physorg.com — (PhysOrg.com) -- Lyric Semiconductor has unveiled a new type of chip that uses probability inputs and outputs instead of the conventional 1's and 0's used in logic chips today. Crunching probabilities is much more applicable to many computing task performed today rather than binary logic.
Aug 19, 2010 View in Crawl 4
ubernogginAug 19, 2010
This is probably cool.
Closed AccountAug 20, 2010
Put both types in a machine and you get right brain left brain. I for one will probably welcome our new computer brain overlords.
kaliorAug 21, 2010
oh, I get it
Closed AccountAug 21, 2010
I see what what you did there
earthtonesAug 19, 2010
Certainly an interesting idea. They should call it the "poker chip". Heh.
paranor01Aug 20, 2010
It's probably an interesting idea, no certainty with probabilities.
earthtonesAug 20, 2010
that might be a good point.
lederhosedAug 20, 2010
so it can tell me when that monkey will turn out hamlet? yes!
alphalionAug 20, 2010
So, click on an app, and it might open, or might do what it was advertised to do. Don't computers already do this?
paranor01Aug 20, 2010
Does this mean we can now build an infinite improbability drive if we invert the probability inputs?
Closed AccountAug 20, 2010
Came for Douglas Adams reference. Not disappointed.
solecizeAug 20, 2010
maybe
neutron7Aug 20, 2010
I prefer bistromathics myself. Infinite improbability drives are hard on the whales.
oninboninAug 20, 2010
Splat!
paranor01Aug 20, 2010
"Oh no. Not again."
timsaucemnAug 25, 2010
They first need to build a finite improbability calculator, and then calculate the improbability of the infinite improbability drive. Only then, can the infinite improbability drive be created.
Closed AccountAug 20, 2010
The various outputs of this chip:
● As I see it, yes
● It is certain
● It is decidedly so
● Most likely
● Outlook good
● Signs point to yes
● Without a doubt
● Yes
● Yes – definitely
● You may rely on it
● Reply hazy, try again
● Ask again later
● Better not tell you now
● Cannot predict now
● Concentrate and ask again
● Don't count on it
● My reply is no
● My sources say no
● Outlook not so good
● Very doubtful
Closed AccountAug 20, 2010
Magic 8 bit?
hutch619Aug 20, 2010
Magic byte
chadsexingtimeAug 20, 2010
"Outlook not so good"
True, but microsoft released it anyway
gerrykAug 20, 2010
A computer that doesn't operate using logic?
Behold the femputer!
ikorkyiAug 20, 2010
i like this.
chuckdontsurfAug 20, 2010
THAT DOES NOT FEMPUTE!
Closed AccountAug 21, 2010
CANNOT FEMPUTE
kaliorAug 21, 2010
goddammit, I read chuckdontsmurf's comment in Farnsworth's voice
flanativeAug 20, 2010
Next up, Intel to sue over their use of the Pentium 60 math co-processor unit....
hanniballecterAug 20, 2010
Binary, as a default language of electronic signals, is an old hold-over from the early days of electronics where it's hard to accurately tell what the signal is saying because of degradation, interference or whatever. That's why they either make it so that the signal either denotes a 1 or a 0 (or an on and an off), because it's easy to figure out.Comment is buried, click here to see the rest.
skeloothAug 20, 2010
Thank you capt. Obvious
stap1egunAug 20, 2010
I'm having trouble understanding why this is more useful than just simulating probability in binary form. Is it just more efficient somewhat like a GPU is optimized for graphics calculations, or is it a fundamentally new method of computing similar to quantum computing?
I would be grateful if someone could explain this for me. Thanks!
bobby1978Aug 20, 2010
Imagine a box. Next imagine a cat inside that box. Now imagine millions of such boxes with cats inside them. Then imagine a knife (electrons), which produces results. This is probably how it works.Comment is buried, click here to see the rest.
gunit99Aug 20, 2010
wat
appleofdischordAug 20, 2010
The cats are now diamonds.
Closed AccountAug 20, 2010
I think you are on the right track with your "GPU" analogy. Built in processing is many times faster than in-memory processing. The article mentioned the difference between digital NAND gates Bayesian NAND gates. I'm assuming the chip would provide a very-low-level instruction set for setting up Bayesian equations.
http://en.wikipedia.org/wiki/Bayesian_probability
stap1egunAug 20, 2010
Ok, that's kind of what I figured especially after reading the comments on the article page too. Thanks for the help!
homercles337Aug 20, 2010
The only thing i can think of is that computers are limited with respect to random seeds. We do fine with tricks using the CPU clock, but its still pseudo-random. For stochastic processes (statistics) this is important since many analyses rely on "consensus results" (ie, many simulations with different random seeds) because no one random seed is truly random, but many random seeds approach random. That being said, many basic statistics are deterministic and would not benefit from a genuine stochastic seed. Mostly, im not excited about this and im a computational scientist.
caughtthinkingAug 20, 2010
Next up, http://en.wikipedia.org/wiki/Technology_in_The_Hitchhiker's_Guide_to_the_Galaxy#Infinite_Improbability_Drive
spritomAug 20, 2010
Is it powered by a nice hot cup of tea?
