Picture this as a classification problem. Given a set of inputs, the AI will make a classification of threat or non-threat. In order to do this, we must train the AI using example inputs for threats, as well as sample inputs for non-threats. One way to gather example input for non-threats would be to collect the news that intelligence has classified as 'chatter'. Non-threats would be gathered by using news from periods of time when there were no terrorist attacks.
take a big pattern-matching AI system, and "teach" it using the media records preceding other big terrorist events. 9/11, London, Madrid, etc.
It should be noted, AI systems that rely on training need many examples to learn from. One would not write a face detection system using 50 sample faces and expect it to perform very well. In work that I have done in this field, we tend to use many thousands of training examples. I'm not sure if we would have sufficient data for this type of training, as large terrorist attacks of the type listed above are fairly rare.
Cringely aims to slap some sense into Microsoft, Apple, and IBM altogether
Um, IBM makes its money through enterprise-level applications and services, with some hardward. Apple plays the hardware/music/software game. You may as well "slap some sense" into Boston Market, Sears, and Starbucks for not joining Google's model.
Re:Web 2.0: Where solutions don't need problems?
on
Web 3.0
·
· Score: 1
Yes, some people might get a bit of internet fame for creating some bit of software that has rounded corners and gradients, and you can update stuff without the page refreshing, but in my development cycles if I were to propose this:
Planning Phase
Development Phase
Testing Phase
(now we have a working, accessible application)
Development Phase 2 (AJAX it up while maintaining accessibility)
Testing Phase 2
Release
That doesn't seem like a good development cycle for an AJAX application, having worked on several. The refactoring alone would be a nightmare. People tend not to just "AJAX it up" a little bit, what would be the point? If you are smart about using AJAX, it should probably be built into the core of your application. The correct development cycle should be:
Planning Phase (planning for AJAX)
Development Phase (setting up a strong AJAX framework)
Testing Phase
AJAX is one of the technologies that a lot of people are ragging on right now, simply because people think it is the second coming and are throwing it into every webpage. In reality, you should use the technology that is right for the project, not just to "AJAX up" your cousin's band's webpage.
Look, not to be a jerk, but I probably wouldn't buy a computer online that was selling for half the retail price. I'm sure many would disagree, and tell me that I'm missing out on some good deals.
I was once defrauded of around $1,500 for a laptop on ebay. Hundreds of other people had bought the same laptop from the same "ebay store" and they retailed at around $3,000. I don't know if I did anything stupid because there were 50 other people that also bought the auctions this store had made
It may just be me, but I don't think I could ever make a purchase such as this over the Internet. Ebay is fine for buying some rare CD's, collectibles, and maybe even the fake Oakley's the parent mentioned. If I'm buying a computer, I'm definitely going to either go to a store or go to a reputable online seller like Apple.com. There are clear consequences for a company like Best Buy that rips you off. The consequences for an online seller are all together different.
It almost strikes me as reasonable that online fraud investigation would take so long. After all, you could be dealing with an investigation that spans several states, not to mention countries. Ebay could certainly hire a few more employees to combat fraud. Still, I imagine that this isn't as easy as it seems. Ebay lists so many items that a hands-on approach wouldn't make much of a difference. It seems that manpower should be diverted to more autonomous methods of fraud detection.
Lets test this theory out and take a look at the wikipedia entry for a British band that was popular in the early 80's, Theatre of Hate. http://en.wikipedia.org/wiki/Theatre_of_Hate
Led by singer/songwriter Kirk Brandon, the original group also consisted of: guitarist Simon Werner, bassist Jonathan Werner and drummer Jim Walker.
Incorrect, Simon and Jonathan were in a previous band with Kirk Brandon, The Pack.
Theatre of Hate garnered much early attention as a live act and in 1981 made their debut with the concert LP "He Who Dares Wins Live at the Warehouse Leeds".
Incorrect, the album was "He Who Dares Wins Live in Berlin".
Shortly after the album's release however, Brandon fired the remainder of the band an assembled a new line-up consisting of: guitarist Billy Duffy, bassist Stan Stammers, saxophonist John Lennard and drummer Nigel Preston (who was soon after replaced by Luke Rendle).
