I'm doing a postdoc in CompSci now, and its nowhere near a six-figure salary. And knowing the payrates in quite a few countries for postdocs I call bullshit. Link a single position that pays six figures for a postdoc.
You've hit the nail on the head when you say application*s*. Given that this article is a dupe, and we all bashed ajaxWrite last time it was up, I'm suprised that nobodies mentioned the other apps. This guys plan was to realise a new app every week. So far he's got a sketch program, and a something for videoediting. The sketch program (like the ajaxWrite) is exactly the kind of simple programing assignment that you would get somebody to do to learn a new language. Its not drowning in features, although of course he compares it to Inkscape. I didn't look at the videoediting as thats not really my thing.
One amusing thing about his hype machine is that he has a claim that the idea for the word processor came from another guy, and that he injected $50000 as VC to get it off the ground. Odd thing is three weeks ago he made that claim about the word processor. It seems that this claim shifts through the applications as they are written. That is some impressive use of 200,000 quarters on a piece of string trick.
AnalogDiehard doesn't want to read the article, and she wants to draw attention away from the fact. This is just a pre-canned comment with no relevence to the article as a tactic to lend legitimacy to her karma-whoring.
That's one of the most insightful posts that I've ever read on this site. A really good breakdown of what it is to apply the knowledge behind discrete math to a problem, rather than solve problems in discrete math.
And in response to the other reply above, yes, a tree is a type of graph but you've missed the point. If an algorithm can be coded over trees rather than graphs, then it should be, as it will simplify the design, and reduce the complexity.
Dogma : An authoritative principle, belief, or statement of ideas or opinion, especially one considered to be absolutely true.
Please note the difference between this and your definition of dogma. As regards a cult, here is a loose, but working definition;
1. An organisation that uses dogma to control the beliefs of its followers
2. An organisation that refuses to discuss this dogma to people that have not been indoctrinated.
3. An organisation that demands that all of its followers ungo an indoctrination to bind them to the organisation.
4. - 12. may vary depending on the cult.
I notice that you've dodged the question of what 12 steps you adhere do, and that you ask us to suspend disbelieve that it's not psuedoreligous or a cult without showing us the program. That is how your behaviour is dogmatic.
No, as I stated, you acted dogmatically in the way that you flamed the original poster. I never used the phrase "incontrovertably true", so either you're confused about which post was mine, or about the meaning of the word dogmatic. The generic 12-steps that you've linked to are psuedo-religous. I asked for your specific flavour as I wondered why you acted like a cultist when you claimed that you weren't. Saying that your fellowship's traditions include not disclosing the traditions certainly sounds likes the actions of a cult.
You've responded quite harshly to an insightful post on the subject. The OP made a comment that 12-step programs are psuedo-religious and cultish, but in defending them you've acted in the dogmatic way that those criticisms suggest. Purely out of interest, and not to start a flamewar here, what are the 12 steps of your program?
You are simplifying many aspects of the problem. It is possible that we have a working example of a solution, and yet we cannot create another. In order to build upon a previous example we need to reverse engineer the working solution. This is not a simple task. We have people in our department researching computational neuroscience and I've seen some of their seminars. The problem is so complex that there is no other problem that we can even use as a metaphor for it.
We have open problems in different fields of science that have resisted all attempts to solve them. The problem with Kurzweil (and his supporters) is that they *assume* that AI will fall into the solvable category if we throw enough resources at it. This assumption is not necessarily true.
Simply saying that an increase in technology will allow us to build upon existing biological systems is a fallacy. Even ignoring the ethical issues of such experimentation on sentient (or near-sentient) beings, there is the technological challenge.
Assume that we are lucky once, and manage to perform a modification to a biological brain. This does not *automatically* lead to further increases. Even to the increased intelligence of the modified organism, the problem may still be too complex to be tractable. In which case the next advance is still waiting for another lucky strike. And given the complexity of the organism the odds of such a lucky strike are small.
