Yes, it's a step up from the current "Allow all/deny all/ask all" model, but IMHO, this is still useless. I may not want to reject all third-party cookies, just most of them. Sure, there's the Trusted Sites zone, but I don't consider any website worthy of the label "Trusted".
This is still very coarse-grained control. I can do a million times better than this with my nice little Junkbuster and get rid of banner ads while I'm at it. I don't see this change as really doing very much for privacy protection compared to existing free software. Sure, it's a nice gesture from MS, but I think that's all it is --- throwing us a bone so they can say "See? We're very concerned about privacy. Ignore that digital watermark we attached to all your Office documents."
I don't have the book because I probably wouldn't bother to buy a full Mindstorms kit, but what makes you say that a breadth-first search is necessarily bad? I can see how it's not much help for the Mindstorms guru who wants to know more, but for someone who's just starting out, a broad review (with specific examples) seems like a good thing.
I got started in the robot lab of a small liberal arts college. They didn't have a whole lot in the way of facilities; a few of their bots were expensive research jobs from Pioneer, but mostly it was a matter of students screwing parts together. Said parts mainly came from the more expensive catalog kits; in my last days they finally acquired a 3D printer. (Of course, you have to have some mechE background and CAD experience to use a 3D printer.)
If you want to do this academically, the easiest thing is to search your school, find out who has robots, and go work with those people. Be prepared for an enormous amount of frustration; academic robotics research is mainly the study of an infinite number of Things That Don't Work.
If what you want is home robotics, the Mindstorms kit may be an option. However, you may have been looking for a cheaper solution. Lego is still about the best solution (IMHO) for the body of a home hobbyist's robot, because it's fairly cheap and very easy to redesign as needed. Buy a big box of Lego Technic and order spare connectors as your cat eats them or they get lost in the couch. There are many alternatives to Mindstorms for controllers; many people swear by the BASIC stamp, for example. This is a case where you're going to have to search the web and catalogs to find some parts within your price range. The more soldering and splicing and assembly you're willing to do, the better the price you can get.
Note that for "advanced" stuff you will also need a workbench which lets you drill and cut and shape parts with reasonable accuracy. You're also likely to need some electronics supplies --- a soldering iron and a standard set of resistors at the very least. The initial outlays here could also eat into your budget.
Your situation sounds like that of a coder who'd like to play with robots but really doesn't have much in the way of dedicated workspace, tools, or know-how. Given your claimed lack of electronics skills, you probably want to avoid kits until you're more comfortable. In your situation, I'd start with Mindstorms: you get plenty of tutorial advice and can even start working with a friendly little GUI at first. You might pick up the O'Reilly book on Mindstorms to help you out. That setup should give you a flavor for what robot-building is really like and should cost you $200 maximum. It'll take a while to exhaust all the capacities of a mildly hacked Mindstorms set (the most you'll need is extra Lego or sensors), and by then you'll be beyond the realm of Ask Slashdot.
I can see the tree stuff; those are interesting examples, though they make my brain hurt to read them (I'm of the opinion that good Scheme is liberally interspersed with comments that help you to know what block you're currently in; after about the third lambda I lose track.)
As far as the list-multiplier goes, though, wouldn't it be just as simple to say
Y'know, I earned my bachelor's in CS a year ago, TA'ed two functional-language-based classes, and wrote extensions to Scheme that made it useful for mobile agent programming, and I still don't know what call/cc is for. When the hell do you actually need something like that?
Could somebody explain to me why it's a good idea to layer this on top of HTTP? As far as I can see, it doesn't really solve the firewall problem; the article specifically says they'll run on different ports so that admins can firewall all IPP traffic. Last I checked, HTTP actually used a fairly inefficient file-transfer mechanism. Given that the modern corporate document is filled with images and charts and sundry multimedia (and therefore bloats pretty quickly), wouldn't it make more sense to layer this on top of, say, FTP? Why is it layered on top of anything at all? What does HTTP provide that IPP needs?
I'm sure there are some answers to these questions, because I believe in the IETF and I personally have only ever designed toy protocols, but I don't see the answers in the article.
"Once modified, the artificial chromosome can be duplicated hundreds of thousands of times. The company says that the amount of protein produced rises in step with the number of gene copies..."
I intially read this as them doing chromosome amplification within a single cell. (Such a thing is not unheard of; it happens in certain cancers, although it's usually not a whole chromosome being amplified.) On reparse, I think you're right; it's more a matter of them inserting extra copies. IMHO, that's not the right solution; you can't regulate that very well. I'd say it'd be a better idea to insert strong promoter elements along with the new genes if you wanted them to be hyperproductive. (That way, if the gene temporarily needs to be turned down, you can administer a compound which inhibits the activator protein (although that may have issues depending on what other genes share that promoter).)
Biologists have been making bacterial artificial chromosomes, or BACs, for years. BACs were a key component of the Human Genome project. They're conceptually quite simple; you take any ordinary chromosome and yank out the regions that actually code for stuff.
