It's not that direct, and I see no reason that a recurrent network couldn't learn "common sense". (Well, at least as well as people can.) But if you need to include all the information involved in acquiring common sense, then you've radically increased the data requirements, including lots of time-series data sets, etc. The cheap way to do this is probably to embody it in a body with lots of sensors. Now one source of this information would be a fleet of automated cars...
The problem is that this depends on storage capacities continuing to increase at the same rate they have been, and processor capabilities doing the same. (OTOH, for neural nets you don't need more capable CPUs, as distributed nets work quite well, and solve a lot of the cooling problems.)
So it's doable. Whether it will be done is another question. It's generally cheaper to use the lowest amount of computing power that's needed to do the job. (Generally. Cellphones are an excellent counterexample, however.)
It would also help to require that they not have been proven to have been doing unethical work during the past, say, five years. (I didn't say illegal, I said unethical. Unfortunately, that makes the term "proven" a bit difficult to define. Also the term unethical. So you'd need to set down certain minimum requirements that would substitute.)
I think you're missing the point. The point is not to do an accurate reconstruction of missing information, but to do a plausible one. If the data is missing, this approach cannot regenerate it.
He didn't say encrypted, but rather cryptographically signed. That's essentially a modified hash of the picture, or well-defined parts of the picture. It generally wouldn't be visible if you were looking at the picture.
OTOH, I also doubt that the general run of cameras used for security do this signing. I suspect that he had a specific variety (not brand) of camera that he was thinking of when he said "security camera" which is a lot different from the normal web-cam that is mounted on a porch.
Well...not exactly back to normal. The faker bots will improve their ability to fool people into thinking they're real.
Also, the intent is probably impossible for even a superhuman AI to accomplish (except by judging something like volume of posts, which ordinary recipients don't have access to). A twitter post often doesn't contain enough information to decide whether it was posted by a human or by a bot. As the faker bots improve, they'll be able to handle longer segments of connected text, and possibly to ever respond reasonably. (Eliza, Parry, Doctor, etc. show that ordinary human responses are often shallow enough to easily fake. And Eliza was *supposed* to be a counter-example.)
The actions of his that I have noticed do not signify someone who's trying to get rid of or minimize the state, but rather someone who's trying to use it to his own advantage.
I may well claim to be a libertarian, I haven't read anything he's written, but that's not what the actions of his that I have notices show. What actions has he taken that cause you to believe he is trying to eliminate or minimize the state rather than to use it to his own advantage?
IIRC, Mr. Thiel went to Stanford which is where it hooked up with some friends or acquaintances and together built a company that got him that fortune. So he's not exactly a new import into Silicon Valley.
So some, unknown fraction, of the data can be validated as accurate, the those who are offering the data know in advance what can be validated.
I'm sorry, but to me this says "You can split the data into two chunks. One chunk you can validate, and the other you can't. We know ahead of time which pieces you can validate, but trust us, the rest is honest."
Please note, this doesn't mean I believe they're lying this time. It just means I don't trust the protocol. (I actually have no opinion about whether they're lying this time. This could just be setting a protocol up for later abuse after it's been accepted. Or it really could be that they don't understand why they aren't trusted.)
The problem isn't that learning to program is getting harder, it's actually slightly easier. But the approach to teaching it is wrong. If you want to teach the mass of students (a bad idea, by the way) then you need to introduce them to programming with something like Scratch, or Logo and turtles. The assembler level is totally the wrong place for anyone these days. An IBM 7094 or Z80 or I6502 was relatively simple. Even the M68000 wasn't too bad. But modern processors are just to complex to be a reasonable starting point.