One analogy is the difference between the increase in silicon fabrication technology, and the increase in software development. One is a simple problem that we can throw engineering resources at in order to gain incremental increases. The other is largely intractable, and when progress does come it does so in sporadic spurts of growth. The result is an exponential increase in chip performance (so far) against a linear (at best) increase in software performance.
Assumptions about which class of problem the solution to AI belongs to have no merit at all.
Err, where are we producing human level intelligences today? Other than real actual people... AI is a genuinely hard problem. There is no guarantee that it will be solved. Whilst some problems have fallen to human endevour there are other problems that have not. Try looking at a list of open problems in number theory sometime. These have had some of the brightest minds of the past 500 years try and fail to solve them. A workable AI may fall into the same category.
I think that the GP has a fair point. It's not that the GPU and the CPU are comming together, it is more like there are two separate roles. In one role you want a vector that can run *lots* of numerical code with simple control flow. Graphics, physics and a bunch of other stuff can be executed on this processor. But for the CPU you want something that can handle complex control flow. As complex control flow is tricky (pipeline stalls, branch prediction, hiding memory latency...) the easiest way to do it is through multi-threading. So you have one chip that does all of the parallel processes in lockstep taking advantage of the simple control flow to predict ahead. And there is a different chip that does all of the processes at separate rates using multitasking to hide the latency, rather than prediction.
Thanks for the link. There's no authorative reference in there, but it does the match the figures that I found from a google. The main point in the article is quite interesting - that while we have fairly good population figures for humans there is no definite figure for the number of cars. I guess that your 750mil figure was fairly accurate after all...;^)
I don't know what you googled as you didn't provide a URL but your 750 million cars figure sounds a bit fishy. While your overall argument that most people don't own a car, the population in America is ~325 million, and Europe is ~750 million. Both have a saturated car market with multiple ownership per household. That would be around 1billion cars alone. China, India and other places to a decreasing degree are buying cars like crazy. Their ownership may be much lower but there is ahug base of old Soviet vehicles and obsolete western models across the world. I'd guess (hand-waving) that there are 1.5 - 2 billion cars in circulation...
My ISP is Plusnet, the one mentioned in the piece. Their line of argument is that they have never sold an "unlimited" packaged. Furthermore their literature does state clearly that it is a contended product with a contention ration of (I forget exactly) about 30:1. So other users on the same contended line do have an effect on your usage. And since they started shaping the p2p traffic performance has picked up massively for everyone else.
There is no password protection on most mobile phones, they just have pin entry when they're first switched on. That is no good if the phone is already powered up.
Cancellation would not solve this problem. You are kind of missing the point. The journalist that the GP is referrering to showed how easy it is to set up tracking on somebody elses phone. Once setup there is *no* indication that the service is active. Thus you wouldn't know to cancel it, or who with.
That's quite impressive, I'll have to have a look at their site later and see how they are doing it. It doesn't appear to give full spatialisation, I'm guessing that it's supposed to sound like it's a sound source rotating around my head. That could be down to the fullsize headphones that I'm wearing at the moment. Some of it sounds similar to the effects that we created where there is some element of depth, but it just sounds like it's panning right-left across the back of your shoulders. There was also a small amount of up/sound spatialisation, but not very strong.
With those quibbles, it's still the best demo that I've heard yet. Thanks for the link.
I call bullshit. We did some research into this a few years ago and producing spatialised sound through eyebud headphones is not realistic. Each person has a unique head-transform based on the curves of the earlobes amongst other things. What is feasible is telling people that it is spatialised audio and then allowing a suggestive / placebo effect to work. As they're going for surround rather than full spatialisation it is also possible that they've just boosted the bitrate on the bass frequencies so that it sounds less tiny. This gives a 'fuller volume' effect which most people will accept as surround if you suggest it.
Sorry for the rant, but we spent a long time dicking around with spatialised audio until we did 'blind' listening trials to confirm that it was bollocks. Also, if they do get some kind of spatialisation working then it would require earbuds as oppoosed to fullsize headphones. The fullsize ones have a sound source outside of the ear canal which means that the outer ear picks up its own localisation cues.
