[quote=Wired]Darpa has a well-earned rep for some of the most ambitious, over-the-top research programs of all time. But this might be the most over-the-toppest of all. The very first step? Create a unified mathematical language for everything the military sees or hears.
The armed forces are overwhelmed by all the data its various sensors are sniffing out. They want a singe data stream that combines drone video feeds, cell phone intercepts, and targeting radar. Darpa’s solution, found in the brand-new Mathematics of Sensing, Exploitation, and Execution program is to design an algorithm that teaches the sensors how to interpret the world — how to think, how to learn and what data, accordingly to collect.
Sensors “process their signals as if they were seeing the world anew at every instant,” Darpa laments in its call for algorithms. To put it in Philosophy 101 terms, existence is, to a sensor, what William James called a “blooming, buzzing confusion“: an unmediated series of events to be vacuumed up, leaving an analyst overloaded with unsorted data. Wouldn’t it be better if a sensor could be taught how to filter the world through a perceptual prism, anticipating what the analyst needs to know?
That’s the specific military application of MSEE. But to get there, Darpa take a rather unconventional path. To get the “economy and efficiency that derives from an intrinsic, objective-driven unification of sensing and exploitation,” it wants to create an “intrinsically integrated” algorithm for the machines to interpret reality. “All proposed research must describe a unifying mathematical formalism that incorporates stochasticity fundamentally,” Darpa tells would-be designers.
In other words, one mathematical formula has to teach machines how to create order out of the chaos of the world around them, and to use that common ontology to develop a “learning capacity and expected rate-of-learning.” Naturally, human interaction is to be limited: the sensors should “learn in unsupervised or semi-supervised fashion” instead.
“Specifically excluded is research that results primarily in evolutionary improvements to the existing state of practice,” Darpa writes. You think? It even italicizes that passage in its bid. If you’re going to teach an infrared sensor pod how to make sense of the shapes it observes, there’s no half-stepping allowed.
By the time the MSEE’s produced a prototype — about three and a half years (!) — multiple types of sensors ought to be able to orient themselves using the algorithm. Specifically, Darpa says that an MSEE prototype has to “furnish sensor output products” from imagery and video, communications intercepts and the tracking of a moving target. If your algorithm can train those very distinct sensors how to determine for themselves what relevant data is, you’ll have gone a long way to draining oceans of data into a customizable kiddie pool for military analysts.
Oh, and you also may have introduced a new kind of artificial intelligence to machines used to track people and deadly weapons of war. At the least, you’ll have designed a mark-one Cylon, one that might recognize other sensors as its kin. Darpa — notably! — is silent on the most critical question of all: what will reality look like to a sensor?[/quote]
[url=http://www.wired.com/dangerroom/2011/01/darpa-wants-a-new-language-to-explain-everything/][source][/url]
This reminds me of the thread about how someone is trying to explain language to computers with 4D vectors.
Never change, DARPA.
Never change.
I had never thought of it that way. Since mostly everything code is binary 1's and 0's, the thought of a unified math language sounds incomprehendable. It would be the same to try and recreate the sensory input decoding in the human brain.
I want to head DARPA.
Sounds like those matrix screens that the operator guys inside the ship watch.
Say what you will about the U.S.'s massive military budget but has been responsible for many DARPA inventions in the past.
[QUOTE=RBM11;27356925]Say what you will about the U.S.'s massive military budget but has been responsible for many DARPA inventions in the past.[/QUOTE]
The Military budget can be cut in half and DARPA wouldn't feel a thing.
[QUOTE=Beafman;27356437]I had never thought of it that way. Since mostly everything code is binary 1's and 0's, the thought of a unified math language sounds incomprehendable. It would be the same to try and recreate the sensory input decoding in the human brain.[/QUOTE]
Which is what we also need.
Can someone put this in layman's terms for the less mathematically inclined? It sounds like what they're suggesting is to have computers/machines process information in a similar manner to the human brain.
[IMG]http://i275.photobucket.com/albums/jj288/DarkDrakonus/Rex.jpg[/IMG]
Always one step closer.
[QUOTE=Canuhearme?;27369275]Can someone put this in layman's terms for the less mathematically inclined? It sounds like what they're suggesting is to have computers/machines process information in a similar manner to the human brain.[/QUOTE]
It seems to me that they want to develop an algorithm that allows programs inside drones/"robots" to automatically learn from its environment without human interaction.
Edit:
I'm thinking maybe "avoiding obstacles", "tracking moving objects" and so on.
[QUOTE=Swebonny;27369360]It seems to me that they want to develop an algorithm that allows programs inside drones/"robots" to automatically learn from its environment without human interaction.
Edit:
I'm thinking maybe "avoiding obstacles", "tracking moving targets" and so on.[/QUOTE]
That would match with the design parameters of the "BIG-DOG". Robots need to be able to see what's around them and build a plan to avoid or overcome it.
A sort of Decision making engine.