Incorrect. Stammers, Lennard, and Rendle were already in the band. Rendle was fired and replaced with Preston.
Another concert recording, "Live at the Lyceum", followed in 1982 before Theatre of Hate entered the studio with producer Mick Jones of the Clash to record their first non-live album debut, "Westworld", which went on to reach the UK Top 20. The album also spawned the Top 40 single "Do You Believe in the Westworld?".
The only correct sentences, although Westworld made the top 10.
In late 1982, Theatre of Hate released another live album entitled "He Who Dares Wins: Live in Berlin."
Incorrect, see the above correction.
In early 1983 the Theatre of Hate disbanded. Brandon went on to front Spear of Destiny and guitarist Billy Duffy formed the group Southern Death Cult, which would later become enormously successful after shortening their name to The Cult.
Incorrect, Billy Duffy joined The Death Cult, along with Ian Astbury, the singer for disbanded Southern Death Cult.
So, out of 7 sentences, only 2 were correct. Why was it again that I should trust Wikipedia as a source of information?
Just because the website exists, doesn't mean that I am obligated to correct misinformation.
For all I care, my microwave could function as my computer, as long as I can carry out day-to-day activities and use a normal sized keyboard and screen.
His essential point seems to be: using wireless technology, we don't care where the computer is, as long as we have a wireless keyboard and large screen. I imagine anyone would agree with that. Groundbreaking? No.
Wow, and since a prize is offered, all those masters students, phd students, and researchers who work in natural language text compression will actually start to publish their work, as opposed to whatever else they were doing.
Seriously though, the Ansari X-prize worked because it was applied to a problem that just neeeded money, not ground breaking research.
To think of this technology as simply search is close-minded. What Apple, MS, and others are envisioning goes far beyond the basic concept of search.
One of the main benefits is Open By Content: why should I drill through a folder hierarchy when I can locate my files based on keyword content, etc. If you expand on this concept, why should I even care where my files are in my computer? Couldn't all of my files be in the same directory if I had an easy way to prune my view of that directory? The concept of a file hierarchy may be useful to programmers and the back-end of the file system, but it isn't useful for the everyday user.
I lay all of the files on my computer out in a logical structure. However, this structure is of no use to someone who is not me, my layout may not be intuitive to them. I may like structure, but my mom still saves everything onto her desktop.
Imagine if you will, if the user didn't have any concept of where their files were. All they do is create them and save them. When they want a file they just search for it. The result? The computer has just become that much more simple to the every day user.
The true beauty is the fact that several hundred people have not read the above post and continue to make uninformed posts of their own. Just look below!
It is true that hollywood would have you think that face recognition works by looking at points on the face, but in reality this is rarely the case.
The best point-based algorithm I can think of would be Elastic Bunch Graph Matching, and it typically requires user interaction during the training stages. I imagine that users of a camera phone would not want to concern themselves with this process. Furthermore, it doesn't work very good.
It is more likely that a subspace projection method is being used, such as PCA ( a basic example, I know), in which the user's face is projected into a high dimensional space for comparison by some distance metric.
You bring up a good point. Read any face detection literature and you will see countless examples of systems thinking that hand drawings are faces, as well as soccer balls, pumpkins, or some wrinkle on a shirt.
A conventional face detection algorithm will easily be fooled by a photograph. Some form of face detection algortihm will be necessary to localize the face so that it can be cropped and registered. Once the face has been cropped and registered, any traditional face recognition agorithm, such as PCA, LDA, or any other subspace method will likely be fooled into thinking that the picture is indeed a face.
As a researcher, I have yet to see any commercial face recognition algorithm that performs notably better then the above methods in a controlled third-party study.
This is not a problem. I work with face recogniton and detection algortihms that have no problem working with such small images.
The real issue to look at is that the state of the art face recognition and detection algorithms do not work very good, unless they are in very controlled settings. In fact, varying the illumination on a face makes it harder to recognize than manipulating many other factors, such as facial hair, glasses, etc. I find it hard to believe that this phone will be able to take images in a controlled lighting situation.
The fact is, many news stories and movies will lead you to believe that face recognition and detection work much better than they do.
This is missing the point to an extent. IBM does not care about Linux on the desktop or laptop. IBM sold its entire PC business because it no longer finds PCs to be as profitable.
