Silicon Brains That Think As Fast As a Fly Can Smell 84
Nerval's Lobster writes "Researchers in Germany have discovered what they say is a way to get computers to do more than execute all the steps of a problem-solving calculation as fast as possible – by getting them to imitate the human brain's habit of finding shortcuts to the right answer. A team of scientists from Freie Universität Berlin, the Bernstein Center Berlin, and Heidelberg University have refined the idea of parallel computing into one they describe as neuromorphic computing. In their design, a whole series of processors designed as silicon neurons rather than ordinary CPUs are linked together in a network similar to the highly interconnected mesh that links nerve cells in the human brain. Problems fed into the neuro mesh are broken up and processed in parallel, but not always using the same process. The method by which neuromorphic processors handle problems varies with the way they're linked together, as is the case with neurons in the brain. The chips are designed to copy the layout and functions of brain cells, but the way they're interconnected is based on another highly efficient biological model. 'The design of the network architecture has been inspired by the odor-processing nervous system of insects,' said one of the researchers. 'This system is optimized by nature for a highly parallel processing of the complex chemical world.' In tests using real-world datasets, the prototype was able to match the performance of specialized Bayeseian pattern-matching systems. Even better, the stable decisions reached by 'output neuron populations' take approximately 100 milliseconds, which is the same speed required by the insect nervous systems on which the network design is based, according to the paper."
flies can smell? (Score:2)
In order to properly evaluate this story I would need to know the rate at which flies smell. although presumably silicon ICs can move faster than that.
Re: (Score:2)
But does it run linux?
Imagine a beauwolf cluster?
Or are we not doing that anymore?
you mean a swarm of flies
Re:flies can smell? (Score:4, Funny)
These must be time flies
You know, the ones that like arrows
Re: (Score:2)
No, they're certainly not time flies. Or maybe they are... over sixty years ago they were calling the giant mainframe computers that were less powerful than a musical Hallmark card "electronic brains." Now that they're using silicon instead of vacuum tubes, ignorant fucks who have no idea whatever how computers work (my sister, when her grandson asked how computers worked, shrugged and said "it's magic") are calling them "silicon brains".
Brains do more than compute. These are not brains.
Fast as a fly can smell.. (Score:1, Funny)
Is that fast enough to make the Kessel Run in less than twelve parsecs?
Um ... (Score:3, Funny)
Re: Um ... (Score:1)
Shut up before we get an argument that lasts a hundred lightyears.
Re: (Score:2)
I guess Ep IV was before your time
and Han fired first
Re: (Score:2)
For a BRIEF second there... (Score:1, Offtopic)
Guess that means that mine can't think as fast as my house banana flies can smell.
Re: (Score:2)
...I read it as: Silicon Breasts That can Think... Guess that means that mine can't think as fast as my house banana flies can smell.
Silicon breasts sound a bit hard to me
Short Cuts (Score:3, Funny)
That's the last thing we need: robot overlords who keep taking shortcuts. Next thing you know, they'll kill all humans and then go bankrupt from ill-advised mortgages!
Re: (Score:2)
That's the last thing we need: robot overlords who keep taking shortcuts. Next thing you know, they'll kill all humans and then go bankrupt from ill-advised mortgages!
Which is why we should let them run the banking systems so they go bankrupt first. Oh, wait...
uh... what? (Score:1)
The method by which neuromorphic processors handle problems varies with the way they're linked together, as is the case with neurons in the brain.
First of all, no one knows how neurons are linked together in the brain(*).
Second of all, as far as anyone can tell, the cerebral cortex is a repeated pattern of small structures ("Cortical Columns" [wikipedia.org]) which are, again - as far as anyone can tell, wired identically.
There's some variation: The afferent and efferent layers have thicker neuronal sections which correspond to "amplifiers" needed to send and receive signals to the rest of the body, the pre-frontal cortex is an endpoint layer, and there's lots of or
Re: (Score:1)
First of all, no one knows how neurons are linked together in the brain
I figured it out, but I'm waiting for my paper to be published. TBD for now.
Re: (Score:2)
Re:uh... what? (Score:5, Interesting)
You're pointing to articles on high end mammalian brain structures when TFA is referring to the most basic structures in an insect brain. Also, this blanket assumption that no one could possibly understand the complexity of a small group of neurons is way out of date.
Olfactory circuits are pretty well understood. This isn't the first simulation of neurons mimicking an olfactory bulb at the single neuron level. We've been watching videos and seeing presentations of these models for years now. What is neat, here, is that they're modeling a somewhat realistic hardware instantiation of a model (as in, this is something which maybe could be built).
I come at this from the other end. I make the chemical sensing hardware that mimics the response of a biological chemical sensor (an artificial insect 'nose'). There are long running collaborations between my field and neuromorphic computing folks to develop a combined sensor-processor that can electronically understand smell in the same way a living thing does. I have to sit through their talks on modeling neurons, and they have to sit through my talks on nanosensor arrays.
