Chess is nowhere near that simple. Counting material is a very poor evaluation function unless you also somehow count position. I've seen games where the winning move was to sacrifice a queen for a pawn. But counting material is a QUICK evaluation function, so you can cut off most loosing lines by gross material disparity n-moves in the future. This will occasionally cause you to prune a winning line, but not often. However as you get closer to the current board position material difference becomes an increasingly poor evaluation function, and position becomes increasingly important. But there is no quick evaluation function for position, so you can't use it to prune deep searches, as there are too many nodes.
I think you're wrong. The problem is that current computers are vastly weaker than the brain at problems that a neural network is well adapted to learning. A secondary problem is deficiency in sensors and manipulators.
But what's really missing in the current AIs is a deep motivational stack. They are currently operating with a very shallow heap of motivations. E.g., if you were to ask Alpha Go why it bothered to play go, and it could even understand the question, it wouldn't be able to tell you. True, ask any go player, and the answer you get is generally a false statement due to lack of insight, but Alpha Go couldn't reach to that level.
Maybe. But I still wish Java had non-object structs. I like being able to save binary images to disk without a bunch of serialization...and pull them back without a bunch of deserialization.
For that matter, if they're going to add so many features into the language, why don't they add a persistent storage B+Tree. I rarely need or want SQL, but a built in B+Tree would be immensely useful. And I don't mean one elaborated the way libdb (SleepyCat) is...more like the way it was, only built into the language so it could automatically handle non-reference based data. C has a good reason for not including that, it's emphasis is minimalism. C++....well, there are decent libraries you can use, but I'm not convinced. Still, they put their emphasis on libraries with minimal-to-no runtime overhead. Java doesn't have those excuses. (OK, originally it was aimed as an embedded language, and then it was going to be a language that ran totally in browsers...but when they gave up on those choices...)
All developers don't need, or get access to, all of the code. So that argument fails. You would, of course, need to move many of them. I'm sure the EU would object to acquiring a lot of well paid tax payers and a company that generates huge amounts of cash.
The thing is, it would be a quite expensive proposition for Apple. I'm rather sure they COULD do it, but I'm even surer that they wouldn't.
Perhaps the Cayman Islands, then? Or make a deal with both Russia and the Ukraine to let them buy the Ukraine?
But I'm not sure that just because the EU recognizes US banking/taxation laws it would automatically enforce a case that was clearly unfounded. Particularly when the case against enforcement was supported by a powerful corporation. You might, of course, need to move all development out of the US. But the TPP and an equivalent treaty in process with Europe are going to make import taxes expensive to the imposing country. And the process being set up for trials isn't friendly to governments being charged by a corporation.
While true, that's not useful, because of the contrafactual if. They can include one time pads in the ROMs, but since it's a mass market device, they will be accessible by anyone with the purchase price and some expertise.
They could do something nearly equivalent, but it would require an extreme amount of effort, and so, as I said, I can't imagine them doing it.
It's one thing to make a received communication undecipherable, it's another to protect something in transit. And even in the first case if nobody knows the key to unlock the code, then you can't retrieve the message yourself.
Now streaming communication can be secure AFTER then handshake, but if someones recording everything, they are also recording the handshake, and so they know the key that's been decided upon.
Foolproof encryption is impossible if there is physical access, and if it's desirable that somebody ever be able to read the message. You can do encryption that can't be cracked without the key, but coercion of key access is not an unknown part of the process. (OK, in this case the person you'd want to coerce it from is dead, but that may be equivalent to "you don't want anyone to ever be able to retrieve the message", which is one of the workable cases.)
The problem with one-time pads is that they depend on a shared secret, and if you overshare it, then it's no longer a secret. And no algorithm can actually be equivalent to a one time pad, much less be able to generate the precise same series of numbers as that on another machine. If you use a deterministic algorithm, then it's not random and is potentially crackable. If you don't, then you can't get the same set of random numbers as your partner. The only way around this is out-of-band communication.
Any country cheap enough that they could buy the entire thing, government included. Apple could probably afford any of several countries, but Luxemborg would have the advantage of being a member of the EU, and thus hard to act against.
That's not clear. It would certainly be a lot of trouble, but they could set it up so that new model Apple-Apple communication over the networks (including cell phone) would be unbreakable. But it would be a LOT of trouble, and I can't imagine them bothering to do so. The metadata would still be obvious, of course.