Closed AccountAug 20, 2010
So it's like a mentat,, right?
regeyaAug 20, 2010
Sounds like just the thing to control an improbability drive.
Closed AccountAug 20, 2010
That's not new.
http://en.wikipedia.org/wiki/Fuzzy_logic
Hardware using fuzzy logic have existed for decades.
Closed AccountAug 20, 2010
Fuzzy logic isn't new. Having this particular technology on a chip is. The main benefit is faster, real-time calculations.
honoredmuleAug 21, 2010
This sounds fundamentally more generalized /in low-level application/ than fuzzy logic coded on a traditional chip. Here the whole point is that the probability calculation is essentially handled by a non-discrete application of low-level electronic "simulation" (i.e. NOT calculation via boolean gates). Implementing a chip that hard-codes fuzzy logic routines using traditional logic gates (or even n-valued gates) is barely even similar and certainly not the point of this article. This is more like a specialized analog computer, where the specialization is one where analog's tradeoff of discrete accuracy for speed does not render it completely useless.
I would imagine the difference in performance might be comparable to the difference between optimized adders that cascade all carry bits in one step and their predecessors that consume a clock cycle to handle each pair of nth-place bits + carry bit. That we can already perform such calculations with existing specialized hardware or software isn't a terribly interesting point.
fuzzynyankoAug 20, 2010
Reminds me of the above poster using cats as an example
hardwalkerAug 20, 2010
this is insane o.o
this is how nature works, irobot and s**t it's COMING
solecizeAug 20, 2010
and it's going to vacuum your floors in a improbable fashion!
raydarAug 20, 2010
sorta like a roomba right now.
hardwalkerAug 20, 2010
i actually have one of those, my cats love it
drmangrumAug 20, 2010
I'm confused on how this would operate. Determining probability takes a very specific data set in order to operate. Amazon can offer product suggestions because they can easily cross reference a purchase trend among buyers purchasing a single item. If you're buying a skateboard, you're probably interested in skater-culture tee-shirts and such.
How is a chip going to determine these things without a supporting database? Only way I could see this working is if it facilitates aggregating data and integrates the math functions directly.
honoredmuleAug 21, 2010
At the core of these data set matches are still lower level math computations and specifically non-discrete calculations of match quality (i.e. skater-culture tee-shirt A is a .87943 similarity match with skateboard purchase). In this case, similarity and probability are different interpretations of the same value and use the same discrete set of numeric-only inputs.
pabstyloudmouthAug 20, 2010
Actually i was working on making a transistor with trinary code but it is not working out so well, "maybe" does not come in hand very often.
Closed AccountAug 20, 2010
Can anyone actually explain this in plain English?
(I'm expecting the answer to be "probably")
neutron7Aug 20, 2010
probably.
Closed AccountAug 21, 2010
I gave the answer...why does he get more diggs? :-P
nyxerebosAug 20, 2010
It's a math co-processor for statstical stuff. Often one needs to calculate with probabilities, perhaps a great many probabilities when using stochastic methods. This chip does so faster than general purpose computers. At least, this is what I gather from the article. It's not especially clear on the details.
http://en.wikipedia.org/wiki/Stochastic
http://en.wikipedia.org/wiki/Monte_Carlo_method
Closed AccountAug 21, 2010
I said ENGLISH :-P
honoredmuleAug 21, 2010
It's a black-box statistics (probability) calculator. The interesting and useful innovation is that the innards of that black box are highly optimized at a low (electronic) level rather than just coding complex calculations using traditional binary circuitry. This bypasses some mathematically inaccurate behavior caused by the traditional computer's finite concept of numbers as well as in theory paving the way for such calculations to be performed a few orders of magnitude faster than with traditional chips.
duckylamAug 20, 2010
The improbability drive should surface any day now...
mentosanAug 20, 2010
A robot using built with these microprocessors would be something like Marvin
http://www.youtube.com/watch?v=dxp1AqWleT4
blacklilyninjaAug 20, 2010
I'm reminded of all the men who chose not to read the instructions.
lethargicmonkeyAug 20, 2010
so it produces an answer, and says, "yeah, its probably right... but hey, don't melt me down if i'm wrong."
volfie99Aug 20, 2010
Why am I turning into a penguin...?
samuraighostAug 20, 2010
Because it's better than turning into an apple or a multi-colored piece of glass.
aminy23Aug 21, 2010
So when there is an article that requires intelligence to understand, the pro-Linux comments actually get dugg up.
samuraighostAug 21, 2010
It's like preaching to the choir.