Re:uh... what? (Score:4, Funny)
I come at this from the other end.
The butt?
Re: (Score:2)
I am talking patch-boards. pots, switches, lots of meters and analog chart recorders here, fellas, without a keyboard in sight. One programmed these with straight math, where you literally wired your equation into the machine.
Back in my day, I could solve problems on those things thousands of times faster than I could on a digital min
Great. Low-quality evolutionary "solutions" (Score:4, Interesting)
Solutions that evolution produces (whether real or simulated) typically suck, as they are typically just good enough for the training criteria and may even completely fail longer term. This really is nonsense, unless you have very low quality requirements. And, unlike a solution based on understanding how to solve something, this bio-inspired stuff cannot easily be improved incrementally from seeing how it performs in practice.
Re:Great. Low-quality evolutionary "solutions" (Score:5, Interesting)
Re: (Score:3)
BTW: Not exactly the link I was looking for, but same topic: http://www.genetic-programming... [genetic-programming.com]
In a final real-world test, Koza chose a filter circuit to solve a design problem that a scholarly engineering journal had deemed too difficult to solve. "The tenth-order elliptic asymmetric bandpass filter was touted as being difficult to design, but we were easily able to solve it," Koza said.
To be fair, Koza did have to double the size of the population used to evolve a bandpass filter-up to 640,000 circuits-thereby multiplying the time it took the computer to evolve a "best" circuit. He had to devise a more extensive fitness measure by which the members of the evolving population were measured against one another. The problem took four days to run, on a 64-CPU parallel processor.
This article is from 1996, so I guess the same algorithm would be even faster now.
Re: (Score:1)
Thanks for this, interesting. +
Perhaps the argument of the effectiveness of evolutionary processes as a design tool revolves around the specificity of a problem (as gweihir points out below).
Maybe the more broad the problem the lower the potentials and greater the iterations needed to refine and vice versa.
Re: (Score:2)
Oh, sure, if you have exactly defined criteria, evolutionary processes work. But a) you rarely have a complete spec b) once you step outside of that spec, you are typically screwed and c) you can do nothing to predict how changes in the criteria will impact an evolutionary optimized thing, while for one where you actually have a working theory, you can say how robust its performance are.
The typical situation in science and engineering is that you do not have a full set of criteria and that you do not know t
Re: (Score:2)
Sounds like the textbook definition of the GP's, "[...] this bio-inspired stuff cannot easily be improved incrementally from seeing how it performs in practice.". I think that being able to improve things in that way is important, too, since these kinds of evolutionary processes are like rolling a ball down a hill to find the lowest point; it's good at finding the local
Yes, Adrian Thompson's Discriminator GA (Score:4, Informative)
See On The Origin of Circuits [damninteresting.com]:
"As predicted, the principle of natural selection could successfully produce specialized circuits using a fraction of the resources a human would have required. And no one had the foggiest notion how it worked."
"Dr. Thompson peered inside his perfect offspring to gain insight into its methods, but what he found inside was baffling. The plucky chip was utilizing only thirty-seven of its one hundred logic gates, and most of them were arranged in a curious collection of feedback loops. Five individual logic cells were functionally disconnected from the rest-- with no pathways that would allow them to influence the output-- yet when the researcher disabled any one of them the chip lost its ability to discriminate the tones. Furthermore, the final program did not work reliably when it was loaded onto other FPGAs of the same type."
"It seems that evolution had not merely selected the best code for the task, it had also advocated those programs which took advantage of the electromagnetic quirks of that specific microchip environment. The five separate logic cells were clearly crucial to the chip's operation, but they were interacting with the main circuitry through some unorthodox method-- most likely via the subtle magnetic fields that are created when electrons flow through circuitry, an effect known as magnetic flux. There was also evidence that the circuit was not relying solely on the transistors' absolute ON and OFF positions like a typical chip; it was capitalizing upon analogue shades of gray along with the digital black and white.'"
Dr. Thompson's publications seem to be difficult to find in free viewing form on the Internet, but the daminteresting article gives the gist of it: evolution will eventually make use of whatever characteristics are available to solve a problem.
Evolutionary Solution != "Genetic" Algorithm Deriv (Score:1)
Re: (Score:2)
What's more, there are some apparently void elements in the circuit, but still the circuit stops working when these elements are removed.
++?????++ Out of Cheese Error. Redo From Start.
Re: (Score:1)
Solutions that evolution produces (whether real or simulated) typically suck,
Your life is confirmation bias.
Evolution produces excellent solutions because they have survived changing conditions over long time spans. Evolution inherently optimises for the long term.
Seven billion humans are rather bad at finding solutions to all but the most trivial problems because they deliberately optimise for the short term.
