http://arstechnica.com/gadgets/2016 /08/ibm-phase-change-neu

>Charles_Darwin.jpg

Richard Dawkins is a phony """"""""intellectual""""""""" who does not compare to Darwin in the slightest.

u
whot
m80

>>8245483

Nice, not singularity-nice, but still pretty nice

File: big_thumb_fb9682e28c1e89d7be1e8d33276f2283[1].jpg (48KB, 736x414px) Image search: [Google]

48KB, 736x414px

DELET THIS

>>8245483
At this pace there's little that can be done to prevent machines from taking over everything. Expecting some humans will be able to defeat their limits and advance to the next level because that's the only way to win.

File: 4jnh2h.jpg (76KB, 512x600px) Image search: [Google]

76KB, 512x600px

>>8245483

I was now when it happened.

Is this hardware trainable, like a software neural net? The entire secret sauce that makes deep learning work at all is the error backpropagation algorithm + stochastic gradient descent; the article mentions backprop but the diagram only shows it going intra-neuron, from the threshold back to the generator, presumably to help control its output. Also, I'm not even sure whether backprop (which relies on differentiability and smoothly variable parameters) would even work with spiking neural networks, which are very much discrete and stochastic with variable probability rather than outputting a number of variable magnitude.

Neural networks are generally not very expensive to *evaluate*, even without special hardware (beyond a good GPU); it's *training* that's the real bitch.

>>8245493

What has Richard Dawkins contributed to evolutionary biology, anon?

Can this be used to replace meat neurons?

File: 3322.jpg (27KB, 500x446px) Image search: [Google]

27KB, 500x446px

>>8245573

>meat neurons

>>8245483
Oh yes, I am really excited to hear what the literal inventor of memes has to say about this

>>8245532
>Neural networks are generally not very expensive to *evaluate*, even without special hardware (beyond a good GPU); it's *training* that's the real bitch.

>Second, these phase-change neurons are the closest we've come to creating artificial devices that behave like biological neurons, perhaps leading us towards efficient, massively parallel computer designs that apply neuromorphic approaches to decision-making and processing sensory information. IBM says that their new work is complementary to research being carried out into memristor-based synapses, too.

I would assume when they say "artificial devices that behave like biological neurons" they mean that the actual hardware itself behaves like neurons and can be taught to do something, even without a virtual software neural net layer on top of it.

>So far, IBM has built 10x10 crossbar arrays of neurons, connected five of those arrays together to create neuronal populations of up to 500 neurons, and then processed broadband signals in a novel, brain-like way. (In technical terms, the neurons showed the same "population coding" that emerges in biological neurons, and the signal processing circumvented the Nyquist-Shannon sampling theorem).

When they connect more of these it exhibits "brain-like" behavior so seems it's at hardware level. We just need software to make use of this 'behaviour'

http://www.nature.com/nnano/journal/v11/n8/full/nnano.2016.70.html

In any case here is the actual article. Perhaps you can make more sense of it.

>>8245573
Technically, it's a step in the right direction. Though, we would have no idea how to actually replace our "meat neurons", surgically I mean, or whatever.

>>8245568
Que?

>>8245568
memes

>>8245601

>Richard Dawkins is a phony """"""""intellectual"""""""""

In your view, what has Dawkins contributed to biology?

>>8245611
memes

>>8245609
>>8245615

Yes, yes he did.

Also contributed massively to the gene's eye view of evolution.

And gave us the notion of the extended phenotype.

Great ethologist and geneticist.

>>8245568
"Bitches ain't nuffing but hoes" Richard Dawkins, who was not involved in a study, has reportedly mentioned this.

>>8245483

Holy fuck balls.

If [read: when] we acheive a complete neural mapping of a brain, we can begin constructing a like for like artificial one.

>I jizzed

>>8245601
You say ¿Qué?

>>8245619
>Also contributed massively to the gene's eye view of evolution
fair enough

Just sign me up for that consciousness transfer now senpai.

>>8245622
http://stanfordflipside.com/2013/10/study-confirms-that-bitches-as-suspected-aint-shit-but-hoes-and-tricks/

We see in this seminal study by his contemporaries how he was actually right.

>>8245862

No. You're full of idiotic programming.

We'll transfer you and then edit the shit out of you.

>>8245483

Most people don't realize that the military has used what could be considered to be "neural ASICs" in actual military gear for a few years now

This thread made me realize that I might even be able to get my unwashed hands on one https://en.wikipedia.org/wiki/TrueNorth

>>8245483
Investment on these researches in education would produce FAR MUCH MORE EVEN ARTIFICAL NEURONS

okay, this isn't my area of expertise, so this may be a stupid question, but what exactly do they mean by saying this circumvents the Shannon-Nyquist Theorem?