The new thing here is putting one in and having it stick around between generations. I suspect, although they don't say for sure, that this was done by breeding modified mice to modified mice, so that every newborn had two copies of the artificial chromosome. I say this because otherwise, with every cell division only one of the daughter cells would have the added chromosome. This means that in each mating, only three-fourths of the kids would have a copy; if you mate these offspring two more times, you're going to lose a significant number of your artificial chromosomes.
The idea that one can insert genes into a "safe" spot instead of having them integrate into the main genome is a good one. However, I'd be a bit worried about just inserting a gene without any promoter/represser elements and then amplifying that gene's function simply by adding more copies. The nucleus is not going to like having a large number of extra chromosomes floating around. IMHO, if you really want to do gene therapy to affect the descendants, you want to cut out the existing gene copy and put the new gene in exactly the same place. (Yes, if you wanted to add a totally new gene, these chromosomes would be a good place for it. Personally, I would not accept a new gene anytime soon, because I don't have faith in humanity's ability to get any technology right on the first pass.)
ASL as signed by the deaf does have some grammar; there's a canonical order in which you make signs, and reversing the order can alter the meaning. (At least, this is what I was told when I was taught some basic signs and phrases by an interpreter earlier this year.) It doesn't appear that Koko's humans consider the meaning to have changed when she alters sign order. Therefore, I would argue that what Koko uses is not actually a sign language, but is the sign equivalent of pointing and grunting.
But isn't the purpose of language, when boiled down to simplest terms, to communicate? If Koko wants a banana and is able to communicate that through sign language (as opposed to the toddler's point and grunt technique) isn't she communicating?
Communication != language, though. When your organs send hormones into the blood to signal other organs to do something, that's communication. When these packets fly across the fibers, that's communication. When a dog expresses displeasure by biting you, that's communication. However, none of these are language. Language, as I've repeatedly said, must contain not only words but also syntax. (Some people also add conditions about semantic understanding, but I'm not going there. Suffice to say that syntax is necessary but may not be sufficient. Similarly, language is probably a necessary condition of intelligence, but may not be sufficient.)
Practically speaking, what's the difference between the use of sign language and point-and-grunt behavior, assuming one can produce different kinds of grunts? IMHO, they're about equally expressive. (Maybe add pantomimed actions to the grunting as well.)
As far as self awareness, I thought some primates had been able to make the connection that the reflection in a mirror was their own image and not that of another.
Some studies have been published that claim this. Many have raised objections. At least some of the objectors are closed-minded. At least some of the studies were done by biased observers. Some studies have been done that refute the originals. Some studies have been done that refute the refutations. Lots of angry letters to the editor have been written. I did say it was a subject of great debate. I personally fall on the non-self-aware side of things; my belief is that if there were self-aware animals, they would have done something to get humans to stop killing them. (Note that Koko's keepers also claim she has a human-level IQ, which means that gorillas should be quite capable of organizing to stop gorilla hunters if they're actually sentient. (I personally don't think an IQ score is necessarily significant proof; I'd like to know just how they modified the tests so that the gorillas could take them.))
It certainly does. Unfortunately, if you look at the actual published papers of primate sign language use, the real picture becomes a bit more clear. Once again, these animals don't show syntactic constructions --- they chain symbols together, so that if Koko wants a banana, you may get any of "give Koko banana", "banana give Koko", or "give banana Koko". Which one gets used most is basically going to depend on how promptly each is rewarded with a banana. (This is also why I'd expect to see less of "Koko give banana", because a human is not going to parse this as a request for a banana quite as easily.) Also notice that all her replies are interpreted for semantic content by a human who has begun with the expectation that Koko has a complete command of English/ASL. There's some mighty big bias there. (Of course, I'm biased to say that this is not true language.)
I personally consider a site with a vested interest in proving that Koko is intelligent/sentient/a "person"/whatever to be an unreliable source. The stuff about her demonstrating productive combination is interesting, but I also kind of wonder how many times she didn't do this when presented with a new object. (That's the general problem here --- a few instances of a behavior that could be interpreted in a certain way are being heavily generalized, and I would say they're being overgeneralized.)
Now, the mirror behavior... that I'm not sure about. There's a raging debate about this in the cognitive psych community, but nobody has a good definition of self-awareness, so we can't prove things one way or another (yet).
Koko can certainly communicate in a reasonably rich fashion with those not of her own species, which is itself impressive, but I'm not seeing language or a definite self-awareness here. IMHO, the best proof of true primate sentience would be a gorilla indicating to its keepers that it wanted to write an essay proving it was sentient, and then following through on that desire. I want to see multiple isolated gorillas sign "I think, therefore I am."
To put things in more Slashdottish terms, Koko doesn't pass the Turing test.:-) (Actually, that's another thing about language . Words aren't the basic unit of language; phonemes are. I've never heard of an animal which was able to master the concept of a keyboard.)
What about Linus? He's got the trademark on the word "Linux", IIRC. Would he not have the power to demand that a company making Linux-broken objects stop using his trademark (since it dilutes the mark to have it associated with stuff that doesn't work)? If this is not strong enough, it seems that he could simply trademark "Certified Linux Compatible" with a little Tux logo and tell us to look for that.
