Have you got any idea how difficult it is to refute an experimental outcome, at least in the less exact sciences? It's not only that you can create a gazillion possible deviations between your set-up and the one from the article (making direct comparison difficult), you will also need to run it with a pretty large subject group if you want to have enough power (making it expensive and time consuming), and then you're going to have problems publishing your article (reviewers and editors don't like null effects). In short, there is no profit in it. Most people, and researchers are people, are in it for the money, prestige, whatever, and replicating a study generally doesn't get you funding, prestige, publications. So guess what happens? The world, at least the part that does experimental psychology, gets stuck with 90% junk publications. And that's being conservative.
They haven't solved the natural language problem, clearly. Asking "What is the average length of a human" leads to the interpretation "mean, human(animal) body length", which is the correct thing. Then it says it doesn't have enough data. The question "how tall is the average human", which is essentially the same, leads to the well-known "Wolfram|Alpha isn't sure what to do with your input.".
But, at least it knows where the centre of the earth is. I asked it: "what is the centre of earth", and it told me:
Input interpretation: Earth (planet) | location Centre,Alabama,United States
So, there you have it, on highest authority: Alabama.
O noes! It turns out that it doesn't understand "centre" to mean the same as "center". The question "what is the center of earth" leads to 0deg S, 0deg W, which is, unfortunately, wrong and not even funny...
It is the Supreme Court's task, at least in other countries, to set the rules for once and for all, not to address incidents. Based on the available evidence, it has decided that such voting problems are unacceptable *in general*.
As a researcher with more than 10 years of experience in processing natural language in various ways, I can only say that I don't believe it until I see it. Of course it is possible to generate reasonable answers to a fairly large set of questions if you throw enough resources at the development, but I doubt this system will be able to give a reliable answer to any interesting question, unless of course it makes the user disambiguate his/her question to the point where he/ has answered it himself. That would be cool for me, but not really a crowd pleaser.
Experimental psychology is not easy (IAAECP). Interpreting the outcomes is even more difficult, so it seems. In the abstract, the authors write "The studies also showed that players high in trait aggression were more likely to prefer or value games with violent contents, even though violent contents did not reliably enhance their game enjoyment or immersion." So, violent people like violent games, and nobody needs violence in order to enjoy gaming.
On the other hand, this study cannot assess the effects of playing many hours per day over a long period. So, the conclusion might just as well be: ban violent games, since no-one needs it and their effects cannot be determined, but we do know violent people like them.
Let's see what this does to my karma...
Or this quote: "It is ridiculous claiming that video games influence children. For instance, if Pac-man affected kids born in the eighties, we should by now have a bunch of teenagers who run around in darkened rooms and eat pills while listening to monotonous electronic music."
I would strongly recommend starting with a text book on Cognitive Psychology, or reading it in parallel. AI tends to overlook the fact that intelligence is a human trait, not the most efficient algorithm for solving a logic puzzle. Anderson's book can be recommended: http://bcs.worthpublishers.com/anderson6e/default.asp?s=&n=&i=&v=&o=&ns=0&uid=0&rau=0.
You can do all sorts of word puzzle games with regexps, provided you've got a decent word list to begin with. E.g., find all words that do only contain vowels: ^[aeiou]+$ (this is of course language dependent and in English would fail to recognize the -y- as a vowel, but that's not necessary in this case). Or try to find good names for that killer app you're developing. Say you call it Super Video Program. Then you could search for words containing the letters s.*v.*p.* and come up with "silver plate". Ok, not a great example, but you know what I mean. Or you want to spell "access" differently. We know that a and e can sometimes be interchanged; x, cc, ks, xc, and cs; and s, ss and th. Then you come up with a regexp like '^[ae](x|cc|ks|cs|xc|xs|qs)[ae](s|ss|th)$' and find access, axes, excess, and exes.
And you can search for those impossible entries in cross words, for which you will only need ^, . and $.
Amen to that. It has been said before, of course, but never in such a clear, yet funny way.
