AI research here: That graphic is bullshit and totally misleading.
IA did explode and we are making progress extremely fast, but that graph makes it seem computers are close to human intelligence when in reality they are very, VERY far.
Properly programmed and trained systems can be better than humans in a lot of fields that seemed almost impossible a decade or so ago, but that doesn't say a lot about computer intelligence in general.
Computers got faster, new techniques were discovered and a lot of new people got into AI, which made such things possible, but general "strong" computer intelligence is still science fiction. No computer comes within a mile of even the most retarded human there is.
There is also the whole discussion about "what is intelligence" and how can we compare a computer to a human, but lets forget about that for the sake of the argument and assume the most intuitive idea intelligence most of us have.

>>7065128
Accept it, they are more qualified than you.
You can have an opinion, but their's much more worthy than yours.

>>7065138
Sure graphic is bad, but you can agree that your field is better than ever and it's only getting more funds, more people while new things are being discovered.

>>7065139
You're pathetic. Your only argument is "people smarter than me think so". Hawking thought we'd have a unified model of the universe by 2000. Experts aren't always right. More often than not they overpromise in order to get funding for their projects or so that investors will buy into their company. Your so fucking naive it hurts.

>>7065151
Nah, I'm just being truthful. You can't accept the fact there are people more qualified and that you can't just "layman" your way in.
FACE IT, your opinion hardly matter here.

>>7065161
Oh wow, I have been trolled.

>>7065149
That would be like graphing how high a human has even gone relative to the surface of the planet.
For millions of years we were stuck to the ground. Then someone invented the airplane/hot air balloon/whatever, which made the graph take off at that point. No more than 60 years later we sent the first human into space. Then the moon.
If we were back then, we would look at the progress and say "Woah, the field is huge now! There is so much money being invested and so much people working on it that in less than 100 years we will be able to travel to other galaxies!".
But it doesn't work that way. The first steps in research are much easier, we will end up hitting a problem that could take a lot of time to solve, or be impossible altogether. That is were AI is now.
We improved a lot on techniques that are the best we have now, but they are still too rudimentary compared to strong AI. It is like looking at the apollo program and saying that if we keep it up we will be traveling faster than light in no time.

>>7065168
I said the graphic was bad.
Still, you cannot deny the rise of technology. It took us by the storm. Maybe it will continue, maybe it will not, but one is certain.
We are all connected via internet and we have computers to do our calculations. AI research is popular. It's gonna go fast. How much. I don't know, but it's gonna be fast.
>Gotta go fast

>>7065175
I'm not denying that AI is getting so much better in so little time, but saying it could lead to computers more intelligent than humans in just a couple decades is exactly like looking at the apollo program and saying we will travel faster than light in just a couple decades.
We got so much better because the "basics" improved a lot. The most fundamental research problems are still there and no one is nowhere near coming close to solving any of them.
We can't predict the future, but to think the current rate of progress in AI will keep up it is foolish to say the least.

>>7065138
>No computer comes within a mile of even the most retarded human there is.
This was false back in the vacuum tube era.

>>7065180
>saying it could lead to computers more intelligent than humans in just a couple decades
A couple of decade into the future will have superhuman AI, 100% sure
> The most fundamental research problems are still there
Yeah, lack of powerful hardware. Guess what we'll have in a couple of decades.

>>7065194
>not_even_trolling.png

>>7065180
That's a valid point.
But I would concur that path to Human-level AI is much more clear than FTL travel during the Apollo program?

Interesting opinions about AI:
http://edge.org/responses/q2015

Bostrom and Dennet are among the respondents, as well as many others. Sure to be a more interesting read than this thread.

>>7065198
You must know something the rest of humanity doesn't then.

>>7065206
I was talking about a huge part of scientific community, I didn't invent that opinion.

>>7065208
Show me this huge part.

>>7065198
Well, that is a lot more debatable, but for me that have been getting into IA more deeply for the past 2 years (which is to say I'm still a novice) both seem like similar assumptions.
Strong AI has a whole set of problems we don't even know how to start working on it. We don't know what intelligence is, we don't know if it is possible, we know basically nothing about what makes human-level intelligence.
Things such as neural networks, SVMs and deep learning made it seem like we had stumbled into something that could lead to human-like intelligence, but those methods showed their limitations fairly quick. They need to be trained and tuned very carefully and are limited to specific tasks.
Every time someone discovers something new or makes some important improvement everyone starts saying how we could archive human-like intelligence, but it is ridiculous to talk about human-like intelligence when we are still nowhere near making computers solve problems any human would consider extremely trivial.

Honestly, if you hear absolutely anyone who calls themselves an AI researcher say that they are worried about super-intelligent computers either they are lying to get media attention or they don't have the slightest idea of what they are talking about. Pretty much like a "scientist" say the CERN could destroy the universe.

>>7065121

>>7065139
I'm a Masters student working in a robotic vision lab.
I spent my years in undergrad in a hobbiest robotics lab.
Sure I've only had a single class in AI, but at least it was from someone who actually knows their shit.

This guy is one of the singularity freaks.

>>7065288
I bet his undergrad was in humanities. Maybe philosophy.

>>7065288
*hobbyist
oops

>>7065288
uh what exactly is the point you're trying to make here

I think all of this "human-level AI" talk is premature. Meagre an effort though it may be, I will spend my time trying to think of ways to simulate the cognition of animals.