That's good to hear. Are there any reports of people using it over NFS? One of the problems with subversion used to be that once a repository was corrupt you were screwed. I think that the repair tools have improved a lot since then, but I'd still be a bit dubious until lots of other people had already tested it;^)
Non-root users don't have access to the server for security reasons. And the sysadmin doesn't want any non-critical processes running on the box. Basically it does nothing but NFS.
SVN does rock, and I use it over CVS any chance that I get... but the NFS issue does suck. Not being able to host repositories on mounted drives is a bitch, especially at work where all of the machines are locked down and everything is done over NFS. Luckily our sysadmin is fairly understanding and made local partitions for us to use, but the repositories on them still have to be cron'd onto an NFS mount for backup.
There are some noises that you can avoid the data corruption by switching the berkley db for another options, but even the subversion book calls this 'experimental' at the moment.
Not only has he failed to grasp that this construct is just a model of reality, he's gone one step further along the road to idiocy. He manages to use 'self-referential' in his 'proof' but doesn't get as far as 'frame-of-reference'. Surely these two concepts are similar enough that learning one (in the context of physics) would lead you to the other? His whole argument is that dt/dt makes no sense, but dt'/dt would. And given that you have to measure velocities with respect to some frame of reference in relativity anyway, it's kind of obvious.
No, your comparison is completely wrong. Islam is not a race, it's a religion. We limit freedom of speech to prevent racism for a very good reason - you are born with a colour of skin, and there is nothing that you can do to change that. Religion is a choice. When people make choices they are based on opinions, and in a free democracy we must be able to ridicule and disagree with other peoples opinions. If we cross that line even once, and say that some opinions are so important that we are not even allowed to satirise them, then it is a slippery slope and the end of democracy. Without a free and informed debate on the things that people believe, we have started to restrict what people are allowed to believe in.
Slashdot Mirror
I'm doing a postdoc in CompSci now, and its nowhere near a six-figure salary. And knowing the payrates in quite a few countries for postdocs I call bullshit. Link a single position that pays six figures for a postdoc.
You've hit the nail on the head when you say application*s*. Given that this article is a dupe, and we all bashed ajaxWrite last time it was up, I'm suprised that nobodies mentioned the other apps. This guys plan was to realise a new app every week. So far he's got a sketch program, and a something for videoediting. The sketch program (like the ajaxWrite) is exactly the kind of simple programing assignment that you would get somebody to do to learn a new language. Its not drowning in features, although of course he compares it to Inkscape. I didn't look at the videoediting as thats not really my thing.
One amusing thing about his hype machine is that he has a claim that the idea for the word processor came from another guy, and that he injected $50000 as VC to get it off the ground. Odd thing is three weeks ago he made that claim about the word processor. It seems that this claim shifts through the applications as they are written. That is some impressive use of 200,000 quarters on a piece of string trick.
AnalogDiehard doesn't want to read the article, and she wants to draw attention away from the fact. This is just a pre-canned comment with no relevence to the article as a tactic to lend legitimacy to her karma-whoring.
That's one of the most insightful posts that I've ever read on this site. A really good breakdown of what it is to apply the knowledge behind discrete math to a problem, rather than solve problems in discrete math.
And in response to the other reply above, yes, a tree is a type of graph but you've missed the point. If an algorithm can be coded over trees rather than graphs, then it should be, as it will simplify the design, and reduce the complexity.
Is your post written in some natural language form of lisp?
Dogma : An authoritative principle, belief, or statement of ideas or opinion, especially one considered to be absolutely true.
Please note the difference between this and your definition of dogma. As regards a cult, here is a loose, but working definition;
1. An organisation that uses dogma to control the beliefs of its followers
2. An organisation that refuses to discuss this dogma to people that have not been indoctrinated.
3. An organisation that demands that all of its followers ungo an indoctrination to bind them to the organisation.
4. - 12. may vary depending on the cult.
I notice that you've dodged the question of what 12 steps you adhere do, and that you ask us to suspend disbelieve that it's not psuedoreligous or a cult without showing us the program. That is how your behaviour is dogmatic.
