Someone else who COMPLETELY missed the point of the exercise. Believe it or not, the answer to "Before this hotel mogul's elbow broke through it, a Picasso he owned was worth $139 million; after, $85 million" is NOT "http://www.luxist.com/tag/metropolitan+museum+of+art". It is "Steve Wynn". Yes, you, a human being, can read the question and understand that you are supposed to be giving a person's name as the answer. And you, a human being, can read through the results the Google returned and find the name of the person who damaged the piece of art. Google is not doing any of that for you. THAT is the difference between Google and Watson. And if you think it takes just 'a tiny bit of code' to do those things you have no idea what you are talking about.
The purpose of the challenge behind Deep Blue was not just to play chess, it was to demonstrate that some problems could be solved by massively parallel processing, and that IBM had developed hardware and software to do that. Deep Blue was an IBM 9176, and IBM sold a lot of them (none of them used to play chess).
Similarly, the purpose of this challenge is not 'play Jeopardy!', but to demonstrate that IBM has software and hardware that can do natural language processing. Watson will probably never again play Jeopardy!, but the techniques learned by doing it will certainly show up in other applications.
General visual processing is not a trivial task. OCR of white block letters on a plain blue background in a known fixed position is a trivial task, and would have added no more than a few microseconds to the processing time.
Obviously Watson's configuration was chosen so that it could come up with the response in the amount of time it had available. If that time was shortened (by having to do OCR or something) they could have chosen a different configuration (more processors, etc). And do you really think having to OCR block letters on a clear background would have taken more than a few microseconds anyway?
The feature they talked about on Nova was that they tell Watson the CORRECT answer after someone gets it. They showed the example of a category that required you to name a month, but Watson answered the first question with something other than a month. The next two questions were answered by humans, who gave correct month answers. Then Watson answered a question with a month, having learned that was what was expected.
I don't think that would change anything. From watching the show, it appears that Watson is designed to have it's best answer in a fixed amount of time, and does not use more time even if available. For instance, on one of the Double Jeopardy questions (which is not time constrained), it very quickly came up with it's answers, but was not very confident in them. It did not use the additional time 'thinking'. So, slowing down the rate at which they fed it the clues would make that available processing time shorter, but they could probably overcome that by adding more processors, etc.
Yes, it is very different. Jeopardy requires an actual answer (question), often just one or two words. That means it not only has to search for the pertinent terms, but it also needs to understand exactly what is expected as an answer. Google returns a list of documents where YOU are likely to FIND the answer. And of course another very big difference is the whole 'appropriate query' aspect. Why should the humans have to make an appropriate query?
You are glossing over the important point that when you transfer the 50000 copies you are also transferring the license, which still says you can't sell them. And if you aren't transferring the license (as required by the license) then you are in violation of the license, and the 50000 copies are not 'lawfully made', because absent the license you aren't allowed to make any copies.
More like Costco offers a 'Costco Visa Card', which says has an opt-in feature that says 'Costco will look at the things you buy with this card to improve your Costco shopping experience'. Then Costco notices that a whole bunch of people are using the card to purchase product X at Sams, so they say 'hey, maybe we should sell product x'. Then the Sam's fanbois come out yelling about how Costco is 'stealing' stuff from Sams.
This all happened in December. When the experiment was ready, about 20 Google engineers were told to run the test queries from laptops at home, using Internet Explorer, with Suggested Sites and the Bing Toolbar both enabled. They were also told to click on the top results. They started on December 17. By December 31, some of the results started appearing on Bing.
How do you know they are 'collecting and parsing search result by google'? What if MS has a theory that says 'after a user does a search, any and all links visited in the next 5 minutes are assumed to be related to that search'? Sure, MAYBE the result came from Google. MAYBE it came from a user selecting a result from Google, then following a link on THAT page (which Google did not return as a result). MAYBE Google came up with nothing, so the user got the link from a friend.
This 'sting' is pretty poorly designed, as the whole purpose of it is to be able to make the claim that MS is 'stealing' results. Did they try any other methods of getting nonsense data into Bing, that didn't involve 'stealing' Google results (search Google for nonsense term, get no results, go to random URL, repeat a bunch of times...)?
They could do that. They could also dispense with the whole 'game' idea and just have a single box that reveals how much you won. But remember, the whole idea is to sell tickets. If all the numbers are hidden, no-one that has 'lucky numbers' or any such thing is going to buy the ticket if they can't see any numbers first.
The bar code on the back of the ticket (the one you can see) is just for selling the ticket/inventory type things. There is another barcode that has the scratch off coating on it that is scanned to determine if it is a winner.
The Facebook postings aren't used as evidence per se. They are used to get other testimony. For instance: is this your Facebook page? Did you make these posts? Were they true? Is this a picture of yourself and someone tagged as Joe Smith at a party? For my next witness, I call Joe Smith...
Kind of answered your own question didn't you? What chances do you think a plaintiff who refuses to testify has? Hopefully they are as close to zero as you can get.