1) Introspect on your own knowledge of words. Most people have a sense that there's a bit more than just an association between a sound sequence and a set of physical objects --- there's a sense of understanding that this is not just any set of tones, this is a word, and it is attached to an abstract concept. How on earth will you teach a parrot what love is? What hate is? The definition of "taste"? This is the difference between a human and a parrot: you can teach a human new words using only other words. To teach a parrot new words, you must present physical objects (if you want the words associated with things).
2) Ability to use single words or memorized phrases is not language, and ability to see a word on a page and say that word is not the complete definition of reading. Language has syntax. No animal to date has ever demonstrated the ability to learn or use syntactic constructions. From your description, your sister was not using syntax yet. (Look at it another way: if I taught you five Chinese phrases without giving you meanings, but you knew that certain phrases would get Chinese people to give you food if you said them at the right time of day, would you claim to speak Chinese? (Of course, that raises the question of just how many of these phrases you'd need to know before you did actually have a functional understanding of Chinese. Read about the "Chinese Room Argument" if you care.))
Yes, I'm aware of that; for that matter, I suspect the original poster is as well, since said story was posted to Slashdot. That's not really what he was asking, though.
The EEG thing and eye tracking are reasonable ways of doing WIMP interfaces, but they can't handle typing. Subvocalization, at least according to the chair of our communication disorders department, is not really something that one can pick up; he claims that in order to get vibrations that the computer could interpret, you'd have to pretty much make all the actual sounds. Wearables are likely to work more on a basis of chorded keyboards or something similar, IMHO.
A long time. We don't know how the speech centers of the brain really work. We definitely don't have the ability to make the kind of electrode array needed to read the entire dataflow, and we lack the computational power to process that data in realtime (although that last is trivial compared to the first two).
It is something being actively worked on with sensory prostheses, but in order to directly read cognitively signifcant signals, we're going to have to obtain a much better understanding of mental processing. You're going to see true direct input about the same time neuroscience solves the strong AI problem (give or take a few decades).
At the current rate of knowledge increase, though, this should happen within the average Slashdotter's lifetime.
Actually, relative heat loss is determined by an animal's surface area/volume ratio, and larger animals have lower SA/V. An elephant retains body heat much better than a mouse, and the mouse compensates for this by having a metabolism from hell.
So, in theory, if one grew a mouse the size of a dog and left its metabolic rate alone, yes, it actually might have troubles with overheating.
Couple problems there. First off, tumors are actually pretty tricky to recognize in the early stage; on a pathology slide, it could easily be written off as normal tissue variation. Secondly, the immune system doesn't destroy moles because immunologically speaking, they are healthy normal tissue. The difference between a mole and a melanoma is basically a matter of DNA damage leading to altered surface proteins.
Surface antigens are the only known way to recognize individual cells. (If you just look at a single cell under ordinary light with no particular context, you'd have a very hard time telling if it was a cancerous cell or normal tissue, especially if all you see is the outside.) This means that even if we put in a nanobot swarm, that swarm is vulnerable to at least some of the same tactics which tumors use to evade the immune system. (For example, many tumors shed large amounts of soluble protein into the fluid around them; this keeps the immune system distracted, much as a flare could distract a heat-seeking missile.)
Diagnosis of cancer before it's a macroscale-visible lump is not an easy thing. Almost all tumor diagnoses are made by seeing the lump on a radiogram, taking a chunk out of it, and using some nasty chemical stains to visualize that tissue. These aren't methods which we can use inside a living patient, and as I mentioned before, the natural solution is highly suboptimal.
I wouldn't go thinking of these as the cure for cancer. Microsurgery will certainly cause more tumors to be operable, and thus save lives, but it won't cure cancer the same way we've cured smallpox. IMHO, there may never be a complete eradication of cancer; it'll just come down to being as much of a worry as being hit by a meteor.
This story, posted yesterday, has a ton of thoughts about ASPs attached to it (primarily negative, at least at my threshold).
IMHO, the current Internet and the current behavior of normal corporate/personal computer systems is such that ASPs are bound to be plagued with security, privacy, and general ease-of-use issues. I believe that ASPs are often a poor idea because they put everyone under UCITA terms: if a developer doesn't like you, they can remove your ability to work (at least with their tool, and if all your files are in encrypted proprietary formats, you're screwed).
That said, sure, there might be a place for them; as another poster said, it might be nice to be able to rent a package just for an afternoon or a week. OTOH, I would be very unhappy to have my entire system be composed of rented applications. I believe that we'll see the same phenomenon as with housing: the renters will be the ones who only need a short-term resource or can't afford to actually buy.
I don't attribute it to direct intention as in "Make rejecting cookies hard, so that the user won't do it." I would say, rather, that the default for any UI is "annoying to use", and that the only deliberate decision went something like:
"Boy, cookies are hard to disable, aren't they? Do you think we should fix it?"