No, as I stated, you acted dogmatically in the way that you flamed the original poster. I never used the phrase "incontrovertably true", so either you're confused about which post was mine, or about the meaning of the word dogmatic. The generic 12-steps that you've linked to are psuedo-religous. I asked for your specific flavour as I wondered why you acted like a cultist when you claimed that you weren't. Saying that your fellowship's traditions include not disclosing the traditions certainly sounds likes the actions of a cult.
You've responded quite harshly to an insightful post on the subject. The OP made a comment that 12-step programs are psuedo-religious and cultish, but in defending them you've acted in the dogmatic way that those criticisms suggest. Purely out of interest, and not to start a flamewar here, what are the 12 steps of your program?
You are simplifying many aspects of the problem. It is possible that we have a working example of a solution, and yet we cannot create another. In order to build upon a previous example we need to reverse engineer the working solution. This is not a simple task. We have people in our department researching computational neuroscience and I've seen some of their seminars. The problem is so complex that there is no other problem that we can even use as a metaphor for it.
We have open problems in different fields of science that have resisted all attempts to solve them. The problem with Kurzweil (and his supporters) is that they *assume* that AI will fall into the solvable category if we throw enough resources at it. This assumption is not necessarily true.
Simply saying that an increase in technology will allow us to build upon existing biological systems is a fallacy. Even ignoring the ethical issues of such experimentation on sentient (or near-sentient) beings, there is the technological challenge.
Assume that we are lucky once, and manage to perform a modification to a biological brain. This does not *automatically* lead to further increases. Even to the increased intelligence of the modified organism, the problem may still be too complex to be tractable. In which case the next advance is still waiting for another lucky strike. And given the complexity of the organism the odds of such a lucky strike are small.
One analogy is the difference between the increase in silicon fabrication technology, and the increase in software development. One is a simple problem that we can throw engineering resources at in order to gain incremental increases. The other is largely intractable, and when progress does come it does so in sporadic spurts of growth. The result is an exponential increase in chip performance (so far) against a linear (at best) increase in software performance.
Assumptions about which class of problem the solution to AI belongs to have no merit at all.
Err, where are we producing human level intelligences today? Other than real actual people... AI is a genuinely hard problem. There is no guarantee that it will be solved. Whilst some problems have fallen to human endevour there are other problems that have not. Try looking at a list of open problems in number theory sometime. These have had some of the brightest minds of the past 500 years try and fail to solve them. A workable AI may fall into the same category.
I think that the GP has a fair point. It's not that the GPU and the CPU are comming together, it is more like there are two separate roles. In one role you want a vector that can run *lots* of numerical code with simple control flow. Graphics, physics and a bunch of other stuff can be executed on this processor. But for the CPU you want something that can handle complex control flow. As complex control flow is tricky (pipeline stalls, branch prediction, hiding memory latency...) the easiest way to do it is through multi-threading. So you have one chip that does all of the parallel processes in lockstep taking advantage of the simple control flow to predict ahead. And there is a different chip that does all of the processes at separate rates using multitasking to hide the latency, rather than prediction.
Thanks for the link. There's no authorative reference in there, but it does the match the figures that I found from a google. The main point in the article is quite interesting - that while we have fairly good population figures for humans there is no definite figure for the number of cars. I guess that your 750mil figure was fairly accurate after all... ;^)
I don't know what you googled as you didn't provide a URL but your 750 million cars figure sounds a bit fishy. While your overall argument that most people don't own a car, the population in America is ~325 million, and Europe is ~750 million. Both have a saturated car market with multiple ownership per household. That would be around 1billion cars alone. China, India and other places to a decreasing degree are buying cars like crazy. Their ownership may be much lower but there is ahug base of old Soviet vehicles and obsolete western models across the world. I'd guess (hand-waving) that there are 1.5 - 2 billion cars in circulation...