The real weakness of the current system is that if you record the initial handshake which establishes the session key, then it is *relatively* easy to decrypt things, even with otherwise secure encryption. And you're going to need to factor in that the NSA is known to be working on quantum computers. (What success is unknown, so you've got to assume success.) This means that current approaches aren't useful even with longer keys. You need something else (and I'm no expert). It's made more difficult because you can assume the feds will buy and study any mass-market device.
The constitution pretty nearly explicitly says "Anything that isn't listed the federal government is forbidden from doing." So if you say the constitution is silent on the matter, you're saying the feds are forbidden.
That's true. The other day I saw someone "jaywalking" diagonally across the street in a motorized wheel chair while texting. That's a danger that didn't exist previously. I'd say they were going diagonally through the intersection, but actually they were just west of the intersection, where a car turning couldn't see or expect them.
The shady business practice of Microsoft continue in the present. They are collecting legal payoffs from nearly every Android phone sold for "secret patents", i.e., they say they have patents, but they won't tell you what the patents are.
Then there's the stuffing of the delegation to various standards committee meetings so that no opposing parties could speak. And several other equally immoral and unethical practices within the last decade.
It's true that Balmer has been replaced, but early reports don't speak any more highly of his replacement.
In this particular case it's their shady legal maneuvering that causes me to be dubious of them. You may have a case that's a clear win, but if you can't afford to present it that doesn't do you much good against a well-heeled heel.
Yeah, but those are what became (in the 1950's) called androids.
However robot did used to mean autonomous machine (again, in the 1950s). So C3PO and R2D2 were later called robots by that meaning. Unfortunately telefactor is too clumsy a word for newspapers, and the media are notorious for not respecting fine distinctions of meaning. So currently robot is almost without an understandable meaning. You know that it means either some sort of machine, or something acting as the speaker presumes a machine would act, and that's about it.
If sonar pings "do not tell you what is actually there, it just gives you a theory", then you can say the same thing about vision and touch. A good sonic image can be more detailed than a visual scan in cloudy water. So if looking at it counts, then so should sonic imagery. And nobody has mentioned detail. Which is a gradation with no sharp drop-offs or edges.
As for touch... if I touch it with a telefactor, have I touched it? What if the telefactor is purely a kinesthetic sensor? Why would that count different from listening to it? Etc.
There are a lot of places where the existence of sharp boundaries is an artifact of a limited number of ways of action or perceiving. "Light tweesers" are a proof that merely illuminating something is to slightly move it. (There are other proofs.) So if you measure closely enough you can't perceive something (except by depending on ambient differences) without moving it around. And there is a smooth gradation from illumination all the way to robot shovels and beyond.
Laws are the process of creating artificial distinctions that don't exist in the physical universe. The universe doesn't have many sharp distinctions, but people are so formed as to perceive them, because the in-between areas are either quite unusual or infrequent. I'm considering the density variations of the air between two people to be an example, and I'm having a hard time describing in words something which is obvious from the lack of vacuum. There's a decrease in density over a small distance as you pass through your skin into the air. It's a smooth transition over a very short distance. Then it's relatively un-dense through the air until close approach to the other person where there is another smooth change in density over a short distance of their skin.
If you look closely enough at the physical universe you don't find sharp distinctions. But those distinctions are very important to entities living at the scale that humans live and living as the velocities we commonly use. Laws should clarify those bounds in a way useful to humans, just as language does, and as casual thinking does. We live our lives at the human scale.
So. Telefactors, robots, etc. If a robot (not a telefactor) discovers something, why is that different than if a dog discovers something? You are responsible for what your dog does (within limits), but you don't get an automatic claim to what he discovers. You have to act. A telefactor, however, is closer to a (non-autonomous) car. The problem is that it isn't you, but it's acting as a representative of you. So perhaps it should be considered as your agent, and the laws should interpret it as your agent?
Yes, the personnel at MS has changed. The corporate policy has changed. But there's no believable evidence to indicate that they've become someone to become dependent on...and considerable evidence that it's still a bad idea.
N.B.: I'm not claiming this as an example of it happening. This just happened automatically as the system changed and the old programs didn't. Any code that isn't actively maintained is likely to eventually suffer from this problem.