Open Source is not and cannot be the answer to everything. Open Source is mainly interesting for developers. There is a lot of it out there and almost all of it has a very limited potential. I guess that is because most developers cannot complete the product to a professional level and then support it for a longer period. There is no incentive to make them do so, apart from a bit of recognition. Recognition comes from other developers; users only bitch, and other developers can actually help you sort out your problems. Consequently, developing for other developers is much more rewarding. Hence the success of the Linux kernel and GNU tools.
Anything at a higher level is not aimed at developers, but at users. The only reason to make them competitive is developer incentive. Money, if you like. Open source does not make you any money though, and other rewards are scarce. A few companies have stepped in to support open source development (think OpenOffice), which makes development move a lot faster, and even then they don't get it right. E.g. OpenOffice's presentation package consumes much more resources than PowerPoint (under OSX, at least).
This is not a complete analysis of why Open Source is not the panacea some believe it to be. But while the situation stays as it is, $GEEKY_FEATURE is the way to go.
And anyway, in OSX you can tell the Finder to add songs to iTunes if they appear in a specific directory. Would that be a download directory, perhaps?
You have probably never heard of random processes. From this data you cannot rule out the possibility that a radio signal would almost never interfere, but just very, very occasionally would. If the chances of such a fluke would be 1 in 100 million hours, that would still warrant total prohibition of wireless equipment, since with 11000 flights per day there is one such event per year (according to some, there are more than 25000 flights per day in the US alone).
And what do you think Qantas is going to retort? That the malfunction was caused by radio signals from passengers' electronic devices. Duh! Look at it. A computer starts spewing "random data". That can only be caused by random radio waves from random clicking with a wireless mouse. No, in a few months time, everyone bringing a wireless mouse on board will be considered a terrorist.
I work on both PC and Mac (and some linux variant from time to time) and the dock and the task bar are quite different things. E.g., you can drag an application to the dock from anywhere and it will become a kind of shortcut there. Applications that are not by default in the dock appear there when launched. A little thingy under an application shows whether it is running or not. You can also drag documents, folders, etc. in the dock and they can do different things there (just open, or show a list of contents, or a hierarchical menu). You can click on a dock icon and get a menu of application specific commands (e.g. a browser can show the most visited urls, an e-mail app can offer get-new-mail/compose, etc.). And applications can update their icon with little badges to show progress, new mail received, the date, whatever. And you can change its position, size, and magnification. That's pretty different from the Windows task bar where every open document gets an icon+text entry, and which has a start button, clock, etc.
Oh, come on. Excel has one huge disadvantage: it knows no structure apart from a matrix. If your data consists of a time series with different measurements in different columns, you can do fine in Excel (apart from some obvious problems that have been pointed out; one that was strangely overlooked was the limited number of rows and columns).
If your data is structured or develops continuously over rows and columns alike or has more than two dimensions, you're screwed. E.g., a simple setup with time, condition, subject index and outcome is horribly messy in Excel. Then try doing EEG or fMRI data analysis with that...
Well, the different opinions above already show there is not an easy interpretation, but it clearly says "functional". In neurospeak, that means: we're not interested in neural firing patterns or genetic expression of FOXP21 in a rat's hippocampus while learning to navigate a maze with electric shock plates during different phases of starvation. Instead, we're interested in language, vision, motor planning, memory, etc. The "mathematically consistent" bit is to me (IAACNCM: I Am A Computational Neuro-Cognitive Modeler) nonsense. If you can run it on a computer, it should be consistent enough. Predictive doesn't mean anything: if you've got a good model, it is predictive by itself (modulo hardware requirements...).
So yes, this is a reformulation in the vagest terms of long running AI and cognitive psychology research, which DARPA is funding heavily since the 70s.
I call BS. I've got a 6-yr old and neither she nor her friends nor any nephews/nieces have ever mistaken TV for reality. At age 3, they know that what they see on TV isn't real, especially cartoons. And the utterance "Why? We'd just pop up again." seems to complex for a 3 yr old: it requires searching for the reasoning behind your statement and a way to refute it, and sounds too verbose.