My ISP is Plusnet, the one mentioned in the piece. Their line of argument is that they have never sold an "unlimited" packaged. Furthermore their literature does state clearly that it is a contended product with a contention ration of (I forget exactly) about 30:1. So other users on the same contended line do have an effect on your usage. And since they started shaping the p2p traffic performance has picked up massively for everyone else.
There is no password protection on most mobile phones, they just have pin entry when they're first switched on. That is no good if the phone is already powered up.
Cancellation would not solve this problem. You are kind of missing the point. The journalist that the GP is referrering to showed how easy it is to set up tracking on somebody elses phone. Once setup there is *no* indication that the service is active. Thus you wouldn't know to cancel it, or who with.
Happy meal first.
Order a Big Mac later.
Fries with that?
My god its late and I need to go home
That's quite impressive, I'll have to have a look at their site later and see how they are doing it. It doesn't appear to give full spatialisation, I'm guessing that it's supposed to sound like it's a sound source rotating around my head. That could be down to the fullsize headphones that I'm wearing at the moment. Some of it sounds similar to the effects that we created where there is some element of depth, but it just sounds like it's panning right-left across the back of your shoulders. There was also a small amount of up/sound spatialisation, but not very strong.
With those quibbles, it's still the best demo that I've heard yet. Thanks for the link.
I call bullshit. We did some research into this a few years ago and producing spatialised sound through eyebud headphones is not realistic. Each person has a unique head-transform based on the curves of the earlobes amongst other things. What is feasible is telling people that it is spatialised audio and then allowing a suggestive / placebo effect to work. As they're going for surround rather than full spatialisation it is also possible that they've just boosted the bitrate on the bass frequencies so that it sounds less tiny. This gives a 'fuller volume' effect which most people will accept as surround if you suggest it.
Sorry for the rant, but we spent a long time dicking around with spatialised audio until we did 'blind' listening trials to confirm that it was bollocks. Also, if they do get some kind of spatialisation working then it would require earbuds as oppoosed to fullsize headphones. The fullsize ones have a sound source outside of the ear canal which means that the outer ear picks up its own localisation cues.
Reading the dude's list of suggestions I don't think that he's buying much of their product at all. At least, nothing after the initial loss leader...
That's good to hear. Are there any reports of people using it over NFS? One of the problems with subversion used to be that once a repository was corrupt you were screwed. I think that the repair tools have improved a lot since then, but I'd still be a bit dubious until lots of other people had already tested it ;^)
Non-root users don't have access to the server for security reasons. And the sysadmin doesn't want any non-critical processes running on the box. Basically it does nothing but NFS.
SVN does rock, and I use it over CVS any chance that I get ... but the NFS issue does suck. Not being able to host repositories on mounted drives is a bitch, especially at work where all of the machines are locked down and everything is done over NFS. Luckily our sysadmin is fairly understanding and made local partitions for us to use, but the repositories on them still have to be cron'd onto an NFS mount for backup.
There are some noises that you can avoid the data corruption by switching the berkley db for another options, but even the subversion book calls this 'experimental' at the moment.
Religion is always a choice. I didn't say what the alternative choice was ...
Not only has he failed to grasp that this construct is just a model of reality, he's gone one step further along the road to idiocy. He manages to use 'self-referential' in his 'proof' but doesn't get as far as 'frame-of-reference'. Surely these two concepts are similar enough that learning one (in the context of physics) would lead you to the other? His whole argument is that dt/dt makes no sense, but dt'/dt would. And given that you have to measure velocities with respect to some frame of reference in relativity anyway, it's kind of obvious.
No, your comparison is completely wrong. Islam is not a race, it's a religion. We limit freedom of speech to prevent racism for a very good reason - you are born with a colour of skin, and there is nothing that you can do to change that. Religion is a choice. When people make choices they are based on opinions, and in a free democracy we must be able to ridicule and disagree with other peoples opinions. If we cross that line even once, and say that some opinions are so important that we are not even allowed to satirise them, then it is a slippery slope and the end of democracy. Without a free and informed debate on the things that people believe, we have started to restrict what people are allowed to believe in.