Sorry, but have you ever tried to retrofit old code to work with a new set of libraries. It can almost always be done, but it is often quite difficult. And it's quite possible to write patents that apply, or appear to apply, to all the obvious feasible ways of doing so. This has worked in the past. The only tricky part is to get the system dependent on your changes for two or three versions before you start noticed enforcement actions.
I've got several of old programs that won't work on Linux after about 2.2. (I'm vague because I've never bothered to check out the precise cut-off. But I have a VM installed to run them on.)
They've said that too often before, and every time they've lied. To trust them this time would be foolish.
That said, sometimes they do offer a hook with decent bait that it's possible to take without getting caught. Which is where not trusting them comes in. Be sure the bait is worth the risk, and avoid getting hooked...or poisoned.
Change some of the code, wait a couple of years, file patent infringement claims.
Well, I haven't read the Eclipse license, so that might not work, but it's of a piece with what they've done before. So perhaps a modification of that, where Eclipse usability becomes dependent on a Microsoft plug-in that isn't a part of their donated code.
My feeling about MS is, "If it's a sleazy trick, then MS is either working on it or kicking themselves for not thinking of it first.".
A few years ago this would have been a foolish idea, but now electrodes can be permanently attached to neurons with no apparent damage, so it becomes a lot more reasonable. And there's one style of them that can spread out and connect to a bunch of different neurons in the same area.
So a lot of the enabling technologies have already been developed. I must admit that I don't expect this to work, but it's no longer a totally ridiculous idea.
But it wasn't original with him, he just took it seriously. E.E.Smith may be the first to write about it in "Skylark of Valeron" around 1946. Unless you want to include E.M.Forrester "The Machine Stops", but it wasn't clear what was going on there.
Depending on where you live. Some places have noticeably higher or lower background radiation level. Denver is supposed to have a background level of 6 mSv/year or higher.
Slashdot Mirror
Chess is nowhere near that simple. Counting material is a very poor evaluation function unless you also somehow count position. I've seen games where the winning move was to sacrifice a queen for a pawn. But counting material is a QUICK evaluation function, so you can cut off most loosing lines by gross material disparity n-moves in the future. This will occasionally cause you to prune a winning line, but not often. However as you get closer to the current board position material difference becomes an increasingly poor evaluation function, and position becomes increasingly important. But there is no quick evaluation function for position, so you can't use it to prune deep searches, as there are too many nodes.
That said, it's simple compared to go.
I think you're wrong. The problem is that current computers are vastly weaker than the brain at problems that a neural network is well adapted to learning. A secondary problem is deficiency in sensors and manipulators.
But what's really missing in the current AIs is a deep motivational stack. They are currently operating with a very shallow heap of motivations. E.g., if you were to ask Alpha Go why it bothered to play go, and it could even understand the question, it wouldn't be able to tell you. True, ask any go player, and the answer you get is generally a false statement due to lack of insight, but Alpha Go couldn't reach to that level.
Maybe. But I still wish Java had non-object structs. I like being able to save binary images to disk without a bunch of serialization...and pull them back without a bunch of deserialization.
For that matter, if they're going to add so many features into the language, why don't they add a persistent storage B+Tree. I rarely need or want SQL, but a built in B+Tree would be immensely useful. And I don't mean one elaborated the way libdb (SleepyCat) is...more like the way it was, only built into the language so it could automatically handle non-reference based data. C has a good reason for not including that, it's emphasis is minimalism. C++....well, there are decent libraries you can use, but I'm not convinced. Still, they put their emphasis on libraries with minimal-to-no runtime overhead. Java doesn't have those excuses. (OK, originally it was aimed as an embedded language, and then it was going to be a language that ran totally in browsers...but when they gave up on those choices...)
All developers don't need, or get access to, all of the code. So that argument fails. You would, of course, need to move many of them. I'm sure the EU would object to acquiring a lot of well paid tax payers and a company that generates huge amounts of cash.
The thing is, it would be a quite expensive proposition for Apple. I'm rather sure they COULD do it, but I'm even surer that they wouldn't.
Perhaps the Cayman Islands, then? Or make a deal with both Russia and the Ukraine to let them buy the Ukraine?
