Wow, to think I came to slashdot in order to have an informed discussion about Siri, the seemingly impressive voice recognition tech. Nope, all fanboys. Slashdot definitely ain't what it used to be.:-\
Its definitely a subset of AI, but if you are interested in Machine Learning then you should check out the Deep Learning Tutorials [deeplearning.net]. They cover most of the building blocks of "Deep Learning", which you can think of as the new wave of Artificial Neural Networks. The tutorials include complete theoretical (and mathematical) descriptions of the model, as well as Python/Theano implementations. Pre-requisites would be a good math background (first year calculus should suffice), basic probability theory and coding in Python/numpy. You can learn Theano as you go along.
ps: sorry for the repost, forgot to log-in beforehand.
There is no way they can recognize 300 objects in real world conditions. I work in machine learning (academics) and the current record for generic object recognition sits at around 54-57% for the Caltech 101 database (contains images of 101 different objects). So basically the algorithms of the best and brightest minds in academia (LeCun, Poggio, Lowe, etc) get it wrong half the time !!
If any government officials are listening... Please don't waste our tax money on this !
There have been many breakthroughs in neural networks recently, which allow us to train "deep architectures" (with many hidden layers). This was not feasible with traditional backpropagation. This work by Hinton/LeCun/Bengio has led to a resurgence in the field of ANNs, with some experts now believing general AI to be attainable within the next decade.
Anyone interested should have a look at Geoff Hinton's Google Tech talk on the matter. A very interesting talk for anyone in machine learning. He does a way better job of explaining it then I could. Fast forward to 21:30 for the live demo.
This lobbying and pressure has got to stop. We're a sovereign country with our own laws for crying out loud.
You don't like how things work around here, just don't "export" or do business here. Lets face it, if it wasn't profitable for them to operate in Canada they wouldn't.
So mind your own business, try as best you can to make a buck if you so choose, but let us worry about how we run our own country.
</rant>
On the bright side of things, maybe if the radio stations have to pay to play music, maybe they`ll start playing higher quality content (aka. real music), instead of regurgitating top-20 garbage and promoting the recording industry`s agenda.
Maybe once these projects pick up speed, people will get together and start working out of a common workplace to increase efficiency ! Hmmm... but then of course they'll need a big building to fit everyone, some form of financing, cubicles....
Hey... wait a minute ! This sounds familiar...
Nope, false alarm. What a new & radical concept ! This could change everything !
As far as "generic object recognition" goes, we are VERY far from a Holy Grail. State-of-the-art algorithms so far have a 45-55% successfull recognition rate, when dealing with only 101 objects categories (Caltech 101 database). Basically, with only 101 object to choose from, your "search engine" would get it wrong half the time. Not very useful if you ask me. Let alone with hundred of thousands of categories as he claims.
On top of that, the best and brightest are already working on this problem at MIT, in Dr. Poggio's lab (Computational Intelligence), who along with Marr, started the field of computer vision. The problems encountered are still at the theory level, not in the implementation. So a GPU implementation shouldn't change much.
disclaimer: i am also a graduate student working on the problem, who also happens to have graduated from the University of Ottawa
The idea is that with enough data, you could extract the "why" automatically. For example, if you rated all Arnold Scwarzenegger 5, then its probably because you like Arnold. If however you gave a rating of 1 to Kindergarden Cop, as well as "The Game Plan" and a bunch of similar movies, the system could also infer, that as much as you like Arnold, you don't like kids movies starring washed up "action" movie stars.
This is the whole idea behind the field of "machine learning": inferring causes/relationship/structure from raw data.
Is this feasible ? Maybe, maybe not. Is this easy ? Definitly not. But then again if it was, they wouldn't put a 10$ Million prize tag on it.
"Virus writers will write something that searches around for the right place to patch"
No, they won't be able to do that. At that point, they haven't gained execution yet.
Buffer overflows require you to jump to code which is in a known place in memory (usually libraries), which in turn slingshots you back to the exploit code stored on the stack (or other). Without knowing where to jump to, your malicious code will just sit there in memory, not doing anything.
