Silicon brain
Researchers have built a chip with the equivalent of 200,000 neurons and 50 million synapses in an effort to mimic a human brain in silicon. I, for one, welcome our simple-minded overlords. From Technology Review:
Although the chip has a fraction of the number of neurons or connections found in a brain, its design allows it to be scaled up, says Karlheinz Meier, a physicist at Heidelberg University, in Germany, who has coordinated the Fast Analog Computing with Emergent Transient States project, or FACETS.Building A Brain On A Silicon Chip (Thanks, Marina Gorbis!)
The hope is that recreating the structure of the brain in computer form may help to further our understanding of how to develop massively parallel, powerful new computers, says Meier...
FACETS has been tapping into the same databases. "But rather than simulating neurons," says Karlheinz, "we are building them." Using a standard eight-inch silicon wafer, the researchers recreate the neurons and synapses as circuits of transistors and capacitors, designed to produce the same sort of electrical activity as their biological counterparts.
the latest
latest episodes
Isn`t this how Skynet started?
While cool, also a tad unnerving.
This is starting to approach the level of complexity of an insect brain.
There are all kinds of things insects can do that no computer can do today. Imagine an artificial insect that flies as well as an insect can, and that could sniff out particular substances as well as an insect can. There are certainly some interesting possibilities.
The insect sniffing is due to sensory apparatus, not neurons. So this won't help with that.
I seem to recall reading about a project to model a fly (fruitfly, of course) brain, a few years ago. Did anything ever come of that?
most real brains i know are pretty stupid... they make lots of mistakes and are inclined to grossly simplify things in the interest of computational efficacy. i much prefer the cold hard, logical machines to these :)
Definitely an impressive feat, but those who believe that a powerful computer is the biggest obstacle to strong A.I. are bound to be disappointed. Calling a computer chip an "artificial brain" because it has the same number of neurons is like calling your dining room table an "artificial horse" because it has the same number of legs.
That said, I'm really looking forward to how this technology will affect my PlayStation 5.
I disagree, Sean. The sensors aren't nearly as tricky as computing the right reaction to the sensors: they don't provide a direction, only an intensity.
And kfractal: let me know when you have a cold, hard, logical machine that recognizes faces as well as a three-year-old child.
This approach has been tried before, with pretty conspicuous failure. Maybe faster and denser logic will make the difference... but I suspect that there are levels of self-organization in the biological system that go beyond what can can be done easily in hardware or quickly in software.
To quote John wayne's Ethan Edwards, ''That'll be the day!''
When will our Turing-wannabes meet some Fraa Orolo reallybes?
4649 wishes to hear about scientific investigation of consciousness done by researchers who appreciate the Chinese Room arguments as Searle presents them. (He's got an awesome lecture series on some teaching company discs!---I think, or thought. Whereas Google gives a promising link to tpb, a cursory search at teach12.com came up only with a series done by one Professor Grim.)
That's very interesting. However, from what it seems, they'll just be playing with connection weights, like any simulated artificial neural network. What it's not modeling, however, if the huge architectural changes that occur in a brain's early life. Neurons grow and die and reach out to neighbors in a manner that is directly correlated to sensory experience -- a understimulated brain has fewer connections. It is this ground-up wiring of the brain (coupled with half a billion years of evolving a basic structure from which to build) that provides far more intelligence than the day-to-day or month-to-month adjusting of weights.
arent there people working on processors that work in a holographic sense - that already mirrors the way the brain works in alot of ways??? wherede i read about this - these two research groups should get together and have a "brain child"
Gunnm/Battle Angel Alita reference anyone?
I really don't understand why you could want to use such an imperfect model... I mean, the human brain is FAR from perfect... seems to me we should aim much higher... after all, if I want a human brain I'd just have a child...
and it if did work like a human computer wouldn't it just be like "rainman"? people who function on higher levels only do so at cost to "other" functions...
I'd say the best "well rounded" human I have ever met definitely would not make a good computer...
life is imperfect... I demand much closer to perfection from my computer...