Closed AccountAug 20, 2010
It just uses Bayesian NAND gates instead of traditional NAND gates. It still makes use of transistors like every other chip, it just has a different suite of on board logic functions. Not that this isn't cool, but this isn't the high scify advancement that one would infer from the title. The benefit here would be performance with hard wired functions versus in memory simulation.
Closed AccountAug 20, 2010
So approximately 0.0000000001% of the time, the computer will respond with a simple "f**k off and die" screen of death?
honoredmuleAug 21, 2010
You have to admit, that would be a huge improvement on the current 0.0001% of the time.
ironrexAug 20, 2010
One step closer to "The Heart of Gold" http://en.wikipedia.org/wiki/Heart_of_Gold_%28spaceship%29#Heart_of_Gold
unbannedaccountAug 20, 2010
Exactly, and just about as fictional. Makes for a good story, but come on.
"uses probability inputs and outputs" yah... a bit over simplified.
"there are many potential applications. Ben Vigoda stated: "To take one example, Amazon's recommendations to you are based on probability."
Oh, by all means use my taxes DARPA so I get better book suggestions.
mrawesomemanAug 20, 2010
So, they've been working on this for 4 years, and won't have a working prototype for another 3 years?
unbannedaccountAug 20, 2010
This sounds like another tech that is made out to be much cooler than it really is.
drmangrumAug 20, 2010
That's taking an oversimplified look at the potential uses. In a military setting, probably is a huge factor in determining Order of Battle applications like predicting technology acquisitions and tactical analysis of a battlefield situation.
That's just scratching the surface. If the end result is stable enough, it would also be highly useful in both the AI and Modeling & Simulation fields.
unbannedaccountAug 20, 2010
Fair enough, but maybe we can put the money to pure beneficial research?
Why must new tech always be based off finding better ways to kill people, with the ancillary benefit of modeling, AI, curing diseases, etc. etc.
drmangrumAug 20, 2010
The money has to come from somewhere. Until the private sector is willing to take the risk on new technologies, it will be dominated by military funding.
Not all military applications are used to find new ways to kill people. Many applications are finding ways to MINIMIZE causalities and collateral damage.
appleofdischordAug 20, 2010
Probability computation is insanely useful in a lot of different fields. Pretty much any sort of research in any field will benefit from improvements in probability computation.
anothersoldierAug 20, 2010
Show me a computer chip that computes logic, and I'll show you a pair of trousers full of poo.
philipz78Aug 20, 2010
All computers use logic, so poo away!
anothersoldierAug 21, 2010
Computers do not use logic - they follow pre-determined algorithms that always have linear outcomes. No exeptions.
fredfredricksonAug 20, 2010
They should use it to develop psychohistory.
carterblueAug 20, 2010
This is how me make computers that can learn on their own.
cmatthes1Aug 20, 2010
Oooh... A computer chip that produces outputs of, "Meh", and "I guess". I can see how this is desired more than yes and no.
appleofdischordAug 20, 2010
Official press release:
http://www.lyricsemiconductor.com/news.htm
More detail, less speculation and faux-science.
appleofdischordAug 20, 2010
With one of these processors, scientists might be able to compute an entire genome sequence in 10 minutes for perhaps $10 in cost.
http://www.technologyreview.com/computing/26055/page1/
Not sure why physorg thinks amazon recommendations are more impressive.
ayajuliaAug 20, 2010
Am I the only one who thought about Probability armor from Sid Meier's Alpha Centauri when I read this? Man, I loved my Trance Probability Garrisons....
satirenineAug 20, 2010
Soon, we will have created the worlds first analog computer. We'll be running on solar before you know it.
digitxpsAug 20, 2010
Let's hope we can soon find out whether it's useless or not...oh wait.
gilgamesh999Aug 20, 2010
FToA: (From this other article) It says instead of inputting a 1 or a 0 that a probability is input and a probability is output.
"Lyric Semiconductor's chip accepts probabilities as inputs instead of ones and zeros, and the output is also a probability -- the odds that the two input probabilities match up. Rather than the usual NAND gates characteristic of conventional transistor schemes, the chips employ what are known as Bayesian NAND gates, named for the statistician Thomas Bayes whose field of thought is the basis for the idea."
http://www.searchthenetnow.com/tag/bayesian-nand/
So basically humans now have a higher level chip.
Like when computers first came out programmers programmed in binary.
This chip is somewhat the equivalent of what FORTRAN did for software but is actually for hardware.
(At least by what I am understanding. Correct me if im wrong.)
Sidenote: It seems like a good leap towards AI. Maybe the C++ equivalent chip inputs visuals to.
georgesvaneedzeAug 21, 2010
I believe it will provide a great amount of help to Noetic Science.
gkiltzAug 21, 2010
Probability only actually exists in the presence of logic.
Probability without logic would be like gravity without mass. NOT GOING TO HAPPEN!!!
netbookbooNov 29, 2010
Chip very small and coin so big