IBM will make money selling hardware and services in an 'on demand' style. To them, hardware and services are the commodities, not software. This is where Linux fits in.
This is no news in the AI community, algorithms such as GA's long ago learned how to add numbers, etc. I won't even go into detail here as I assume most readers are aware of this
This story is merely a case of someone who is excited about their work explaining it to an author who doesn't know as much about the subject matter. The author then turns around and writes a story for the lay-person who is not versed in the field. These people in turn jump to humorous conclusions.
This is a common occurance in magazines such as Discover and Popular Science, as much as I enjoy them. A good example is stories on robots, such as Honda's ASIMO. People see ASIMO do amazing things and assume that in 10-15 years we will have these robots in our homes. What the articles often fail to mention is that while ASIMO can do complex tasks, it has very limited ability to recognize a situation, such as a staircase in front of it, and decide on a course of action to take, such as executing its stair climbing procedure.
The true point of the article is that AI algorithms can teach us things about evolution. To make grand jumps and assume that these programs are even in the same playing field as SkyNet or the Matrix is to miss the main point.
As I said above, this is merely the case of a complex subject being explained in a way that is easy to digest for the masses. Even someone who had only taken a few graduate AI courses would find that many misguided statements are made in the article.
Pepsi pays for the Pepsi cooler and owns it. They can determine if they wish for coke to be placed in it.
Pepsi does not pay to be listed on google when you search for 'pepsi', and thus your analogy is poor.
Because on a laptop, you can control-click with one hand. After all, they tend to have nice, small keyboards. The control button can be pushed with the pinky, the mouse pad with the index finger, and the mouse button with the thumb.
IBM Thinkpads allow the user to tap the mouse pad, an action that is the same as clicking the mouse button located a mere centimeter below. This allows for one-finger mouse manipulation, instead of two. I don't hear anyone crying out for Macintosh to take on this functionality for the laptops.
While I'm at it, where's the little red nipple?
Has anybody actually thought about how this would really work? As a researcher and developer of face detection and face recognition technology, I have seen the top of the line algorithms and programs developed by some of the top companies in this field, and this sounds a little optimistic.
The system would require that your camera had software on it that could first detect all of the human faces in the image and then identify the correct person's face to blur. Although the top algorithms can do very impressive things in very specific environments, they do not perform as well as movies and news stories would have us think.
Slashdot Mirror
Picture this as a classification problem. Given a set of inputs, the AI will make a classification of threat or non-threat. In order to do this, we must train the AI using example inputs for threats, as well as sample inputs for non-threats. One way to gather example input for non-threats would be to collect the news that intelligence has classified as 'chatter'. Non-threats would be gathered by using news from periods of time when there were no terrorist attacks.
take a big pattern-matching AI system, and "teach" it using the media records preceding other big terrorist events. 9/11, London, Madrid, etc.
It should be noted, AI systems that rely on training need many examples to learn from. One would not write a face detection system using 50 sample faces and expect it to perform very well. In work that I have done in this field, we tend to use many thousands of training examples. I'm not sure if we would have sufficient data for this type of training, as large terrorist attacks of the type listed above are fairly rare.
Cisco was rumored to be looking at a purchase of Nintendo earlier in the year, so the idea of Nintendo being bought is not outlandish in itself.
So the fact that there was a similar, entirely unfounded, story earlier this year makes this new story not outlandish???????
If I make a claim similar to anything Dvorak already predicted, does that make it more sound?
When will Hollywood learn to churn out an endless number of unimaginative sequels, or countless movies all based on the same basic plot device?
Cringely aims to slap some sense into Microsoft, Apple, and IBM altogether
Um, IBM makes its money through enterprise-level applications and services, with some hardward. Apple plays the hardware/music/software game. You may as well "slap some sense" into Boston Market, Sears, and Starbucks for not joining Google's model.
Planning Phase (planning for AJAX)
Development Phase (setting up a strong AJAX framework)
Testing Phase
AJAX is one of the technologies that a lot of people are ragging on right now, simply because people think it is the second coming and are throwing it into every webpage. In reality, you should use the technology that is right for the project, not just to "AJAX up" your cousin's band's webpage.