How about this scenario: 1. Google rolls out browser, people start using it 2. Google rolls out web apps, that only run their browser. People love the apps. 3. Everyone gets addicted to Google's browser. 4.... 5. PROFIT!
And step 4 would be: once you're forced to run their browser, they'll start "upgrading" it with all sort of features, that monitors your browsing, push more ads, searches your local files. All in the best interest of the customer, of course.
Perhaps you could point this out to the people at Language Log (when it comes up again): some of them studied Chinese, so they look at it from another side, and they certainly don't have the "native" view.
Check out Language Log. They do not only have even funnier examples, but also try to analyze the source of the error, as well as translation problems in other languages. The latest installment in the series of Chinese-English mistranslations is The Sichuan's hair blood is prosperous, or check the whole category: Lost in Translation.
They also collect "Cupertinos", errors introduced by spelling checkers, or have you ever heard of US presidential candidates Barrack Abeam and John moccasin? It's a great log for anyone interested in language.
I finally begin to understand why you're labelled "Anonymous Coward" when you don't use your ID to leave a comment. It is simply appropriate.
And you Sir (the possibility that you're actually a woman is too remote), are a prime example of a coward. Have you got any idea what it is like to have children? How painful a story like this can be? Because we are not text processing robots. We try to imagine what happened and find an explanation. You were trying to give an explanation, as a matter of fact, and a "funny" one at it, just to make your own, disquieting worries go away. Perhaps you thought of your father or mother. How they were when you were just three. You looked at your childhood's pictures. Then you imagined one of your parents killing the other one and then you. Disturbing, isn't it? Is that why you tried to make a joke?
Did you ever look into the eyes of a three year old after you punished him/her? I'm talking about three minutes alone here, nothing more serious than that. How even that can affect them? And how much they will assure you afterwards that they love you very much and will never do it again? Did you see the picture in the NY Times of the Iraqi girl after her parents being killed by (American) gun fire? Do you still think you were funny?
Did no one notice the spelling error "rather that" in the summary? And started swearing at author? Amazing! This Internet thing you hear so much about these days might turn out to be a polite discussion forum after all.
One addition: if you multiply 0.00005309 by 6,000,000, you get 318, which is the number of duplicate matches you can expect in the 6 million database. So, it's not really surprising that the Arizona lab did find a (near) match. Still, the chances of a false match for an individual are 0.00005309, so the question about its effectiveness/usefulness is: how many profiles are compared against the database per year?
Hold it right there, cowboy. This is a perfectly legitimate study, and not the first one either (although the first one on numbers in Piraha). I know Ted Gibson and I can assure you he's a respectable scientist. Do you really think the reviewers of the article (it has been published in a very decent journal, actually) would not have caught an obvious fraud?
I have a Asus WL530g which plays badly with my wife's Acer laptop. Rebooting the router seemed to help sometimes. However, I work on a Mac laptop, and it never has had this problem. Even when the Acer loses the connection, the Mac keeps working. So it's obvious that the error is on the Acer's side. I'm pretty sure it's something called XPCOM, but it's hard to be sure (XPCOM might lock up because of a deeper problem).
Anyway, my point is: router problems are sometimes not in the router, but in the client...
Diffusion imaging is not new and the problems are well-known. Basically, you try to estimate a flow by sampling a lot of points and connect them if they go in (more or less) the same direction. If a flow (in this case a fiber) changes direction too much between sample points, you make a mistake. Also, averaging over 5 people can lead to strange errors, but I guess the authors are competent enough to avoid those pitfalls.
The thing about the hub isn't that interesting: don't think all traffic passes through it. And these fiber tracts are not supposed to do much processing anyway. It does strike me that the map is asymmetrical.
One of the authors is quoted as saying: "This means that if we know how the brain is connected we can predict what the brain will do." That should probably be: from knowing the structure we can partially predict the BOLD response (what you measure in fMRI). So much for journalism.