Now for a motivated subset of students Python or C (or Ruby or, perhaps, Go if there's ever a decent introduction) is reasonable. C will give them a better understanding of how computers work, but Python will let them get interesting results faster. It's a trade-off. If you could get anyone to use it MIXX would be a good place to start, but that's going to require a lot of external incentive. If you want to really understand the basis of programming, build a FORTH or Lisp interpreter in C. But that still won't introduce you properly to concurrency, unicode handling, graphics, or even objects. Programming is a lot wider now than it used to be, and it takes a lot longer to master...most people never do master all of it. I'm really weak on graphics programming. (Well, I started on Fortran IV, and even character strings were strange. To me they were fixed length byte arrays, but just try to map than onto a Python3 string.)
So there's several issues. One thing is picking the right entry point. This has to be varied with the student. Another is limiting your expectations. Very few kids, perhaps none, are going to master all of programming. And the ones who do will spend a decade doing it. But they can learn to handle particular areas fairly quickly.
One nice book for that that I ran across recently was about constructing mazes in Ruby. It used libraries to make the graphics simple, and focused on the maze algorithms. That would be suitable after they already had some basic knowledge.
Oops! I missed that part. I'd been thinking this was a reasonable message, but if it's supposed to cover posts to Slashdot as well as the FreeBSD mailing list, then it becomes totally over the top.
There are pieces of that "code of conduct" where I'd say it was reasonable to expel anyone who violated it anywhere from the community, but most of it seems as if it should only apply to the managed lists.
I think they should have replaced them all with "Don't be abusive to people, either verbally or in person, or through agents."
Even that's a bit overly specific. Perhaps just "Don't post comments abusive towards people. P.S.: ALL threats count as abusive. (Being abusive towards code is fine.)"
You've never tried to program a robot, have you. There's a reason that we switched largely to using neural nets, and that's because it really *is* impossible to be specific.
I know that there are lots of hidden definitions underneath everything you say, but those definitions are themselves ambiguous. And the law is ambiguous enough that different courts will reasonable decide essentially identical cases differently. Not in every case, but there are always edge-cases. Someone once said "hard cases make bad law" when talking, I believe, about precedent. But even that's an oversimplification.
That's one reason why there aren't yet robot lawyers and judges. (Another reason, of course, is the lawyers labor union. For a lot of cases robot lawyers and judges could do perfectly well, but even for those they aren't allowed.)
The thing that's both promising and troubling about this "code of conduct" is the "at the discretion of the FreeBSD Code of Conduct Committee" part. With good governance that could make this a very beneficial code, but with bad governance that could make it extremely oppressive.
My first take, however, was that none of those subjects are appropriate on a technical discussion list anyway, whether you are lauding someone or denigrating them.
And an interesting question is whether this response indicates that at that time McConnell was aware that the Russians were intervening in favor of the Republicans. It is, of course, far from proof, but it's certainly suggestive.
How are you supposed to decide if this is real data? I think you need evidence of it's provenance before it's worth anything. Do you believe it because Microsoft's NBC says that Twitter says that it's accurate?
The problem is that mainstream news *is* fake. Which doesn't make the social media garbage any more reliable. There aren't any reliable sources of news. Local news, from local sources, is often biased in a predictable way, and that's the best you can do. If 90% of their audience can't check what they're reporting, they feel free to get totally creative.
OK, I exaggerate. But not by that bloody much. And you can't even tell which direction they're spinning things, since it generally seems to turn more on which will get them most views rather than which is most accurate, or most favored by some ideologue. And if you're given a choice between that and some propaganda put out by an ideologue, which are you going to believe, and why?
You did say "Alien Life", not advanced technology alien visitors.
If they were "advanced technology alien visitors", then given no more information than that they exist, I'd probably be more fearful than welcoming. If you want to know why you could ask the Aztecs. (OTOH, the neighboring tribes that weren't Aztecs *might* have a different opinion.) Or you could ask the native Hawaiians.
Now these analogies aren't exact, but they are legitimate causes for trepidation.
OTOH, they might help us get safely through the construction of super-human AI. That looks to me like a major bottleneck ahead of us, and is likely a part of the "great filter". As is, of course, not developing a super-human AI. Out current governments are in the long term suicidal.