[QUOTE=Hellghast;27369395]That would match with the design parameters of the "BIG-DOG". Robots need to be able to see what's around them and build a plan to avoid or overcome it.
A sort of Decision making engine.[/QUOTE]
Yeah, I mean the "LittleDog" project has something similar. I don't know if the learning is automatic or human controlled though. Really cool thing, I want one.
[media]http://www.youtube.com/watch?v=nUQsRPJ1dYw[/media]
What other "secondary" applications could this have?
[QUOTE=Swebonny;27369512]Yeah, I mean the "LittleDog" project has something similar. I don't know if the learning is automatic or human controlled though. Really cool thing, I want one.
[media]http://www.youtube.com/watch?v=nUQsRPJ1dYw[/media][/QUOTE]
Holy-
:psyboom:
[QUOTE=Swebonny;27369512]Yeah, I mean the "LittleDog" project has something similar. I don't know if the learning is automatic or human controlled though. Really cool thing, I want one.
[media]http://www.youtube.com/watch?v=nUQsRPJ1dYw[/media][/QUOTE]
I still think this one's a better video (purely because of the ice parts.)
[media]http://www.youtube.com/watch?v=nb18sfiEpF8&feature=related[/media]
[QUOTE=Canuhearme?;27370311]I still think this one's a better video (purely because of the ice parts.)
[media]http://www.youtube.com/watch?v=nb18sfiEpF8&feature=related[/media][/QUOTE]
That noise it makes is kind of creepy.
[QUOTE=FalconKrunch;27369083]The Military budget can be cut in half and DARPA wouldn't feel a thing.[/QUOTE]
Their sole purpose is to develop technology for the military. When they get funding from the government it's military funding.
[QUOTE=Canuhearme?;27370311]I still think this one's a better video (purely because of the ice parts.)
[media]http://www.youtube.com/watch?v=nb18sfiEpF8&feature=related[/media][/QUOTE]
[media]http://www.youtube.com/watch?v=VXJZVZFRFJc[/media]
[QUOTE=Latency;27372016]Their sole purpose is to develop technology for the military. When they get funding from the government it's military funding.[/QUOTE]
I'm just saying that if america cut the spending on a useless war, DARPA could still get the funding it needs.
[QUOTE=FalconKrunch;27374872]I'm just saying that if america cut the spending on a useless war, DARPA could still get the funding it needs.[/QUOTE]
Oh yeah I see. Makes sense since they're always looking for new tech, regardless of whether they're in war or not. I bet wartime allows them to pool more money into it though.
[QUOTE=Latency;27375286]Oh yeah I see. Makes sense since they're always looking for new tech, regardless of whether they're in war or not. I bet wartime allows them to pool more money into it though.[/QUOTE]
Cold wars are the best substitute for a major shooting war, however.
[QUOTE=Canuhearme?;27375325]Cold wars are the best substitute for a major shooting war, however.[/QUOTE]
Yeah, but for a cold war you need the constant threat of being attacked by someone exactly as strong as you, which really sucks.
[QUOTE=ASmellyOgre;27381860]Yeah, but for a cold war you need the constant threat of being attacked by someone exactly as strong as you, which really sucks.[/QUOTE]
Don't forget you need McCarthyism!
[QUOTE=Swilly;27381885]Don't forget you need McCarthyism![/QUOTE]
Not really, considering that lasted what, 4-5 out of the 45 years of the Cold War?
[QUOTE=Hellghast;27369302][img_thumb]http://i275.photobucket.com/albums/jj288/DarkDrakonus/Rex.jpg[/img_thumb]
Always one step closer.[/QUOTE]
That's a bit too far-fetched really. One of these is more likely to come true, in one form or another.
[img]http://images.wikia.com/terminator/images/e/e4/T1.jpg[/img]
[img]http://pds12.egloos.com/pds/200808/31/60/c0001960_48b983d434791.jpg[/img]
[img]http://www.digitaljournal.com/img/2/1/0/8/2/9/i/4/0/4/o/Aierial_Hunter-Killer_from__Terminator_.jpg[/img]
[QUOTE=RayDark;27356498]I want to head DARPA.[/QUOTE]
Didn't you play Metal Gear Solid?
[QUOTE=Canuhearme?;27370311]I still think this one's a better video (purely because of the ice parts.)
[media]http://www.youtube.com/watch?v=nb18sfiEpF8&feature=related[/media][/QUOTE]
They should call it ROBO-HORSE
Also BZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZAAAAAAAAAHAAAAHAHAZZZZZZZZZZZZZZZZZZZZZZZZ
[QUOTE=kidwithsword;27384859]Didn't you play Metal Gear Solid?[/QUOTE]
[B]METAL GEAR???[/B]
[IMG]http://www.blogcdn.com/www.joystiq.com/media/2007/02/metal_gear_awesome_snake.jpg[/IMG]
[QUOTE=BANNED USER;27369612]What other "secondary" applications could this have?[/QUOTE]
Would be good for search engines and robotics in general.
Sorry, you need to Log In to post a reply to this thread.