Ok, I'll byte to. I did my degree in theoretical CS (automata, power of formalisms, complexity, etc.). Great topics. But that can get you a research job. If you want to design, do DBA, or manage programmers, you're going to have a hell of a time if you can't actually program. You really need to have felt the difficulties of building a large scale system before you can be successful at it, and preferably more than once.
You're right: a computer scientist is not a programmer. Instead, he (or rarer: she) should be way above a programmer, but should none the less understand programming and the ridiculous problems that arise in real life programming. Otherwise you're just a side-line theoretic.
About monkeys programming: I think you'll find that their mental skills are not really apt. And teaching fundamentals is not the same as experience. You don't know portability until you've hit your head a few times, to name just one practical example. To write efficient programs, expandable architectures, etc., that's also you can only appreciate after having done it a few times. Simple example: there's no point in implementing the most efficient search algorithm if you're only going to sort 20 data points or so. You have to know when your design is good enough. As someone whom I hold in high esteem said: the perfect is the enemy of the good.
Brain imaging, huh? With one of those nice, big, expensive MRI scanners? All it tells you that the BOLD response in a certain area reduces. This is not always associated with less "attention" or "processing". It is well known that if you repeat a picture, a word or a sound (correct term: stimulus), the BOLD response decreases while reaction time is faster. By that account you could conclude that listening improves reaction time.
The truth is that we know very little about the meaning of the BOLD response. While they might inform us about theories about brain structure, BOLD responses should not be taken as evidence for a theory of behaviour.
Slashdot Mirror
Have you got any idea how difficult it is to refute an experimental outcome, at least in the less exact sciences? It's not only that you can create a gazillion possible deviations between your set-up and the one from the article (making direct comparison difficult), you will also need to run it with a pretty large subject group if you want to have enough power (making it expensive and time consuming), and then you're going to have problems publishing your article (reviewers and editors don't like null effects). In short, there is no profit in it. Most people, and researchers are people, are in it for the money, prestige, whatever, and replicating a study generally doesn't get you funding, prestige, publications. So guess what happens? The world, at least the part that does experimental psychology, gets stuck with 90% junk publications. And that's being conservative.
They haven't solved the natural language problem, clearly. Asking "What is the average length of a human" leads to the interpretation "mean, human(animal) body length", which is the correct thing. Then it says it doesn't have enough data. The question "how tall is the average human", which is essentially the same, leads to the well-known "Wolfram|Alpha isn't sure what to do with your input.".
But, at least it knows where the centre of the earth is. I asked it: "what is the centre of earth", and it told me:
Input interpretation:
Earth (planet) | location Centre,Alabama,United States
So, there you have it, on highest authority: Alabama.
O noes! It turns out that it doesn't understand "centre" to mean the same as "center". The question "what is the center of earth" leads to 0deg S, 0deg W, which is, unfortunately, wrong and not even funny...
It is the Supreme Court's task, at least in other countries, to set the rules for once and for all, not to address incidents. Based on the available evidence, it has decided that such voting problems are unacceptable *in general*.
As a researcher with more than 10 years of experience in processing natural language in various ways, I can only say that I don't believe it until I see it. Of course it is possible to generate reasonable answers to a fairly large set of questions if you throw enough resources at the development, but I doubt this system will be able to give a reliable answer to any interesting question, unless of course it makes the user disambiguate his/her question to the point where he/ has answered it himself. That would be cool for me, but not really a crowd pleaser.
Experimental psychology is not easy (IAAECP). Interpreting the outcomes is even more difficult, so it seems. In the abstract, the authors write "The studies also showed that players high in trait aggression were more likely to prefer or value games with violent contents, even though violent contents did not reliably enhance their game enjoyment or immersion." So, violent people like violent games, and nobody needs violence in order to enjoy gaming.
On the other hand, this study cannot assess the effects of playing many hours per day over a long period. So, the conclusion might just as well be: ban violent games, since no-one needs it and their effects cannot be determined, but we do know violent people like them.