You're assuming that the goal is to come to the same (correct) result each time, but with lots of AI programs the goal is to come up with *some* correct result each time, and their use case is generally in places where you can't define one particular result as correct, though you may be able to define a lot of results as wrong, e.g., finish the sentence "My love is like..." Clearly one possible answer is " a red, red, rose", and clearly " a rutabaga" would need a strange context to be a correct answer. But how would you evaluate " a willow wand"? Many would think that a fine continuation. (I've never been sure why "a red, red, rose" is accepted as a reasonable answer, but Robert Burns wasn't wrong about it being a good completion. And Google gives lots of other weird completions that are also accepted as reasonable, at least in some contexts. ["a candle"???])
This kind of problem doesn't have a correct answer, just wrong ones and a bunch of varying acceptability. And what answers are acceptable can depend a lot on context.
(Please note, the prior paragraph is the description of the variety of problem. Complete the sentence was an example, not a defining epitomization. But its the one that came to mind, and it was easy to describe.)
The thing is, their code wouldn't suffice. You also need the training data set, the order in which the data was presented, the rewards issued, etc.
Even then, a lot of AI programs have a (pseudo) random element in them, so you wouldn't get the same results twice. Unless you used the same seed each time, which would rather defeat the purpose of the random number generator, as that's often supposed to allow you to generate a range of responses that are selected from, so it doesn't look deterministic.
It's been a couple of decades now, so this may be obsolete, but in my experience while CygWin might not break, Linux applications running on it sometimes depended on things that didn't work the same way. It's quite plausible that this is still true.
Did you notice that the identification of who was being wiretapped was hash coded? You can't easily tell who the target it, but it can be easily validated that the target was the one specified.
The advantage is it makes it a bit harder to just go to one judge who always rubberstamps everything "ok". It's not a great improvement, but it's a bit of one. I just couldn't think of anything that would really mean they had to actually show cause.
Slashdot Mirror
It's not that direct, and I see no reason that a recurrent network couldn't learn "common sense". (Well, at least as well as people can.) But if you need to include all the information involved in acquiring common sense, then you've radically increased the data requirements, including lots of time-series data sets, etc. The cheap way to do this is probably to embody it in a body with lots of sensors. Now one source of this information would be a fleet of automated cars...
The problem is that this depends on storage capacities continuing to increase at the same rate they have been, and processor capabilities doing the same. (OTOH, for neural nets you don't need more capable CPUs, as distributed nets work quite well, and solve a lot of the cooling problems.)
So it's doable. Whether it will be done is another question. It's generally cheaper to use the lowest amount of computing power that's needed to do the job. (Generally. Cellphones are an excellent counterexample, however.)
It would also help to require that they not have been proven to have been doing unethical work during the past, say, five years. (I didn't say illegal, I said unethical. Unfortunately, that makes the term "proven" a bit difficult to define. Also the term unethical. So you'd need to set down certain minimum requirements that would substitute.)
I think you're missing the point. The point is not to do an accurate reconstruction of missing information, but to do a plausible one. If the data is missing, this approach cannot regenerate it.
He didn't say encrypted, but rather cryptographically signed. That's essentially a modified hash of the picture, or well-defined parts of the picture. It generally wouldn't be visible if you were looking at the picture.
OTOH, I also doubt that the general run of cameras used for security do this signing. I suspect that he had a specific variety (not brand) of camera that he was thinking of when he said "security camera" which is a lot different from the normal web-cam that is mounted on a porch.
Well...not exactly back to normal. The faker bots will improve their ability to fool people into thinking they're real.
Also, the intent is probably impossible for even a superhuman AI to accomplish (except by judging something like volume of posts, which ordinary recipients don't have access to). A twitter post often doesn't contain enough information to decide whether it was posted by a human or by a bot. As the faker bots improve, they'll be able to handle longer segments of connected text, and possibly to ever respond reasonably. (Eliza, Parry, Doctor, etc. show that ordinary human responses are often shallow enough to easily fake. And Eliza was *supposed* to be a counter-example.)