It may just be me, but I don't think I could ever make a purchase such as this over the Internet. Ebay is fine for buying some rare CD's, collectibles, and maybe even the fake Oakley's the parent mentioned. If I'm buying a computer, I'm definitely going to either go to a store or go to a reputable online seller like Apple.com. There are clear consequences for a company like Best Buy that rips you off. The consequences for an online seller are all together different.
It almost strikes me as reasonable that online fraud investigation would take so long. After all, you could be dealing with an investigation that spans several states, not to mention countries. Ebay could certainly hire a few more employees to combat fraud. Still, I imagine that this isn't as easy as it seems. Ebay lists so many items that a hands-on approach wouldn't make much of a difference. It seems that manpower should be diverted to more autonomous methods of fraud detection.
It almost sounded interesting until I saw that they actually expected people to pay for it. Not only is it not free, but $99.95 a year?
Lets test this theory out and take a look at the wikipedia entry for a British band that was popular in the early 80's, Theatre of Hate.
http://en.wikipedia.org/wiki/Theatre_of_Hate
Led by singer/songwriter Kirk Brandon, the original group also consisted of: guitarist Simon Werner, bassist Jonathan Werner and drummer Jim Walker.
Incorrect, Simon and Jonathan were in a previous band with Kirk Brandon, The Pack.
Theatre of Hate garnered much early attention as a live act and in 1981 made their debut with the concert LP "He Who Dares Wins Live at the Warehouse Leeds".
Incorrect, the album was "He Who Dares Wins Live in Berlin".
Shortly after the album's release however, Brandon fired the remainder of the band an assembled a new line-up consisting of: guitarist Billy Duffy, bassist Stan Stammers, saxophonist John Lennard and drummer Nigel Preston (who was soon after replaced by Luke Rendle).
Incorrect. Stammers, Lennard, and Rendle were already in the band. Rendle was fired and replaced with Preston.
Another concert recording, "Live at the Lyceum", followed in 1982 before Theatre of Hate entered the studio with producer Mick Jones of the Clash to record their first non-live album debut, "Westworld", which went on to reach the UK Top 20. The album also spawned the Top 40 single "Do You Believe in the Westworld?".
The only correct sentences, although Westworld made the top 10.
In late 1982, Theatre of Hate released another live album entitled "He Who Dares Wins: Live in Berlin."
Incorrect, see the above correction.
In early 1983 the Theatre of Hate disbanded. Brandon went on to front Spear of Destiny and guitarist Billy Duffy formed the group Southern Death Cult, which would later become enormously successful after shortening their name to The Cult.
Incorrect, Billy Duffy joined The Death Cult, along with Ian Astbury, the singer for disbanded Southern Death Cult.
So, out of 7 sentences, only 2 were correct. Why was it again that I should trust Wikipedia as a source of information?
Just because the website exists, doesn't mean that I am obligated to correct misinformation.
Exactly.
For all I care, my microwave could function as my computer, as long as I can carry out day-to-day activities and use a normal sized keyboard and screen.
His essential point seems to be: using wireless technology, we don't care where the computer is, as long as we have a wireless keyboard and large screen. I imagine anyone would agree with that. Groundbreaking? No.
Wow, and since a prize is offered, all those masters students, phd students, and researchers who work in natural language text compression will actually start to publish their work, as opposed to whatever else they were doing.
Seriously though, the Ansari X-prize worked because it was applied to a problem that just neeeded money, not ground breaking research.
To think of this technology as simply search is close-minded. What Apple, MS, and others are envisioning goes far beyond the basic concept of search.
One of the main benefits is Open By Content: why should I drill through a folder hierarchy when I can locate my files based on keyword content, etc. If you expand on this concept, why should I even care where my files are in my computer? Couldn't all of my files be in the same directory if I had an easy way to prune my view of that directory? The concept of a file hierarchy may be useful to programmers and the back-end of the file system, but it isn't useful for the everyday user.
I lay all of the files on my computer out in a logical structure. However, this structure is of no use to someone who is not me, my layout may not be intuitive to them. I may like structure, but my mom still saves everything onto her desktop.
Imagine if you will, if the user didn't have any concept of where their files were. All they do is create them and save them. When they want a file they just search for it. The result? The computer has just become that much more simple to the every day user.