Let's see what this does to my karma...
Or this quote: "It is ridiculous claiming that video games influence children. For instance, if Pac-man affected kids born in the eighties, we should by now have a bunch of teenagers who run around in darkened rooms and eat pills while listening to monotonous electronic music."
I would strongly recommend starting with a text book on Cognitive Psychology, or reading it in parallel. AI tends to overlook the fact that intelligence is a human trait, not the most efficient algorithm for solving a logic puzzle. Anderson's book can be recommended: http://bcs.worthpublishers.com/anderson6e/default.asp?s=&n=&i=&v=&o=&ns=0&uid=0&rau=0.
You can do all sorts of word puzzle games with regexps, provided you've got a decent word list to begin with. E.g., find all words that do only contain vowels: ^[aeiou]+$ (this is of course language dependent and in English would fail to recognize the -y- as a vowel, but that's not necessary in this case). Or try to find good names for that killer app you're developing. Say you call it Super Video Program. Then you could search for words containing the letters s.*v.*p.* and come up with "silver plate". Ok, not a great example, but you know what I mean. Or you want to spell "access" differently. We know that a and e can sometimes be interchanged; x, cc, ks, xc, and cs; and s, ss and th. Then you come up with a regexp like '^[ae](x|cc|ks|cs|xc|xs|qs)[ae](s|ss|th)$' and find access, axes, excess, and exes.
And you can search for those impossible entries in cross words, for which you will only need ^, . and $.
Amen to that. It has been said before, of course, but never in such a clear, yet funny way.
Open Source is not and cannot be the answer to everything. Open Source is mainly interesting for developers. There is a lot of it out there and almost all of it has a very limited potential. I guess that is because most developers cannot complete the product to a professional level and then support it for a longer period. There is no incentive to make them do so, apart from a bit of recognition. Recognition comes from other developers; users only bitch, and other developers can actually help you sort out your problems. Consequently, developing for other developers is much more rewarding. Hence the success of the Linux kernel and GNU tools.
Anything at a higher level is not aimed at developers, but at users. The only reason to make them competitive is developer incentive. Money, if you like. Open source does not make you any money though, and other rewards are scarce. A few companies have stepped in to support open source development (think OpenOffice), which makes development move a lot faster, and even then they don't get it right. E.g. OpenOffice's presentation package consumes much more resources than PowerPoint (under OSX, at least).
This is not a complete analysis of why Open Source is not the panacea some believe it to be. But while the situation stays as it is, $GEEKY_FEATURE is the way to go.
And anyway, in OSX you can tell the Finder to add songs to iTunes if they appear in a specific directory. Would that be a download directory, perhaps?
You have probably never heard of random processes. From this data you cannot rule out the possibility that a radio signal would almost never interfere, but just very, very occasionally would. If the chances of such a fluke would be 1 in 100 million hours, that would still warrant total prohibition of wireless equipment, since with 11000 flights per day there is one such event per year (according to some, there are more than 25000 flights per day in the US alone).
And what do you think Qantas is going to retort? That the malfunction was caused by radio signals from passengers' electronic devices. Duh! Look at it. A computer starts spewing "random data". That can only be caused by random radio waves from random clicking with a wireless mouse. No, in a few months time, everyone bringing a wireless mouse on board will be considered a terrorist.
I work on both PC and Mac (and some linux variant from time to time) and the dock and the task bar are quite different things. E.g., you can drag an application to the dock from anywhere and it will become a kind of shortcut there. Applications that are not by default in the dock appear there when launched. A little thingy under an application shows whether it is running or not. You can also drag documents, folders, etc. in the dock and they can do different things there (just open, or show a list of contents, or a hierarchical menu). You can click on a dock icon and get a menu of application specific commands (e.g. a browser can show the most visited urls, an e-mail app can offer get-new-mail/compose, etc.). And applications can update their icon with little badges to show progress, new mail received, the date, whatever. And you can change its position, size, and magnification. That's pretty different from the Windows task bar where every open document gets an icon+text entry, and which has a start button, clock, etc.