Not:
I may well claim to be a libertarian (I used to, but now I'm a minarchist),
but rather:
He may well claim to be a libertarian
The actions of his that I have noticed do not signify someone who's trying to get rid of or minimize the state, but rather someone who's trying to use it to his own advantage.
I may well claim to be a libertarian, I haven't read anything he's written, but that's not what the actions of his that I have notices show. What actions has he taken that cause you to believe he is trying to eliminate or minimize the state rather than to use it to his own advantage?
IIRC, Mr. Thiel went to Stanford which is where it hooked up with some friends or acquaintances and together built a company that got him that fortune. So he's not exactly a new import into Silicon Valley.
So some, unknown fraction, of the data can be validated as accurate, the those who are offering the data know in advance what can be validated.
I'm sorry, but to me this says "You can split the data into two chunks. One chunk you can validate, and the other you can't. We know ahead of time which pieces you can validate, but trust us, the rest is honest."
Please note, this doesn't mean I believe they're lying this time. It just means I don't trust the protocol. (I actually have no opinion about whether they're lying this time. This could just be setting a protocol up for later abuse after it's been accepted. Or it really could be that they don't understand why they aren't trusted.)
The problem isn't that learning to program is getting harder, it's actually slightly easier. But the approach to teaching it is wrong. If you want to teach the mass of students (a bad idea, by the way) then you need to introduce them to programming with something like Scratch, or Logo and turtles. The assembler level is totally the wrong place for anyone these days. An IBM 7094 or Z80 or I6502 was relatively simple. Even the M68000 wasn't too bad. But modern processors are just to complex to be a reasonable starting point.
Now for a motivated subset of students Python or C (or Ruby or, perhaps, Go if there's ever a decent introduction) is reasonable. C will give them a better understanding of how computers work, but Python will let them get interesting results faster. It's a trade-off. If you could get anyone to use it MIXX would be a good place to start, but that's going to require a lot of external incentive. If you want to really understand the basis of programming, build a FORTH or Lisp interpreter in C. But that still won't introduce you properly to concurrency, unicode handling, graphics, or even objects. Programming is a lot wider now than it used to be, and it takes a lot longer to master...most people never do master all of it. I'm really weak on graphics programming. (Well, I started on Fortran IV, and even character strings were strange. To me they were fixed length byte arrays, but just try to map than onto a Python3 string.)
So there's several issues. One thing is picking the right entry point. This has to be varied with the student. Another is limiting your expectations. Very few kids, perhaps none, are going to master all of programming. And the ones who do will spend a decade doing it. But they can learn to handle particular areas fairly quickly.
One nice book for that that I ran across recently was about constructing mazes in Ruby. It used libraries to make the graphics simple, and focused on the maze algorithms. That would be suitable after they already had some basic knowledge.
I know. For awhile I tried to get people to use UStatian, but it never caught on.
Oops! I missed that part. I'd been thinking this was a reasonable message, but if it's supposed to cover posts to Slashdot as well as the FreeBSD mailing list, then it becomes totally over the top.
There are pieces of that "code of conduct" where I'd say it was reasonable to expel anyone who violated it anywhere from the community, but most of it seems as if it should only apply to the managed lists.
I think they should have replaced them all with "Don't be abusive to people, either verbally or in person, or through agents."
Even that's a bit overly specific. Perhaps just "Don't post comments abusive towards people. P.S.: ALL threats count as abusive. (Being abusive towards code is fine.)"
You've never tried to program a robot, have you. There's a reason that we switched largely to using neural nets, and that's because it really *is* impossible to be specific.
I know that there are lots of hidden definitions underneath everything you say, but those definitions are themselves ambiguous. And the law is ambiguous enough that different courts will reasonable decide essentially identical cases differently. Not in every case, but there are always edge-cases. Someone once said "hard cases make bad law" when talking, I believe, about precedent. But even that's an oversimplification.