"Well, let's see. The users don't pay us for the software. The cookie-senders have strategic partnerships with us. I don't feel like putting any effort into that feature, do you?"
The current cookie options are easy to code; I'd say that's sufficient explanation.
If I wrote it with that in mind, of course I would. What kind of coward won't stand up and take responsibility for the quality of his own work?
Ummm... read any software licenses lately? There's a reason they disclaim all warranties beyond "If the CD is scratched and you notice within 90 days of purchase, you can have a new one." Unless you're coding for a very limited environment (like, say, a Space Shuttle's control readouts), it is IMHO very unwise to start making guarantees. This is especially true if your code is exposed to the Internet where anyone can try to break it. "With enough eyeballs, all bugs are shallow" can cut more than one way.
A secure virtual machine for making arbitrary calculations can be very simple indeed; you only really need a few operations. It would be like signing a statement that you totaled a column of numbers correctly; you'd want to check it over until you're certain, and charge extra for the time and worry of that, but it's a simple enough task that you can eventually be certain that you're correct.
Can you? How? Can you prove that your double-checking mechanisms are not in error? Can you prove that everything was implemented correctly? What about the libraries against which you linked? No floating-point bugs in the hardware? Nothing corrupted by a recent system crash (perhaps deliberately induced)?
I see nothing to refute my claim that 100% certainty is impossible. I see some good techniques for getting the uncertainty to very low levels, but you're claiming perfect security.
How do you think hardware designers ever get anything done? There's no magical difference that makes bug-free hardware possible and bug-free software impossible.
This is not strictly true. Hardware is governed by a few simple physical laws. You can simulate and model hardware under realistic conditions in order to test its operating range. Software, on the other hand, is a set of arbitrary finite state machines. It cannot be simulated with nearly the same accuracy or precision --- there are fundamental theorems of computational theory that demonstrate this. It is mathematically impossible to say "this program will always work in this fashion". If you wish, I will post the proof.
I never said anything about that. I was very clearly responding to "I personally wouldn't want to be in charge of maintaining a machine which is set up to accept and execute arbitrary tasks from passing users." and talking about protecting the machine from the tasks (and the tasks from each other).
I'd say that protecting the tasks from the machine also falls within the duties of someone maintaining a machine running arbitrary submitted code. Why do you think this is not the case?
I'm not talking about communications security (which, of course, can never be perfect, for physical reasons; all theoretical communications security models rely on the absolute physical security of certain things, which is impossible in real life), I'm talking about the security of one machine and the processes that run on it.
It's unreasonable to restrict discussion to one machine when talking about ASPs, though. The whole point of such things is that data and code are flying around the network like mad and getting executed all over the place. The communications behavior of the machine is a fundamental part of the system under discussion, and if that can be compromised, I believe my original claim is correct.
BTW, what kind of idiot would let their car be controlled by a distant server over a network? Lines get cut, solar flares disrupt communications, networks go down.
Read the articles, especially the Dertouzos press release. Now extrapolate. Which do you think makes more sense: me hunting around for a button on the panel, or me saying "Car, set cruise control."? Cars are likely to eventually have a limited autopilot. If I want to take a nap and don't want my teenage kids trying to drive, I'll want to be able to lock the controls. (Admittedly, a verbal password is a bit of the wrong idea here --- probably want a voiceprint or other biometric.) According to the article, all of this is going to be handled by the car's computer sending my voice to a distant server, which processes the voice into commands for my car and sends back the commands. There's your remote-controlled car.
To look at it another way, companies can profit if everyone is dependent on stuff being shipped through private networks to private servers via proprietary protocols. The more things that are transmitted and controlled, the better. IMHO, this will lead to some things being networked that shouldn't, but that won't change until some early adopters get hurt or killed.
I agree with you that a well-hacked Mozilla-like program is one option, but there's another one: proxies. As far as keeping cookies off your drive, JunkBuster seems to do a pretty good job, and offers a much more fine-grained control over what's going on than the current option of "Either block all cookies or allow them all or get nagged every five seconds for each individual cookie." (Yes, there's the "trusted sites" zone in IE, but I don't care to mark any site even temporarily trustable.)
Your more general point of "the only way you'll get a cookie-free web experience is hacking one together yourself" is quite correct, though.
OK, I'll bite (and demonstrate some ignorance). What are OS390 and RACF? I have not heard of these things, and if they are a demonstration of a completely secure system, I'd be interested in some pointers to them.
Well, OK, there was the Mac 128K, but that was my dad's.
Anyway, what're you gonna do about it, old-timer? Gonna beat me over the head with your pirated Altair BASIC? Or when I was your age did you compute with an abacus and like it? Was it uphill to the computer room both ways? In the rain? Was 640K enough for you, goldurnit?
Slashdot Mirror
Yes, it's a step up from the current "Allow all/deny all/ask all" model, but IMHO, this is still useless. I may not want to reject all third-party cookies, just most of them. Sure, there's the Trusted Sites zone, but I don't consider any website worthy of the label "Trusted".
