Dead A.I.

These things that you say are terrible (greed, lust, violence) are the very things that allow you to type on your computer. These instincts are what drove evolution in our favor. In addition to that if anyone takes out their lust or violence out on a machine then it will be less crime. And as for the ingenuity of man, our ingenuity isn't based on our industry, it's our ability to adapt. Take off your, no doubt, black sunglasses, open the blinds to your basement and get some fresh air.

>>101879
>>101886
Holy fucking shit OP, your such a fucking faggot sperglord. Have ever considered what is means to be human and what even is consciousness. We don't even know if what consider free choice is really a choice that's defined by past experiences. social structures. What stops me or even you to kill people randomly. I would sage but that article is generally interesting.

>>101892
>consciousness
now you're really overthinking it

>>101879
>implying that will stop me from fucking one
Gynoid qts>3dpd

>>101879

Unless you have a phd/masters in computer science: machine intelligence/machine learning/artificial intelligence, I'm really not interested in your laymen definitions of personhood for machines.

>>101916
You don't need any qualification to see the machine in the article isn't genuinely intelligent.

It's an automaton that solves a certain type of puzzle.

>souls

oh benl
when will your brain start making sense?

also, way to bury the lead:
>Model performs at human levels on standard intelligence test

thanks for the link tho

>>101922
This is not just a puzzle or a game like chess.

>The ability to use and understand sophisticated relational representations is a key to higher-order cognition. Relational representations connect entities and ideas such as “the clock is above the door” or “pressure differences cause water to flow.” These types of comparisons are crucial for making and understanding analogies, which humans use to solve problems, weigh moral dilemmas, and describe the world around them.

This is attempting to crack the very last thing that humans were supposed to be able to do better than any computer.

its another "op posts braindead beliefs rather than news" episode
thanks benl

>>101934
>This is not just a puzzle or a game like chess.

However you want to label the class of IQ problem it solves it has no capability beyond it's programming.

You can show it things to learn but that is ultimately just more programming to solve whatever 'thinking' task you set it.

>the very last thing that humans were supposed to be able to do better
The thing humans do better is learn new tricks.

It is impressive how it generalises from so few examples, certainly a step forward, but it's still a long way from surpassing human intelligence.

>>101879
>Machines will never have souls and spirits,

If you could copy your brain to the finest detail, recreated in a computer, would it have your soul?

http://www.medicaldaily.com/how-brain-works-blue-brain-project-creates-digital-slice-brain-study-neurons-356386

>>101955
It would be like my twin (or rather clone). From the point of creation, it would go off and be influenced by its own unique experiences, potentially becoming someone somewhat different.

>>101953
The recognition AIs the article talks about were never manually pre-programmed or "set" the way you describe. They really do "learn" on their own. Because of the complexity, however, the process is kind of a black box and it is unlikely we will ever "understand" what's happening. That's why we should be careful about how we proceed.

>>101980
>were never manually pre-programmed or "set" the way you describe

They were initially constructed at some point with various capabilities of comparison as described so well here >>101934

All AI have an element of pre-design.

Even a neural network has to be trained with the correct data and configured with a viable number of nodes in a sensible arrangement.

All of this makes them something like one trick ponies.

They typically excel at the task they have been designed for but need reworking pretty much for the ground up to perform well at a significantly different task.

The presentation of the the IQ problem itself constitutes 'training' in the case of this AI which is what makes it unusual.

Much faster to learn than deep learning methods, partly because it is pre-programmed with an understanding of analogy.

This method is far less of a black box than AI of recent years which is one thing that makes it progress of sorts.

http://www.mccormick.northwestern.edu/news/articles/2016/06/making-computers-reason-and-learn-by-analogy.html

This AI is designed to work with analogies which harks back to the logic engine approaches of the early days of AI research.

It makes it naturally designed to solved this kind of IQ problem.

The underlying technique and software components may have application in solving other problems but will still need reprogramming to do so.

It wouldn't surprise me if this approach was part of something that makes waves it the future but I don't think this is the breakthrough that makes computers smarter that people on it's own.

>a biological machine is special and has the ability to do things that the exact same machine but made from techy things can't do
Stay ignorant

>>102270
Humans also have elements of "pre-design", have to be trained, typically excel at the task they have been designed for, and are essentially one-trick ponies. Our mode of reasoning relies heavily on spatiotemporal patterns, and logic is not entirely natural to us. Rather, we're merely adapted to our 3D-perceived environment.

>>102371
Humans are generalists.that can solve a wide range of problems either reason or trial and error.

Usually a combination of the two because most are not very clever.

Despite this most can learn all sorts useful things often from seeing it just once.