That's one reason why there aren't yet robot lawyers and judges. (Another reason, of course, is the lawyers labor union. For a lot of cases robot lawyers and judges could do perfectly well, but even for those they aren't allowed.)
The thing that's both promising and troubling about this "code of conduct" is the "at the discretion of the FreeBSD Code of Conduct Committee" part. With good governance that could make this a very beneficial code, but with bad governance that could make it extremely oppressive.
My first take, however, was that none of those subjects are appropriate on a technical discussion list anyway, whether you are lauding someone or denigrating them.
Gee...I'll have to think about switching to using a Mac. All the Twitter folks are going away.
And an interesting question is whether this response indicates that at that time McConnell was aware that the Russians were intervening in favor of the Republicans. It is, of course, far from proof, but it's certainly suggestive.
How are you supposed to decide if this is real data? I think you need evidence of it's provenance before it's worth anything. Do you believe it because Microsoft's NBC says that Twitter says that it's accurate?
The problem is that mainstream news *is* fake. Which doesn't make the social media garbage any more reliable. There aren't any reliable sources of news. Local news, from local sources, is often biased in a predictable way, and that's the best you can do. If 90% of their audience can't check what they're reporting, they feel free to get totally creative.
OK, I exaggerate. But not by that bloody much. And you can't even tell which direction they're spinning things, since it generally seems to turn more on which will get them most views rather than which is most accurate, or most favored by some ideologue. And if you're given a choice between that and some propaganda put out by an ideologue, which are you going to believe, and why?
You did say "Alien Life", not advanced technology alien visitors.
If they were "advanced technology alien visitors", then given no more information than that they exist, I'd probably be more fearful than welcoming. If you want to know why you could ask the Aztecs. (OTOH, the neighboring tribes that weren't Aztecs *might* have a different opinion.) Or you could ask the native Hawaiians.
Now these analogies aren't exact, but they are legitimate causes for trepidation.
OTOH, they might help us get safely through the construction of super-human AI. That looks to me like a major bottleneck ahead of us, and is likely a part of the "great filter". As is, of course, not developing a super-human AI. Out current governments are in the long term suicidal.
You're assuming that the goal is to come to the same (correct) result each time, but with lots of AI programs the goal is to come up with *some* correct result each time, and their use case is generally in places where you can't define one particular result as correct, though you may be able to define a lot of results as wrong, e.g., finish the sentence
"My love is like..."
Clearly one possible answer is " a red, red, rose", and clearly " a rutabaga" would need a strange context to be a correct answer. But how would you evaluate " a willow wand"? Many would think that a fine continuation. (I've never been sure why "a red, red, rose" is accepted as a reasonable answer, but Robert Burns wasn't wrong about it being a good completion. And Google gives lots of other weird completions that are also accepted as reasonable, at least in some contexts. ["a candle"???])
This kind of problem doesn't have a correct answer, just wrong ones and a bunch of varying acceptability. And what answers are acceptable can depend a lot on context.
(Please note, the prior paragraph is the description of the variety of problem. Complete the sentence was an example, not a defining epitomization. But its the one that came to mind, and it was easy to describe.)
The thing is, their code wouldn't suffice. You also need the training data set, the order in which the data was presented, the rewards issued, etc.
Even then, a lot of AI programs have a (pseudo) random element in them, so you wouldn't get the same results twice. Unless you used the same seed each time, which would rather defeat the purpose of the random number generator, as that's often supposed to allow you to generate a range of responses that are selected from, so it doesn't look deterministic.
It's been a couple of decades now, so this may be obsolete, but in my experience while CygWin might not break, Linux applications running on it sometimes depended on things that didn't work the same way. It's quite plausible that this is still true.
Did you notice that the identification of who was being wiretapped was hash coded? You can't easily tell who the target it, but it can be easily validated that the target was the one specified.
The advantage is it makes it a bit harder to just go to one judge who always rubberstamps everything "ok". It's not a great improvement, but it's a bit of one. I just couldn't think of anything that would really mean they had to actually show cause.