This is still very coarse-grained control. I can do a million times better than this with my nice little Junkbuster and get rid of banner ads while I'm at it. I don't see this change as really doing very much for privacy protection compared to existing free software. Sure, it's a nice gesture from MS, but I think that's all it is --- throwing us a bone so they can say "See? We're very concerned about privacy. Ignore that digital watermark we attached to all your Office documents."
I don't have the book because I probably wouldn't bother to buy a full Mindstorms kit, but what makes you say that a breadth-first search is necessarily bad? I can see how it's not much help for the Mindstorms guru who wants to know more, but for someone who's just starting out, a broad review (with specific examples) seems like a good thing.
I got started in the robot lab of a small liberal arts college. They didn't have a whole lot in the way of facilities; a few of their bots were expensive research jobs from Pioneer, but mostly it was a matter of students screwing parts together. Said parts mainly came from the more expensive catalog kits; in my last days they finally acquired a 3D printer. (Of course, you have to have some mechE background and CAD experience to use a 3D printer.)
If you want to do this academically, the easiest thing is to search your school, find out who has robots, and go work with those people. Be prepared for an enormous amount of frustration; academic robotics research is mainly the study of an infinite number of Things That Don't Work.
If what you want is home robotics, the Mindstorms kit may be an option. However, you may have been looking for a cheaper solution. Lego is still about the best solution (IMHO) for the body of a home hobbyist's robot, because it's fairly cheap and very easy to redesign as needed. Buy a big box of Lego Technic and order spare connectors as your cat eats them or they get lost in the couch. There are many alternatives to Mindstorms for controllers; many people swear by the BASIC stamp, for example. This is a case where you're going to have to search the web and catalogs to find some parts within your price range. The more soldering and splicing and assembly you're willing to do, the better the price you can get.
Note that for "advanced" stuff you will also need a workbench which lets you drill and cut and shape parts with reasonable accuracy. You're also likely to need some electronics supplies --- a soldering iron and a standard set of resistors at the very least. The initial outlays here could also eat into your budget.
Your situation sounds like that of a coder who'd like to play with robots but really doesn't have much in the way of dedicated workspace, tools, or know-how. Given your claimed lack of electronics skills, you probably want to avoid kits until you're more comfortable. In your situation, I'd start with Mindstorms: you get plenty of tutorial advice and can even start working with a friendly little GUI at first. You might pick up the O'Reilly book on Mindstorms to help you out. That setup should give you a flavor for what robot-building is really like and should cost you $200 maximum. It'll take a while to exhaust all the capacities of a mildly hacked Mindstorms set (the most you'll need is extra Lego or sensors), and by then you'll be beyond the realm of Ask Slashdot.
I can see the tree stuff; those are interesting examples, though they make my brain hurt to read them (I'm of the opinion that good Scheme is liberally interspersed with comments that help you to know what block you're currently in; after about the third lambda I lose track.)
As far as the list-multiplier goes, though, wouldn't it be just as simple to say
(define (multiply-list l)
(let ((recursor (lambda (lst accum)
(cond ((null? lst) accum)
((= (car lst) 0) 0)
('else (recursor
(cdr lst)
(* (car lst)
accum)))))))
(recursor l 1)))
The syntax may be off, but I believe that's close to the tail-recursive method, which seems to do just what's needed.
Y'know, I earned my bachelor's in CS a year ago, TA'ed two functional-language-based classes, and wrote extensions to Scheme that made it useful for mobile agent programming, and I still don't know what call/cc is for. When the hell do you actually need something like that?
Could somebody explain to me why it's a good idea to layer this on top of HTTP? As far as I can see, it doesn't really solve the firewall problem; the article specifically says they'll run on different ports so that admins can firewall all IPP traffic. Last I checked, HTTP actually used a fairly inefficient file-transfer mechanism. Given that the modern corporate document is filled with images and charts and sundry multimedia (and therefore bloats pretty quickly), wouldn't it make more sense to layer this on top of, say, FTP? Why is it layered on top of anything at all? What does HTTP provide that IPP needs?
I'm sure there are some answers to these questions, because I believe in the IETF and I personally have only ever designed toy protocols, but I don't see the answers in the article.
From the linked article:
"Once modified, the artificial chromosome can be duplicated hundreds of thousands of times. The company says that the amount of protein produced rises in step with the number of gene copies..."
I intially read this as them doing chromosome amplification within a single cell. (Such a thing is not unheard of; it happens in certain cancers, although it's usually not a whole chromosome being amplified.) On reparse, I think you're right; it's more a matter of them inserting extra copies. IMHO, that's not the right solution; you can't regulate that very well. I'd say it'd be a better idea to insert strong promoter elements along with the new genes if you wanted them to be hyperproductive. (That way, if the gene temporarily needs to be turned down, you can administer a compound which inhibits the activator protein (although that may have issues depending on what other genes share that promoter).)
Biologists have been making bacterial artificial chromosomes, or BACs, for years. BACs were a key component of the Human Genome project. They're conceptually quite simple; you take any ordinary chromosome and yank out the regions that actually code for stuff.