But I'm not sure that just because the EU recognizes US banking/taxation laws it would automatically enforce a case that was clearly unfounded. Particularly when the case against enforcement was supported by a powerful corporation. You might, of course, need to move all development out of the US. But the TPP and an equivalent treaty in process with Europe are going to make import taxes expensive to the imposing country. And the process being set up for trials isn't friendly to governments being charged by a corporation.
While true, that's not useful, because of the contrafactual if. They can include one time pads in the ROMs, but since it's a mass market device, they will be accessible by anyone with the purchase price and some expertise.
They could do something nearly equivalent, but it would require an extreme amount of effort, and so, as I said, I can't imagine them doing it.
It's one thing to make a received communication undecipherable, it's another to protect something in transit. And even in the first case if nobody knows the key to unlock the code, then you can't retrieve the message yourself.
Now streaming communication can be secure AFTER then handshake, but if someones recording everything, they are also recording the handshake, and so they know the key that's been decided upon.
Foolproof encryption is impossible if there is physical access, and if it's desirable that somebody ever be able to read the message. You can do encryption that can't be cracked without the key, but coercion of key access is not an unknown part of the process. (OK, in this case the person you'd want to coerce it from is dead, but that may be equivalent to "you don't want anyone to ever be able to retrieve the message", which is one of the workable cases.)
The problem with one-time pads is that they depend on a shared secret, and if you overshare it, then it's no longer a secret. And no algorithm can actually be equivalent to a one time pad, much less be able to generate the precise same series of numbers as that on another machine. If you use a deterministic algorithm, then it's not random and is potentially crackable. If you don't, then you can't get the same set of random numbers as your partner. The only way around this is out-of-band communication.
You need a signing key, or you can't securely issue system updates.
Or did you think anyone could write code that was perfect the first time?
Any country cheap enough that they could buy the entire thing, government included. Apple could probably afford any of several countries, but Luxemborg would have the advantage of being a member of the EU, and thus hard to act against.
You missed that they also want Apple's signing key. That's the important part.
That's not clear. It would certainly be a lot of trouble, but they could set it up so that new model Apple-Apple communication over the networks (including cell phone) would be unbreakable. But it would be a LOT of trouble, and I can't imagine them bothering to do so. The metadata would still be obvious, of course.
The real weakness of the current system is that if you record the initial handshake which establishes the session key, then it is *relatively* easy to decrypt things, even with otherwise secure encryption. And you're going to need to factor in that the NSA is known to be working on quantum computers. (What success is unknown, so you've got to assume success.) This means that current approaches aren't useful even with longer keys. You need something else (and I'm no expert). It's made more difficult because you can assume the feds will buy and study any mass-market device.
The constitution pretty nearly explicitly says "Anything that isn't listed the federal government is forbidden from doing." So if you say the constitution is silent on the matter, you're saying the feds are forbidden.
That's true. The other day I saw someone "jaywalking" diagonally across the street in a motorized wheel chair while texting. That's a danger that didn't exist previously. I'd say they were going diagonally through the intersection, but actually they were just west of the intersection, where a car turning couldn't see or expect them.
The shady business practice of Microsoft continue in the present. They are collecting legal payoffs from nearly every Android phone sold for "secret patents", i.e., they say they have patents, but they won't tell you what the patents are.
Then there's the stuffing of the delegation to various standards committee meetings so that no opposing parties could speak. And several other equally immoral and unethical practices within the last decade.
It's true that Balmer has been replaced, but early reports don't speak any more highly of his replacement.
In this particular case it's their shady legal maneuvering that causes me to be dubious of them. You may have a case that's a clear win, but if you can't afford to present it that doesn't do you much good against a well-heeled heel.
No, Asimov didn't go in much for humor. But it could be from a parody of such a book.
Yeah, but those are what became (in the 1950's) called androids.
However robot did used to mean autonomous machine (again, in the 1950s). So C3PO and R2D2 were later called robots by that meaning. Unfortunately telefactor is too clumsy a word for newspapers, and the media are notorious for not respecting fine distinctions of meaning. So currently robot is almost without an understandable meaning. You know that it means either some sort of machine, or something acting as the speaker presumes a machine would act, and that's about it.