First, these are civil cases, not criminal. There is no prosecutor and there is no guilty/not guilty.
Second, it is the DEFENSE that is using FB as evidence that the plaintiff is lying.
So, let's say I sue you for $10M because I fell on your property, and I claim I am stuck in bed 24x7 and can't enjoy life at all. Are you really going to say that you wouldn't use pictures I posted on Facebook of me dancing at a party as evidence that my life is not as bad as I am claiming?
Sure you can. However that may put a dent in your credibility with the jury. You would need to convince them that sure, you lie on your Facebook page (where you really have nothing to gain or lose), but you would certainly not lie now (when your side of the story is the one they simply must believe).
You can't jam signals for any reason, disclosed or not. However, NBC could decide to no longer be affiliated with the local TV station, thus 'depriving' you of access to NBC shows, and you would have no recourse.
First, "equality of rights" only applies for certain things (race, religion, color, sex). Secondly, it would not be up to FB to prove anything, they are the ones being sued. It is up to the plaintiff to 'prove' that the reason they were denied access is because of one of the above reasons, and not for some other reason.
I'm not sure that you have to understand why something works to show that it does work. For instance, if you look at the prescribing information for many drugs, you may see statements like 'the mechanism of this action is not known'. We can show that the drug does in fact reduce blood pressure. We can show that it is safe for humans. We can show that it causes these side effects. We have no idea why it reduces blood pressure.
It does make sense like it is. In your car analogy there were two crimes committed: car theft and murder. If they had evidence he took the car, and didn't have evidence that he was the murderer, it would indeed make sense to only charge him with car theft. For the case at hand, I don't know if a cop giving someone information is a crime or not, but even if it was, how would they prove it? She gave the info to a friend, at the friends request. Who would even know a crime occurred?
The fact that she was charged with unauthorized access of a computer and not some sort of breach of contract also makes sense. What if the person accessing the database was not a cop? A cop who is not authorized to do a search is no different from a person off the street, and if they both perform the same action they should both be liable for the same penalties.
There are no laws against 'hacking' or 'cracking'. The laws are against "unauthorized use of a computer". It doesn't matter what method you used to access the computer, just that it is not authorized. Just like unauthorized access to physical property can get you a 'breaking and entering' charge, even if the only breaking you did was to push open an unlocked door.
Slashdot Mirror
Someone else who COMPLETELY missed the point of the exercise. Believe it or not, the answer to "Before this hotel mogul's elbow broke through it, a Picasso he owned was worth $139 million; after, $85 million" is NOT "http://www.luxist.com/tag/metropolitan+museum+of+art". It is "Steve Wynn". Yes, you, a human being, can read the question and understand that you are supposed to be giving a person's name as the answer. And you, a human being, can read through the results the Google returned and find the name of the person who damaged the piece of art. Google is not doing any of that for you. THAT is the difference between Google and Watson. And if you think it takes just 'a tiny bit of code' to do those things you have no idea what you are talking about.
The purpose of the challenge behind Deep Blue was not just to play chess, it was to demonstrate that some problems could be solved by massively parallel processing, and that IBM had developed hardware and software to do that. Deep Blue was an IBM 9176, and IBM sold a lot of them (none of them used to play chess).
Similarly, the purpose of this challenge is not 'play Jeopardy!', but to demonstrate that IBM has software and hardware that can do natural language processing. Watson will probably never again play Jeopardy!, but the techniques learned by doing it will certainly show up in other applications.
General visual processing is not a trivial task. OCR of white block letters on a plain blue background in a known fixed position is a trivial task, and would have added no more than a few microseconds to the processing time.
Obviously Watson's configuration was chosen so that it could come up with the response in the amount of time it had available. If that time was shortened (by having to do OCR or something) they could have chosen a different configuration (more processors, etc). And do you really think having to OCR block letters on a clear background would have taken more than a few microseconds anyway?
The feature they talked about on Nova was that they tell Watson the CORRECT answer after someone gets it. They showed the example of a category that required you to name a month, but Watson answered the first question with something other than a month. The next two questions were answered by humans, who gave correct month answers. Then Watson answered a question with a month, having learned that was what was expected.
I don't think that would change anything. From watching the show, it appears that Watson is designed to have it's best answer in a fixed amount of time, and does not use more time even if available. For instance, on one of the Double Jeopardy questions (which is not time constrained), it very quickly came up with it's answers, but was not very confident in them. It did not use the additional time 'thinking'. So, slowing down the rate at which they fed it the clues would make that available processing time shorter, but they could probably overcome that by adding more processors, etc.
Yes, it is very different. Jeopardy requires an actual answer (question), often just one or two words. That means it not only has to search for the pertinent terms, but it also needs to understand exactly what is expected as an answer. Google returns a list of documents where YOU are likely to FIND the answer. And of course another very big difference is the whole 'appropriate query' aspect. Why should the humans have to make an appropriate query?