The new thing here is putting one in and having it stick around between generations. I suspect, although they don't say for sure, that this was done by breeding modified mice to modified mice, so that every newborn had two copies of the artificial chromosome. I say this because otherwise, with every cell division only one of the daughter cells would have the added chromosome. This means that in each mating, only three-fourths of the kids would have a copy; if you mate these offspring two more times, you're going to lose a significant number of your artificial chromosomes.
The idea that one can insert genes into a "safe" spot instead of having them integrate into the main genome is a good one. However, I'd be a bit worried about just inserting a gene without any promoter/represser elements and then amplifying that gene's function simply by adding more copies. The nucleus is not going to like having a large number of extra chromosomes floating around.
IMHO, if you really want to do gene therapy to affect the descendants, you want to cut out the existing gene copy and put the new gene in exactly the same place. (Yes, if you wanted to add a totally new gene, these chromosomes would be a good place for it. Personally, I would not accept a new gene anytime soon, because I don't have faith in humanity's ability to get any technology right on the first pass.)
ASL as signed by the deaf does have some grammar; there's a canonical order in which you make signs, and reversing the order can alter the meaning. (At least, this is what I was told when I was taught some basic signs and phrases by an interpreter earlier this year.) It doesn't appear that Koko's humans consider the meaning to have changed when she alters sign order. Therefore, I would argue that what Koko uses is not actually a sign language, but is the sign equivalent of pointing and grunting.
But isn't the purpose of language, when boiled down to simplest terms, to communicate? If Koko wants a banana and is able to communicate that through sign language (as opposed to the toddler's point and grunt technique) isn't she communicating?
Communication != language, though. When your organs send hormones into the blood to signal other organs to do something, that's communication. When these packets fly across the fibers, that's communication. When a dog expresses displeasure by biting you, that's communication. However, none of these are language. Language, as I've repeatedly said, must contain not only words but also syntax. (Some people also add conditions about semantic understanding, but I'm not going there. Suffice to say that syntax is necessary but may not be sufficient. Similarly, language is probably a necessary condition of intelligence, but may not be sufficient.)
Practically speaking, what's the difference between the use of sign language and point-and-grunt behavior, assuming one can produce different kinds of grunts? IMHO, they're about equally expressive. (Maybe add pantomimed actions to the grunting as well.)
As far as self awareness, I thought some primates had been able to make the connection that the reflection in a mirror was their own image and not that of another.
Some studies have been published that claim this. Many have raised objections. At least some of the objectors are closed-minded. At least some of the studies were done by biased observers. Some studies have been done that refute the originals. Some studies have been done that refute the refutations. Lots of angry letters to the editor have been written. I did say it was a subject of great debate. I personally fall on the non-self-aware side of things; my belief is that if there were self-aware animals, they would have done something to get humans to stop killing them. (Note that Koko's keepers also claim she has a human-level IQ, which means that gorillas should be quite capable of organizing to stop gorilla hunters if they're actually sentient. (I personally don't think an IQ score is necessarily significant proof; I'd like to know just how they modified the tests so that the gorillas could take them.))
It certainly does. Unfortunately, if you look at the actual published papers of primate sign language use, the real picture becomes a bit more clear. Once again, these animals don't show syntactic constructions --- they chain symbols together, so that if Koko wants a banana, you may get any of "give Koko banana", "banana give Koko", or "give banana Koko". Which one gets used most is basically going to depend on how promptly each is rewarded with a banana. (This is also why I'd expect to see less of "Koko give banana", because a human is not going to parse this as a request for a banana quite as easily.) Also notice that all her replies are interpreted for semantic content by a human who has begun with the expectation that Koko has a complete command of English/ASL. There's some mighty big bias there. (Of course, I'm biased to say that this is not true language.)
:-) (Actually, that's another thing about language . Words aren't the basic unit of language; phonemes are. I've never heard of an animal which was able to master the concept of a keyboard.)
I personally consider a site with a vested interest in proving that Koko is intelligent/sentient/a "person"/whatever to be an unreliable source. The stuff about her demonstrating productive combination is interesting, but I also kind of wonder how many times she didn't do this when presented with a new object. (That's the general problem here --- a few instances of a behavior that could be interpreted in a certain way are being heavily generalized, and I would say they're being overgeneralized.)
Now, the mirror behavior... that I'm not sure about. There's a raging debate about this in the cognitive psych community, but nobody has a good definition of self-awareness, so we can't prove things one way or another (yet).
Koko can certainly communicate in a reasonably rich fashion with those not of her own species, which is itself impressive, but I'm not seeing language or a definite self-awareness here. IMHO, the best proof of true primate sentience would be a gorilla indicating to its keepers that it wanted to write an essay proving it was sentient, and then following through on that desire. I want to see multiple isolated gorillas sign "I think, therefore I am."
To put things in more Slashdottish terms, Koko doesn't pass the Turing test.