If sonar pings "do not tell you what is actually there, it just gives you a theory", then you can say the same thing about vision and touch. A good sonic image can be more detailed than a visual scan in cloudy water. So if looking at it counts, then so should sonic imagery. And nobody has mentioned detail. Which is a gradation with no sharp drop-offs or edges.
As for touch... if I touch it with a telefactor, have I touched it? What if the telefactor is purely a kinesthetic sensor? Why would that count different from listening to it? Etc.
There are a lot of places where the existence of sharp boundaries is an artifact of a limited number of ways of action or perceiving. "Light tweesers" are a proof that merely illuminating something is to slightly move it. (There are other proofs.) So if you measure closely enough you can't perceive something (except by depending on ambient differences) without moving it around. And there is a smooth gradation from illumination all the way to robot shovels and beyond.
Laws are the process of creating artificial distinctions that don't exist in the physical universe. The universe doesn't have many sharp distinctions, but people are so formed as to perceive them, because the in-between areas are either quite unusual or infrequent. I'm considering the density variations of the air between two people to be an example, and I'm having a hard time describing in words something which is obvious from the lack of vacuum. There's a decrease in density over a small distance as you pass through your skin into the air. It's a smooth transition over a very short distance. Then it's relatively un-dense through the air until close approach to the other person where there is another smooth change in density over a short distance of their skin.
If you look closely enough at the physical universe you don't find sharp distinctions. But those distinctions are very important to entities living at the scale that humans live and living as the velocities we commonly use. Laws should clarify those bounds in a way useful to humans, just as language does, and as casual thinking does. We live our lives at the human scale.
So. Telefactors, robots, etc. If a robot (not a telefactor) discovers something, why is that different than if a dog discovers something? You are responsible for what your dog does (within limits), but you don't get an automatic claim to what he discovers. You have to act. A telefactor, however, is closer to a (non-autonomous) car. The problem is that it isn't you, but it's acting as a representative of you. So perhaps it should be considered as your agent, and the laws should interpret it as your agent?
"They" is spokesmen for Microsoft.
Yes, the personnel at MS has changed. The corporate policy has changed. But there's no believable evidence to indicate that they've become someone to become dependent on...and considerable evidence that it's still a bad idea.
N.B.: I'm not claiming this as an example of it happening. This just happened automatically as the system changed and the old programs didn't. Any code that isn't actively maintained is likely to eventually suffer from this problem.
Sorry, but have you ever tried to retrofit old code to work with a new set of libraries. It can almost always be done, but it is often quite difficult. And it's quite possible to write patents that apply, or appear to apply, to all the obvious feasible ways of doing so. This has worked in the past. The only tricky part is to get the system dependent on your changes for two or three versions before you start noticed enforcement actions.
I've got several of old programs that won't work on Linux after about 2.2. (I'm vague because I've never bothered to check out the precise cut-off. But I have a VM installed to run them on.)
They've said that too often before, and every time they've lied. To trust them this time would be foolish.
That said, sometimes they do offer a hook with decent bait that it's possible to take without getting caught. Which is where not trusting them comes in. Be sure the bait is worth the risk, and avoid getting hooked...or poisoned.
Change some of the code, wait a couple of years, file patent infringement claims.
Well, I haven't read the Eclipse license, so that might not work, but it's of a piece with what they've done before. So perhaps a modification of that, where Eclipse usability becomes dependent on a Microsoft plug-in that isn't a part of their donated code.
My feeling about MS is, "If it's a sleazy trick, then MS is either working on it or kicking themselves for not thinking of it first.".
A few years ago this would have been a foolish idea, but now electrodes can be permanently attached to neurons with no apparent damage, so it becomes a lot more reasonable. And there's one style of them that can spread out and connect to a bunch of different neurons in the same area.
So a lot of the enabling technologies have already been developed. I must admit that I don't expect this to work, but it's no longer a totally ridiculous idea.
It though more of "True Names" by Vernor Vinge.
But it wasn't original with him, he just took it seriously. E.E.Smith may be the first to write about it in "Skylark of Valeron" around 1946. Unless you want to include E.M.Forrester "The Machine Stops", but it wasn't clear what was going on there.
Depending on where you live. Some places have noticeably higher or lower background radiation level. Denver is supposed to have a background level of 6 mSv/year or higher.
And when AIs start objecting we'd need to write she/he/it.