Euh... have you been to Iceland ?? The weather is HORRIBLE. July and August are the only summer they get, and even during that period it gets really wet. The women are hotter though, I'll grant you that much !
One way to get rid of them is to tell 'em stories that dont go anywhere. Like the time we went over to shelbyville during the war, I wore an onion on my belt....which was the style at the time...you couldnt get those white ones, you could only get those big yellow ones.................now where was I........oh yeah, the important thing was I was wearing an onion on my nelt, which was the style at the time, you couldnt get those...
Bottom line is that Apple is advertising their LCDs as offering "millions of colors" when in reality they only have 256k. Its false advertising, plain and simple. And yes companies should be held accountable for that.
It has served me very well since my early calculus days in high school, all the way to the present day, where i'm doing my graduate studies in electrical engineering. Trust me, being able to do symbolic calculations on a handheld calculator is priceless (way more than any "graphing" capabilities, so you shouldn't care about monochrome display) ! No more manually computing the integral of e ^ x ^ 2 for you !
Oh and i'm assuming that the uber-geeks of the crowd will be recommending HPs for their "polish notation"...
But trust me, you don't need polish notation when you have a full qwerty keyboard on your calculator. Only downside is of course is that not everyone will let you use them during exams.
This is the reason why efforts such as TrustedBSD and the upcoming features of OSX 10.5 have to be encouraged.
Users should have a way of running applications in sandboxes... In theory, a video codec should have no business registering keystrokes, and communicating over the network (aside from licensing, drm stuff which is not technically codec work), so why don't today's OS allow us to specify that ?!
A simple right-click (apple-left-click), properties - Disable network use, Allow hard drive access to media folder, etc....
Slashdot Mirror
Wow, to think I came to slashdot in order to have an informed discussion about Siri, the seemingly impressive voice recognition tech. Nope, all fanboys. Slashdot definitely ain't what it used to be. :-\
Its definitely a subset of AI, but if you are interested in Machine Learning then you should check out the Deep Learning Tutorials [deeplearning.net]. They cover most of the building blocks of "Deep Learning", which you can think of as the new wave of Artificial Neural Networks. The tutorials include complete theoretical (and mathematical) descriptions of the model, as well as Python/Theano implementations. Pre-requisites would be a good math background (first year calculus should suffice), basic probability theory and coding in Python/numpy. You can learn Theano as you go along.
ps: sorry for the repost, forgot to log-in beforehand.
I call shenanigans on this.
There is no way they can recognize 300 objects in real world conditions. I work in machine learning (academics) and the current record for generic object recognition sits at around 54-57% for the Caltech 101 database (contains images of 101 different objects). So basically the algorithms of the best and brightest minds in academia (LeCun, Poggio, Lowe, etc) get it wrong half the time !!
If any government officials are listening... Please don't waste our tax money on this !
woha ! +3 Informative ?
Guys... THE strip bar doesn't really have open wifi. It was a joke. Now come back to your desks...
There have been many breakthroughs in neural networks recently, which allow us to train "deep architectures" (with many hidden layers). This was not feasible with traditional backpropagation. This work by Hinton/LeCun/Bengio has led to a resurgence in the field of ANNs, with some experts now believing general AI to be attainable within the next decade.
Anyone interested should have a look at Geoff Hinton's Google Tech talk on the matter. A very interesting talk for anyone in machine learning. He does a way better job of explaining it then I could. Fast forward to 21:30 for the live demo.
This lobbying and pressure has got to stop. We're a sovereign country with our own laws for crying out loud.
You don't like how things work around here, just don't "export" or do business here. Lets face it, if it wasn't profitable for them to operate in Canada they wouldn't.
So mind your own business, try as best you can to make a buck if you so choose, but let us worry about how we run our own country. </rant>
On the bright side of things, maybe if the radio stations have to pay to play music, maybe they`ll start playing higher quality content (aka. real music), instead of regurgitating top-20 garbage and promoting the recording industry`s agenda.
Maybe once these projects pick up speed, people will get together and start working out of a common workplace to increase efficiency !
Hmmm... but then of course they'll need a big building to fit everyone, some form of financing, cubicles....
Hey... wait a minute ! This sounds familiar...