What about Linus? He's got the trademark on the word "Linux", IIRC. Would he not have the power to demand that a company making Linux-broken objects stop using his trademark (since it dilutes the mark to have it associated with stuff that doesn't work)? If this is not strong enough, it seems that he could simply trademark "Certified Linux Compatible" with a little Tux logo and tell us to look for that.
Two thoughts:
1) Introspect on your own knowledge of words. Most people have a sense that there's a bit more than just an association between a sound sequence and a set of physical objects --- there's a sense of understanding that this is not just any set of tones, this is a word, and it is attached to an abstract concept. How on earth will you teach a parrot what love is? What hate is? The definition of "taste"? This is the difference between a human and a parrot: you can teach a human new words using only other words. To teach a parrot new words, you must present physical objects (if you want the words associated with things).
2) Ability to use single words or memorized phrases is not language, and ability to see a word on a page and say that word is not the complete definition of reading. Language has syntax. No animal to date has ever demonstrated the ability to learn or use syntactic constructions. From your description, your sister was not using syntax yet. (Look at it another way: if I taught you five Chinese phrases without giving you meanings, but you knew that certain phrases would get Chinese people to give you food if you said them at the right time of day, would you claim to speak Chinese? (Of course, that raises the question of just how many of these phrases you'd need to know before you did actually have a functional understanding of Chinese. Read about the "Chinese Room Argument" if you care.))
Yes, I'm aware of that; for that matter, I suspect the original poster is as well, since said story was posted to Slashdot. That's not really what he was asking, though.
The EEG thing and eye tracking are reasonable ways of doing WIMP interfaces, but they can't handle typing. Subvocalization, at least according to the chair of our communication disorders department, is not really something that one can pick up; he claims that in order to get vibrations that the computer could interpret, you'd have to pretty much make all the actual sounds. Wearables are likely to work more on a basis of chorded keyboards or something similar, IMHO.
A long time. We don't know how the speech centers of the brain really work. We definitely don't have the ability to make the kind of electrode array needed to read the entire dataflow, and we lack the computational power to process that data in realtime (although that last is trivial compared to the first two).
It is something being actively worked on with sensory prostheses, but in order to directly read cognitively signifcant signals, we're going to have to obtain a much better understanding of mental processing. You're going to see true direct input about the same time neuroscience solves the strong AI problem (give or take a few decades).
At the current rate of knowledge increase, though, this should happen within the average Slashdotter's lifetime.
Actually, relative heat loss is determined by an animal's surface area/volume ratio, and larger animals have lower SA/V. An elephant retains body heat much better than a mouse, and the mouse compensates for this by having a metabolism from hell.
So, in theory, if one grew a mouse the size of a dog and left its metabolic rate alone, yes, it actually might have troubles with overheating.
That's pretty cool.
Couple problems there. First off, tumors are actually pretty tricky to recognize in the early stage; on a pathology slide, it could easily be written off as normal tissue variation. Secondly, the immune system doesn't destroy moles because immunologically speaking, they are healthy normal tissue. The difference between a mole and a melanoma is basically a matter of DNA damage leading to altered surface proteins.
Surface antigens are the only known way to recognize individual cells. (If you just look at a single cell under ordinary light with no particular context, you'd have a very hard time telling if it was a cancerous cell or normal tissue, especially if all you see is the outside.) This means that even if we put in a nanobot swarm, that swarm is vulnerable to at least some of the same tactics which tumors use to evade the immune system. (For example, many tumors shed large amounts of soluble protein into the fluid around them; this keeps the immune system distracted, much as a flare could distract a heat-seeking missile.)
Diagnosis of cancer before it's a macroscale-visible lump is not an easy thing. Almost all tumor diagnoses are made by seeing the lump on a radiogram, taking a chunk out of it, and using some nasty chemical stains to visualize that tissue. These aren't methods which we can use inside a living patient, and as I mentioned before, the natural solution is highly suboptimal.
I wouldn't go thinking of these as the cure for cancer. Microsurgery will certainly cause more tumors to be operable, and thus save lives, but it won't cure cancer the same way we've cured smallpox. IMHO, there may never be a complete eradication of cancer; it'll just come down to being as much of a worry as being hit by a meteor.
Yeah, but we all know just how secure existing authentication is.
This story, posted yesterday, has a ton of thoughts about ASPs attached to it (primarily negative, at least at my threshold).
IMHO, the current Internet and the current behavior of normal corporate/personal computer systems is such that ASPs are bound to be plagued with security, privacy, and general ease-of-use issues. I believe that ASPs are often a poor idea because they put everyone under UCITA terms: if a developer doesn't like you, they can remove your ability to work (at least with their tool, and if all your files are in encrypted proprietary formats, you're screwed).
That said, sure, there might be a place for them; as another poster said, it might be nice to be able to rent a package just for an afternoon or a week. OTOH, I would be very unhappy to have my entire system be composed of rented applications. I believe that we'll see the same phenomenon as with housing: the renters will be the ones who only need a short-term resource or can't afford to actually buy.