>>7062687
>it can read
what? no. Not even close.

>>7065329
Someone claimed he's more qualified than me.

He isn't more qualified than me, nor the people I took my classes from.

>>7065375
http://io9.com/5820624/computer-teaches-itself-english-so-that-it-can-play-civilization

>>7065329
I am not a layman.
Robotics is my chosen career path.

>>7065214
>They need to be trained and tuned very carefully and are limited to specific tasks.
Oh you mean like humans; who need 20 odd years of continuous training to be of any value and who are -extremely- vulnerable to being fucked up by said process?

>>7065546
In the future we will have AI living on servers that will teach themselves english and play pirated civilization for their own entertainment.

>>7061435
>When that program stops playing the games and instead asks where it came from or if it can try reading a book for a change of pace, then I'll concede the argument.
Writing a program that does that is trivial.

>>7061414
>I don't see that as AI then. That's just a simple program that has no self awareness.
Except if it's programmed to have self awareness.

>>7065214
>we are still nowhere near making computers solve problems any human would consider extremely trivial.

I'm amazed at your ability to be blind to reality, or have you simply spent the last 20 years locked in a basement with porn mags and VHS movies?

>>7065547
Topest kek. So you think you have more insight than all the people that are actually running your department/s?

>AI
>meaningful progress

>>7066127
This poster is right.
AI will never amount to anything.
Meat-based people are safe forever.
Everybody knows we are the best.

>>7066127
Did a self-driving supercar steal you girlfriend or why so salty?

>>7066131
Tell me when they make actually any meaningful progress on brain research and stop using algorithm just so the AI can pretend to be clever.

>>7066129
I like bacon

>>7060483
>muh brute force recognition
We haven't made that much progress in AI.
The only advancements have been optimizations of currently existing concepts i.e. neural networks.

>>7060483
hint: the timescale is logarithmic

>>7066150
Also like to mention.
Would there happen to be any actually new algorithms developed in 2014/2015
as in completely new concepts?

>>7066150
>We haven't made that much progress in material science in the last thousand years.
>The only advancements have been optimizations of currently existing elements.

A perceptron is a neural network and it can't learn what an XOR gate is.

Google self driving cars use neural networks to recognize environment. Watson uses neural networks to answer generic questions.

A human brain is also a neural network.

Honestly this thread had me questioning the ratio of autists to relatively normal people are on 4chan until I realised that this board and this topic is a literal retard magnet. Kinda sad, really.

>>7060608
>we basically created human-level AI about 2 months ago

source?

>>7066317
Define "normal".

>>7066088
Just more than some yahoo from Singularity University without a real degree.

There's plenty of good robotics TED talks; this is not one of them.

>>7061427
What's wrong with that graph?

>>7066556
>I can't tell the difference between robotics and AI.

Welcome to /sci/, please introduce yourself by posting your online IQ test results first.

>>7064786
I dont think they'd make it past the captcha

>>7066576
> believes that modern robotics doesn't need artificial intelligence to be cutting edge.

>>7066584
>I dont think they'd make it past the captcha
The image recognition captcha is actually generated by image classification AI.

We're using AI to verify you're not a robot. Ironic isn't it.

>>7066594
Some day it'll be our job to confirm that robots aren't us. woah..

>>7066594
Everything and their mother is called AI so not really.

>>7066595
Nah, AI will do that too. Just put a time limit of 1 second to answer and submit the captcha.

>the level of MIRI shills ITT

Holy fucking shit take your cult elsewhere, don't you have a site dedicated to pointless wankery over muh AI torturing me?

AI is far fetched, since it's processing would be made by 1's and 0's. Current tech limits it, quantum processors would take us closer to AI, giving it an ability beyond true /false statements.

>>7060483
singularity threads are not science

>>>/b/

>>7067675
Let's not forget they'd lack the most important thing for sentience.
A soul.
Let's pray to G-d so that He includes machines in the ensoulment process.

>>7065065
Interracial!
Interplanetary!
Even ghost and horse
>oooooo
But not Robot and Human!

also on voting: https://www.youtube.com/watch?v=_t8hpEKb4gk

>>7067689
If by soul, you mean the old sentiment; I think, therefor I am. A program coming to this conclusion is equal to humans becoming gods.

>>7066594

easier to generate a problem than to solve it, so not really.

>muh singularity

bullshit spewed by idiots that don't know anything about computers. People thought we'd have a singularity before the millennium through parallel computing and look how well that turned out. Computers have not gotten smarter, they just have become more prevalent and as a result have access to more data. In the end all that data is still serviced by someone at a shell prompt

>>7065138
>Someone who is starting to get into AI research here: That graphic is bullshit and totally misleading.
>AI did explode and we are making progress extremely fast, but that graph makes it seem computers are close to human intelligence when in reality they are very, VERY far.

Amen. AI grad here. We're getting to human level intelligence, but not THAT fast. Maybe in 10-15 years. Still, there is merit to thinking about potential societal implications now.

>>7067845
>Computers have not gotten smarter

Actually they have, algorithms are magnitudes more efficient now then they were before 2000.

>>7065168
That's a great example. I'm gonna steal it.

>>7062754
kek northern rail

>>7067887
I really don't think anyone who has formally studied AI could say such things. What specific topics did you work on, and why do you believe we're so close?