I agree that ultimately it's pointless to design a machine that is limited in capacity as a human. Star Trek's Data was certainly deficient in certain area's (like humor or emotion), but excelled in computational areas.
Right now we have a hard enough time getting a robot to walk and recognize people. Think about how much data you process when you drive a car. Or more exactly, think about how much data you sift through to extract what you really need. I know scientist say we should have a humanistic level of AI by 2050, but I still think that's a long shot. Unless we move away from the standard digital model to something much more complex, then it's a real long shot. Quantum computing or light based computing might solve these issues.
i kinda have this weird idea that human intelligence is only being limited by outside forces...
maybe its just my bell-non-local quantum-bias shining through again - (i try not to discriminate against mere 3 brained humans... but its hard)
Just to be off-topic some:
So, what's the 'fraction'? Seven eights? Twenty-two sevenths?
/Don't get me started on "Three times less"..
/Re-read "Exhalation"
@ PaulR: Never heard of the phrase "idiomatic expression," have you?
@flytch, bcsizemo:
Humans are still superior to computers in many, many arenas. Face recognition, language recognition and comprehension, ability to hold a conversation are some things that are directly hardwired in our brains.
Strategic decisions without perfect information is another -- anything from playing Go (computers are still crap) to conduction diplomacy and waging war.
Pattern recognition, recognizing that one situation is similar enough to another that the lessons learned from one can be applied to the other -- even if the first situation involves zen koans and the second involves financial decision making, is another powerful function that requires seemingly immense parallel processing and symbolism.
Many of what people think of as "flaws" are actually evidence of a superior processor. When we remember events, for instance, all of them, even those we think we remember quite well, are grossly distorted, with salient features standing out, even exaggerated, and others blurring away. But this is not a flaw: this is exactly what makes them useful. This is what allows us to generalize from them, to learn. Otherwise we'd be like the too-often-brought-up Funes the Memorious, who could remember everything but understood nothing, and could not generalize. If we had to search through every memory we ever had, like photographs, we'd have no conceptual understanding of our past.
The abilities of our brains is immense, and one fails to realize how they can relate to computers if they merely consider computers as dumb word processors or whatever. No one is suggesting that your iPhoto program gets replaced by one that forgets photographs. But to suggest that there aren't vast qualities of the human mind that wouldn't be immensely powerful in a computer is absurd.
@ SamSam:
"Humans are still superior to computers in many, many arenas. Face recognition, language recognition and comprehension, ability to hold a conversation are some things that are directly hardwired in our brains."
It could be argued that voice recognition software trained to recognize English is superior at recognizing English than, say, a five year old Chinese boy who's never taken any English courses. Also, it "learns" to recognize English at a much faster rate than any human being could. Likewise, face recognition software is able to pick a known face out of a large crowd at a billion times the rate of speed that a human being can.
You and I most likely agree that a computer can and will outpace the human brain in certain areas, if not all of them. They have already made a mockery of human cognitive ability in terms of mathematics. And to use your example again, I would suggest that face recognition is another task at which computers are far superior, depending entirely on the scale of the project, as well as the sophistication of the processing and programming used to enable the computer to perform its tasks. It is not as if voice or facial recognition are individual tasks.
When you say that imperfect memory is what enables us to generalize, you make a fair point. However, when there are people with arguably photographic memory, or even something approaching photographic memory, who are able to make such generalizations, I would say that it is an empty overgeneralization. After all, it is specificity, and not generalization, that catches criminals. Most criminologists will tell you that serial killers don't really have that much in common, and yet we build up all sorts of fanciful generalizations about them. This is why criminal profiling has taken a beating from academics in recent years. It's counterproductive, usually. Misremembering things is a human deficit, and I would argue that, in some people at least, generalization borders on self-deception.
And I am still confused as to the "equivalent" of a neuron, Mark. How do you measure something so esoteric? It is not mass we're talking about. So, what then? Computational power? I would say that any chip made of silicon could only be a simulation; not a true equivalent. Could there be a silicon equivalent of a white blood cell? If injected in a human being, would they keep the host alive?