I don't attribute it to direct intention as in "Make rejecting cookies hard, so that the user won't do it." I would say, rather, that the default for any UI is "annoying to use", and that the only deliberate decision went something like:
"Boy, cookies are hard to disable, aren't they? Do you think we should fix it?"
"Well, let's see. The users don't pay us for the software. The cookie-senders have strategic partnerships with us. I don't feel like putting any effort into that feature, do you?"
The current cookie options are easy to code; I'd say that's sufficient explanation.
If I wrote it with that in mind, of course I would. What kind of coward won't stand up and take responsibility for the quality of his own work?
Ummm... read any software licenses lately? There's a reason they disclaim all warranties beyond "If the CD is scratched and you notice within 90 days of purchase, you can have a new one." Unless you're coding for a very limited environment (like, say, a Space Shuttle's control readouts), it is IMHO very unwise to start making guarantees. This is especially true if your code is exposed to the Internet where anyone can try to break it. "With enough eyeballs, all bugs are shallow" can cut more than one way.
A secure virtual machine for making arbitrary calculations can be very simple indeed; you only really need a few operations. It would be like signing a statement that you totaled a column of numbers correctly; you'd want to check it over until you're certain, and charge extra for the time and worry of that, but it's a simple enough task that you can eventually be certain that you're correct.
Can you? How? Can you prove that your double-checking mechanisms are not in error? Can you prove that everything was implemented correctly? What about the libraries against which you linked? No floating-point bugs in the hardware? Nothing corrupted by a recent system crash (perhaps deliberately induced)?
I see nothing to refute my claim that 100% certainty is impossible. I see some good techniques for getting the uncertainty to very low levels, but you're claiming perfect security.
How do you think hardware designers ever get anything done? There's no magical difference that makes bug-free hardware possible and bug-free software impossible.
This is not strictly true. Hardware is governed by a few simple physical laws. You can simulate and model hardware under realistic conditions in order to test its operating range. Software, on the other hand, is a set of arbitrary finite state machines. It cannot be simulated with nearly the same accuracy or precision --- there are fundamental theorems of computational theory that demonstrate this. It is mathematically impossible to say "this program will always work in this fashion". If you wish, I will post the proof.
I never said anything about that. I was very clearly responding to "I personally wouldn't want to be in charge of maintaining a machine which is set up to accept and execute arbitrary tasks from passing users." and talking about protecting the machine from the tasks (and the tasks from each other).
I'd say that protecting the tasks from the machine also falls within the duties of someone maintaining a machine running arbitrary submitted code. Why do you think this is not the case?
I'm not talking about communications security (which, of course, can never be perfect, for physical reasons; all theoretical communications security models rely on the absolute physical security of certain things, which is impossible in real life), I'm talking about the security of one machine and the processes that run on it.
It's unreasonable to restrict discussion to one machine when talking about ASPs, though. The whole point of such things is that data and code are flying around the network like mad and getting executed all over the place. The communications behavior of the machine is a fundamental part of the system under discussion, and if that can be compromised, I believe my original claim is correct.
BTW, what kind of idiot would let their car be controlled by a distant server over a network? Lines get cut, solar flares disrupt communications, networks go down.
Read the articles, especially the Dertouzos press release. Now extrapolate. Which do you think makes more sense: me hunting around for a button on the panel, or me saying "Car, set cruise control."? Cars are likely to eventually have a limited autopilot. If I want to take a nap and don't want my teenage kids trying to drive, I'll want to be able to lock the controls. (Admittedly, a verbal password is a bit of the wrong idea here --- probably want a voiceprint or other biometric.) According to the article, all of this is going to be handled by the car's computer sending my voice to a distant server, which processes the voice into commands for my car and sends back the commands. There's your remote-controlled car.
To look at it another way, companies can profit if everyone is dependent on stuff being shipped through private networks to private servers via proprietary protocols. The more things that are transmitted and controlled, the better. IMHO, this will lead to some things being networked that shouldn't, but that won't change until some early adopters get hurt or killed.
I agree with you that a well-hacked Mozilla-like program is one option, but there's another one: proxies. As far as keeping cookies off your drive, JunkBuster seems to do a pretty good job, and offers a much more fine-grained control over what's going on than the current option of "Either block all cookies or allow them all or get nagged every five seconds for each individual cookie." (Yes, there's the "trusted sites" zone in IE, but I don't care to mark any site even temporarily trustable.)
Your more general point of "the only way you'll get a cookie-free web experience is hacking one together yourself" is quite correct, though.
)
(Forgot to close a paren in previous post.)
OK, I'll bite (and demonstrate some ignorance). What are OS390 and RACF? I have not heard of these things, and if they are a demonstration of a completely secure system, I'd be interested in some pointers to them.
Well, OK, there was the Mac 128K, but that was my dad's.
Anyway, what're you gonna do about it, old-timer? Gonna beat me over the head with your pirated Altair BASIC? Or when I was your age did you compute with an abacus and like it? Was it uphill to the computer room both ways? In the rain? Was 640K enough for you, goldurnit?