Nope, false alarm. What a new & radical concept ! This could change everything !
I call BS on the article as well...
As far as "generic object recognition" goes, we are VERY far from a Holy Grail. State-of-the-art algorithms so far have a 45-55% successfull recognition rate, when dealing with only 101 objects categories (Caltech 101 database). Basically, with only 101 object to choose from, your "search engine" would get it wrong half the time. Not very useful if you ask me. Let alone with hundred of thousands of categories as he claims.
On top of that, the best and brightest are already working on this problem at MIT, in Dr. Poggio's lab (Computational Intelligence), who along with Marr, started the field of computer vision. The problems encountered are still at the theory level, not in the implementation. So a GPU implementation shouldn't change much.
disclaimer: i am also a graduate student working on the problem, who also happens to have graduated from the University of Ottawa
The idea is that with enough data, you could extract the "why" automatically. For example, if you rated all Arnold Scwarzenegger 5, then its probably because you like Arnold. If however you gave a rating of 1 to Kindergarden Cop, as well as "The Game Plan" and a bunch of similar movies, the system could also infer, that as much as you like Arnold, you don't like kids movies starring washed up "action" movie stars.
This is the whole idea behind the field of "machine learning": inferring causes/relationship/structure from raw data.
Is this feasible ? Maybe, maybe not. Is this easy ? Definitly not.
But then again if it was, they wouldn't put a 10$ Million prize tag on it.
"Virus writers will write something that searches around for the right place to patch"
No, they won't be able to do that. At that point, they haven't gained execution yet.
Buffer overflows require you to jump to code which is in a known place in memory (usually libraries), which in turn slingshots you back to the exploit code stored on the stack (or other). Without knowing where to jump to, your malicious code will just sit there in memory, not doing anything.
Can someone tell me why fonts look so horrible on linux ??
God... the pictures on the website make it look like its 10 years old !!
Euh... have you been to Iceland ?? The weather is HORRIBLE. July and August are the only summer they get, and even during that period it gets really wet. The women are hotter though, I'll grant you that much !
One way to get rid of them is to tell 'em stories that dont go anywhere. Like the time we went over to shelbyville during the war, I wore an onion on my belt....which was the style at the time...you couldnt get those white ones, you could only get those big yellow ones.................now where was I........oh yeah, the important thing was I was wearing an onion on my nelt, which was the style at the time, you couldnt get those...
Are you kidding me ?! Can't I just pay $20 million or something ...
Remind me again what the business plan is for free municipal wifi ?
Oh right, there is none.
9 out of ... damnit !
at least i tried...
Ahh the memories...
"Time for another good idea/bad idea:
Good Idea: Trying to soar like the eagle.
Bad Idea: Trying to soar like the...Hindenburg"
narf !
Bottom line is that Apple is advertising their LCDs as offering "millions of colors" when in reality they only have 256k. Its false advertising, plain and simple. And yes companies should be held accountable for that.
Just an after the fact observation, which yes, has held approximately true until now. But its NOT A LAW and stop treating it as such !!
That history of dance guy is gonna be pissed !
It has served me very well since my early calculus days in high school, all the way to the present day, where i'm doing my graduate studies in electrical engineering. Trust me, being able to do symbolic calculations on a handheld calculator is priceless (way more than any "graphing" capabilities, so you shouldn't care about monochrome display) ! No more manually computing the integral of e ^ x ^ 2 for you !
Oh and i'm assuming that the uber-geeks of the crowd will be recommending HPs for their "polish notation"...
But trust me, you don't need polish notation when you have a full qwerty keyboard on your calculator. Only downside is of course is that not everyone will let you use them during exams.
i'm sure people were wondering "why would you want to fly planes into buildings ?"
So that's how Moses managed to cross the Red Sea ?!
This is the reason why efforts such as TrustedBSD and the upcoming features of OSX 10.5 have to be encouraged. Users should have a way of running applications in sandboxes... In theory, a video codec should have no business registering keystrokes, and communicating over the network (aside from licensing, drm stuff which is not technically codec work), so why don't today's OS allow us to specify that ?! A simple right-click (apple-left-click), properties - Disable network use, Allow hard drive access to media folder, etc....