The true beauty is the fact that several hundred people have not read the above post and continue to make uninformed posts of their own. Just look below!
It is true that hollywood would have you think that face recognition works by looking at points on the face, but in reality this is rarely the case.
The best point-based algorithm I can think of would be Elastic Bunch Graph Matching, and it typically requires user interaction during the training stages. I imagine that users of a camera phone would not want to concern themselves with this process. Furthermore, it doesn't work very good.
It is more likely that a subspace projection method is being used, such as PCA ( a basic example, I know), in which the user's face is projected into a high dimensional space for comparison by some distance metric.
Although I am as skeptical as you are, most face recognition algorithms crop the face, eliminating the hair.
You bring up a good point. Read any face detection literature and you will see countless examples of systems thinking that hand drawings are faces, as well as soccer balls, pumpkins, or some wrinkle on a shirt.
A conventional face detection algorithm will easily be fooled by a photograph. Some form of face detection algortihm will be necessary to localize the face so that it can be cropped and registered. Once the face has been cropped and registered, any traditional face recognition agorithm, such as PCA, LDA, or any other subspace method will likely be fooled into thinking that the picture is indeed a face.
As a researcher, I have yet to see any commercial face recognition algorithm that performs notably better then the above methods in a controlled third-party study.
This is not a problem. I work with face recogniton and detection algortihms that have no problem working with such small images.
The real issue to look at is that the state of the art face recognition and detection algorithms do not work very good, unless they are in very controlled settings. In fact, varying the illumination on a face makes it harder to recognize than manipulating many other factors, such as facial hair, glasses, etc. I find it hard to believe that this phone will be able to take images in a controlled lighting situation.
The fact is, many news stories and movies will lead you to believe that face recognition and detection work much better than they do.
This is missing the point to an extent. IBM does not care about Linux on the desktop or laptop. IBM sold its entire PC business because it no longer finds PCs to be as profitable. IBM will make money selling hardware and services in an 'on demand' style. To them, hardware and services are the commodities, not software. This is where Linux fits in.
This is no news in the AI community, algorithms such as GA's long ago learned how to add numbers, etc. I won't even go into detail here as I assume most readers are aware of this
This story is merely a case of someone who is excited about their work explaining it to an author who doesn't know as much about the subject matter. The author then turns around and writes a story for the lay-person who is not versed in the field. These people in turn jump to humorous conclusions.
This is a common occurance in magazines such as Discover and Popular Science, as much as I enjoy them. A good example is stories on robots, such as Honda's ASIMO. People see ASIMO do amazing things and assume that in 10-15 years we will have these robots in our homes. What the articles often fail to mention is that while ASIMO can do complex tasks, it has very limited ability to recognize a situation, such as a staircase in front of it, and decide on a course of action to take, such as executing its stair climbing procedure.
The true point of the article is that AI algorithms can teach us things about evolution. To make grand jumps and assume that these programs are even in the same playing field as SkyNet or the Matrix is to miss the main point.
As I said above, this is merely the case of a complex subject being explained in a way that is easy to digest for the masses. Even someone who had only taken a few graduate AI courses would find that many misguided statements are made in the article.
Pepsi pays for the Pepsi cooler and owns it. They can determine if they wish for coke to be placed in it. Pepsi does not pay to be listed on google when you search for 'pepsi', and thus your analogy is poor.
Because on a laptop, you can control-click with one hand. After all, they tend to have nice, small keyboards. The control button can be pushed with the pinky, the mouse pad with the index finger, and the mouse button with the thumb. IBM Thinkpads allow the user to tap the mouse pad, an action that is the same as clicking the mouse button located a mere centimeter below. This allows for one-finger mouse manipulation, instead of two. I don't hear anyone crying out for Macintosh to take on this functionality for the laptops. While I'm at it, where's the little red nipple?
Has anybody actually thought about how this would really work? As a researcher and developer of face detection and face recognition technology, I have seen the top of the line algorithms and programs developed by some of the top companies in this field, and this sounds a little optimistic. The system would require that your camera had software on it that could first detect all of the human faces in the image and then identify the correct person's face to blur. Although the top algorithms can do very impressive things in very specific environments, they do not perform as well as movies and news stories would have us think.