Looked at from another angle, humans rigidly guess what is needed in advance, so end up acting in coarse discrete steps, while evolution adapts fluidly to what is needed. The s
Re: (Score:1)
Considering you are the product of evolution and therefore the ability understand a problem and solve it in steps is also a product of evolution, I'm not sure how you draw the conclusion that evolutionary solutions typically suck.
Besides that, the universe as a whole is the result of a system of hydrogen atoms evolving over the course of 13 billion years.
Of course if you limit evolution to biological evolution, I'd have to still call bullshit. Who knows what else is out there but here on earth we have an am
Re: (Score:2)
Have you looked at these bodies we have recently? They pretty much suck in most regards. Sure, they are good enough to support survival of the human race (at least up to now, the future does not really look good on that, and think back to the cold war where it was near thing several times), but that is about it.
Besides that, the universe as a whole is the result of a system of hydrogen atoms evolving over the course of 13 billion years.
I see. Sorry for my answer, I found that you have absolutely no clue what you are talking about only after I wrote it.
Re:Great. Low-quality evolutionary "solutions" (Score:4, Interesting)
Solutions that evolution produces (whether real or simulated) typically suck, ...
Evolutionary solutions do not suck at all. In fact they are often brilliant and most optimized solution with lots and lots resilience. The digital camera sensors still do not have the dynamic range of mammalian eye. Robot touch sensors still don't have the dynamic range of our finger tips. We still can't mimic a geckos adaptive suction pads to create a vehicle that runs up in vertical walls. Heck, that little suction holder for my GPS keeps falling down.
What sucks is the evolutionary process that is prodigal in its use of resources and time. On the large mammal end (elephants, gorillas, humans, whales) a typical female produces about 10 off spring (without assistance/interference from humans and modern tech). On the insect level, they produce hundreds of off spring, most of them die, a few survive for the next generation. Evolution, if we were to anthropomorphize it, would not flinch at producing 10 to 100 times more output than needed, picking 10% or 1% of the output and discarding the rest. Trees produce billions of pollen grains and their success rate, measured by how many of them end up as mature trees of the next generation is measured in parts per trillion. What if it takes 10 million generations to find the optimal solution? Well, so be it. Mother Evolution would say. What if entire species specialize too much and lose their ability to adapt for changing environment? Mother Evolution does not care, there are other species willing to fill their niche, should they go extinct.
Guessing computers (Score:1)
The more we get computers to think like humans, the more computers will become just as fucked up as humans. It's a great idea at 1st sight, but I suspect it will lead to such wonderful human conditions as confusion, multiple personalty disorders, not to mention our propensity to make a LOT of mistakes, which kind of defeats the purpose of computers to begin with.
Re: (Score:1)
But if we do it correctly then it will get confused and make the mistake of not killing all humans.
Conversion, please (Score:1, Funny)
How many Libraries-of-Congresses is that?
Old and New (Score:1)
Multi-parallel systems aren't new. Early computing history predicted computers in the future (now) would be massive amounts interconnected CPUs all working together to quickly solve problems. Instead we've gone down the path of faster is better and there are few, if any, companies that are able to take the chance on designing new computer architectures (and programming concepts to go along with them). Sure many tasks seem linear, but as a extremely simple example adding 2 + 3 + 5 + 1 can be done faster o
Frank Herbert... (Score:2)
Years ahead of it's time, obviously.
Neural networks (Score:2)
Re: (Score:1)
I saw silicon (and FPGA) built around NN in the mid 1990's. The first NN's used for computation were published in the 1960's.
The only novel part of this is the "true to biology" part of the topology.
I would be impressed if it could make a better video card - if the paradigm of "neuromorphic" was competitive vs. the current material implemented in silicon.
Ambiguous Title (Score:5, Funny)
Is it
Silicon Brains That Think As Fast As (a Fly Can Smell)
or
Silicon Brains That (Think As Fast As a Fly) Can Smell
And are those flies time flies or fruit flies??
Re: (Score:2)
It was obvious that it was meant to be:
'Silicon Brains that think as fast as a fly (can smell)'
Now the questions I have are:
"How well can it smell?"
or
"Just how good or bad can they smell?" (as in 'like a bed of roses', or a skunks ass?)
Note to TFA's submitter and /. editors:
There was(and still is) very good reasons for the inclusion of punctuation as part of the English language.
Re: (Score:2)
Time flies as fast as an arrow can time flies
So now our bugs can have bugs (Score:2)
I wonder.... (Score:2)
I wonder if linking silicon neurons together to mimic the functioning of actual neurons in the human brain will lead to the same inaccuracies and false assumptions that humans make all of the time?
Literary device (Score:1)
In other news... (Score:2)
"Cat Videos That Viral As Fast As A Llama Can Spit"
Enough with the inane, idiotic, punction-free parallels already! How many football fields is that, again? And where's my car analogy!?