You are glossing over the important point that when you transfer the 50000 copies you are also transferring the license, which still says you can't sell them. And if you aren't transferring the license (as required by the license) then you are in violation of the license, and the 50000 copies are not 'lawfully made', because absent the license you aren't allowed to make any copies.
More like Costco offers a 'Costco Visa Card', which says has an opt-in feature that says 'Costco will look at the things you buy with this card to improve your Costco shopping experience'. Then Costco notices that a whole bunch of people are using the card to purchase product X at Sams, so they say 'hey, maybe we should sell product x'. Then the Sam's fanbois come out yelling about how Costco is 'stealing' stuff from Sams.
If the returned pages were, as you say, completely unrelated to the search terms, then why are multiple people clicking on them?
Um, from the article that started all this:
This all happened in December. When the experiment was ready, about 20 Google engineers were told to run the test queries from laptops at home, using Internet Explorer, with Suggested Sites and the Bing Toolbar both enabled. They were also told to click on the top results. They started on December 17. By December 31, some of the results started appearing on Bing.
How do you know they are 'collecting and parsing search result by google'? What if MS has a theory that says 'after a user does a search, any and all links visited in the next 5 minutes are assumed to be related to that search'? Sure, MAYBE the result came from Google. MAYBE it came from a user selecting a result from Google, then following a link on THAT page (which Google did not return as a result). MAYBE Google came up with nothing, so the user got the link from a friend.
This 'sting' is pretty poorly designed, as the whole purpose of it is to be able to make the claim that MS is 'stealing' results. Did they try any other methods of getting nonsense data into Bing, that didn't involve 'stealing' Google results (search Google for nonsense term, get no results, go to random URL, repeat a bunch of times...)?
They could do that. They could also dispense with the whole 'game' idea and just have a single box that reveals how much you won. But remember, the whole idea is to sell tickets. If all the numbers are hidden, no-one that has 'lucky numbers' or any such thing is going to buy the ticket if they can't see any numbers first.
The bar code on the back of the ticket (the one you can see) is just for selling the ticket/inventory type things. There is another barcode that has the scratch off coating on it that is scanned to determine if it is a winner.
The Facebook postings aren't used as evidence per se. They are used to get other testimony. For instance: is this your Facebook page? Did you make these posts? Were they true? Is this a picture of yourself and someone tagged as Joe Smith at a party? For my next witness, I call Joe Smith...
Kind of answered your own question didn't you? What chances do you think a plaintiff who refuses to testify has? Hopefully they are as close to zero as you can get.
First, these are civil cases, not criminal. There is no prosecutor and there is no guilty/not guilty.
Second, it is the DEFENSE that is using FB as evidence that the plaintiff is lying.
So, let's say I sue you for $10M because I fell on your property, and I claim I am stuck in bed 24x7 and can't enjoy life at all. Are you really going to say that you wouldn't use pictures I posted on Facebook of me dancing at a party as evidence that my life is not as bad as I am claiming?
Sure you can. However that may put a dent in your credibility with the jury. You would need to convince them that sure, you lie on your Facebook page (where you really have nothing to gain or lose), but you would certainly not lie now (when your side of the story is the one they simply must believe).
Of course inline pictures, frames, and blink were positives for the site developers. 'Do not track' not so much.
You can't jam signals for any reason, disclosed or not. However, NBC could decide to no longer be affiliated with the local TV station, thus 'depriving' you of access to NBC shows, and you would have no recourse.
First, "equality of rights" only applies for certain things (race, religion, color, sex). Secondly, it would not be up to FB to prove anything, they are the ones being sued. It is up to the plaintiff to 'prove' that the reason they were denied access is because of one of the above reasons, and not for some other reason.
I'm not sure that you have to understand why something works to show that it does work. For instance, if you look at the prescribing information for many drugs, you may see statements like 'the mechanism of this action is not known'. We can show that the drug does in fact reduce blood pressure. We can show that it is safe for humans. We can show that it causes these side effects. We have no idea why it reduces blood pressure.
It does make sense like it is. In your car analogy there were two crimes committed: car theft and murder. If they had evidence he took the car, and didn't have evidence that he was the murderer, it would indeed make sense to only charge him with car theft. For the case at hand, I don't know if a cop giving someone information is a crime or not, but even if it was, how would they prove it? She gave the info to a friend, at the friends request. Who would even know a crime occurred?
The fact that she was charged with unauthorized access of a computer and not some sort of breach of contract also makes sense. What if the person accessing the database was not a cop? A cop who is not authorized to do a search is no different from a person off the street, and if they both perform the same action they should both be liable for the same penalties.
If he abused his access privileges then it is not 'authorized use'. The laws are against unauthorized use.
There are no laws against 'hacking' or 'cracking'. The laws are against "unauthorized use of a computer". It doesn't matter what method you used to access the computer, just that it is not authorized. Just like unauthorized access to physical property can get you a 'breaking and entering' charge, even if the only breaking you did was to push open an unlocked door.