>>102376
What you perceive as the "generalist" ability is just the wide application of spatiotemporal reason tricks we've evolved to have. Humans are not naturally proficient at reasoning, and require a lot of training to get to a point where it's second nature.

Humans and AI programs aren't so different when it comes to intelligence, partly because we've designed them in our image, but also because "learning" and other phenomena are simply higher order natural laws of the physical world.

>>102393
>Humans and AI programs aren't so different when it comes to intelligence
They're wildly different. Many orders of magnitude in terms of complexity.
Only recursive neural networks come even close to functioning in a similar way.

>>102393
>humans and AI computers aren't so different
Very false. Sylicon microprocessors and organic neural networks are fundamentally different in many many ways. Really the similarities stop at using electric currents to relay information. But brains use neurotransmitters as well and neural connections change constantly to adapt to changing conditions. Individual brain cells are more analogous to computers than brains are, a brain is more like a network of microprocessors, trillions of them. Comparing a computer to a brain is like comparing a star to a galaxy

>>102351
thank you

>>101879
The flesh is not sacred.

>>101879
These articles are garbage written by complete ignorants

>>105803
Any article that over exaggerates the benefits or dangers of machine intelligence can generally be disregarded. Is there are lot of money being pumped into AI research? Yes. Are there a lot of experts in the field? Not really.

In the end, research into AI is fragmented and is more of a corporate arms race more than anything else. As an example: VR. All these big companies working on it but there is no industry standard because everyone want their version of VR to be the de facto standard.

Echo isn't going to be running a Combat Information Center anytime soon, Siri isn't going to be controlling Palo Alto firewalls tomorrow. The research of what they want this technology to do is going in all directions and the future isn't as clear as all these articles want to make you think it is.

>>101879
Seems like researchers have created a machine that can solve specific visual problems. That's pretty neat, but it's just a small piece of the puzzle of developing AI.

Essentially, humans are a collection of processes that influence our actions; our minds are comprised of matter and function by electrical impulses and chemical processes that cause us to behave in a certain way. Theoretically speaking, it's possible that such a mind could be created from basic elements and behave in a way identical to natural, organic, human minds.

Unfortunately, we lack the technology to do so currently, and these crude approximations of intelligence are the best we can achieve. I would argue that a device that can react with specific behaviors to any situation, no matter how menial or basic, would be more similar to AI than a machine that can simply process and decode a singular type of input.

>>101879
Disregard this post, science and AI are great things and I like donkey penis in my asshole.

Judging the importance of this sans the context of the algorithm they is pointless, because it varies wildly depending on how the program contextualizes it's visual discoveries. At minimum, it's able to distinguish objects that are nested, categories them relative to other objects in the same problem, plot their positions and recognize patterns in that positioning and apply it to the objects in the third row. This is interesting, but not earthshaking. But if it's a more generalized algorithm than this could be really amazing.

>>Assuming the brain doesn't purely rely on predictable underlying mechanisms that can't be replicated by math

>>102406

Not that guy, but if you've decided to build a galaxy, you first need to know how to build a star.

AI is a meme

>>105825
Modern firewalls use AI.

>>111063
>mentions Palo Alto firewalls
>thinks he doesn't know that firewalls run rudimentary AI

>>107923
>building star
>literally just put hydrogen in one place until it starts to burn.

>>101879
>who would personify a computer program written by a group of anti-social
>The more mobility they give machines, the more destruction they can do... as psychopaths go off the reservation at random.

hmmm...
Hmmm...
HMMM!!!

>>107923
That's the most retarded thing I've read all... In the past 5 minutes.

>In order to build a plane you must first understand a bird.
>In order to build a stream engine you must first understand a windmill.
>In order to understand a cell phone you must first understand a phone.
>In order to understand asymmetric warfare you must first understand human wave tactics.
>In order to complain about a politician you must first have political experience.
>In order to be a dumbass you must first be a fucking dumbass.

How about making them see it better than we do?

>>111406
What does a galaxy without stars look like?

>>112110

According to current models for spiral galaxies (and probably quasars), a slowly-rotating blob of dark matter around some supermassive black hole or topological defect (in which case it doesn't necessarily have to be a star or remnant) surrounded by interstellar dust of pulsing through dark matter waves into bodies of sufficient density to locally collapse into star formation -- one can imagine freezing it in state and plucking all the stars out -- at that point you have about 2000 years before fusion occurs from the next dust cloud collapse -- probably less near the core.

Anyway, without light it "looks" like nothing. There will be emission of Hawking Radiation from the core in the center of wavelength on order of the radius of the hole, which is too large to stimulate emission from any dust, so none of the galaxy will glow in response. If you are allowed to leave pulsars and neutron stars and white dwarves in place after plucking out all the stars though those will glow just fine, very very dimly.