Oh, come on. Excel has one huge disadvantage: it knows no structure apart from a matrix. If your data consists of a time series with different measurements in different columns, you can do fine in Excel (apart from some obvious problems that have been pointed out; one that was strangely overlooked was the limited number of rows and columns).
If your data is structured or develops continuously over rows and columns alike or has more than two dimensions, you're screwed. E.g., a simple setup with time, condition, subject index and outcome is horribly messy in Excel. Then try doing EEG or fMRI data analysis with that...
Well, the different opinions above already show there is not an easy interpretation, but it clearly says "functional". In neurospeak, that means: we're not interested in neural firing patterns or genetic expression of FOXP21 in a rat's hippocampus while learning to navigate a maze with electric shock plates during different phases of starvation. Instead, we're interested in language, vision, motor planning, memory, etc. The "mathematically consistent" bit is to me (IAACNCM: I Am A Computational Neuro-Cognitive Modeler) nonsense. If you can run it on a computer, it should be consistent enough. Predictive doesn't mean anything: if you've got a good model, it is predictive by itself (modulo hardware requirements...).
So yes, this is a reformulation in the vagest terms of long running AI and cognitive psychology research, which DARPA is funding heavily since the 70s.
I call BS. I've got a 6-yr old and neither she nor her friends nor any nephews/nieces have ever mistaken TV for reality. At age 3, they know that what they see on TV isn't real, especially cartoons. And the utterance "Why? We'd just pop up again." seems to complex for a 3 yr old: it requires searching for the reasoning behind your statement and a way to refute it, and sounds too verbose.
How about this scenario: ...
1. Google rolls out browser, people start using it
2. Google rolls out web apps, that only run their browser. People love the apps.
3. Everyone gets addicted to Google's browser.
4.
5. PROFIT!
And step 4 would be: once you're forced to run their browser, they'll start "upgrading" it with all sort of features, that monitors your browsing, push more ads, searches your local files. All in the best interest of the customer, of course.
Perhaps you could point this out to the people at Language Log (when it comes up again): some of them studied Chinese, so they look at it from another side, and they certainly don't have the "native" view.
Check out Language Log. They do not only have even funnier examples, but also try to analyze the source of the error, as well as translation problems in other languages. The latest installment in the series of Chinese-English mistranslations is The Sichuan's hair blood is prosperous, or check the whole category: Lost in Translation.
They also collect "Cupertinos", errors introduced by spelling checkers, or have you ever heard of US presidential candidates Barrack Abeam and John moccasin? It's a great log for anyone interested in language.
I finally begin to understand why you're labelled "Anonymous Coward" when you don't use your ID to leave a comment. It is simply appropriate.
And you Sir (the possibility that you're actually a woman is too remote), are a prime example of a coward. Have you got any idea what it is like to have children? How painful a story like this can be? Because we are not text processing robots. We try to imagine what happened and find an explanation. You were trying to give an explanation, as a matter of fact, and a "funny" one at it, just to make your own, disquieting worries go away. Perhaps you thought of your father or mother. How they were when you were just three. You looked at your childhood's pictures. Then you imagined one of your parents killing the other one and then you. Disturbing, isn't it? Is that why you tried to make a joke?
Did you ever look into the eyes of a three year old after you punished him/her? I'm talking about three minutes alone here, nothing more serious than that. How even that can affect them? And how much they will assure you afterwards that they love you very much and will never do it again? Did you see the picture in the NY Times of the Iraqi girl after her parents being killed by (American) gun fire? Do you still think you were funny?
Step forward, admit the joke was in bad taste.
Did no one notice the spelling error "rather that" in the summary? And started swearing at author? Amazing! This Internet thing you hear so much about these days might turn out to be a polite discussion forum after all.
One addition: if you multiply 0.00005309 by 6,000,000, you get 318, which is the number of duplicate matches you can expect in the 6 million database. So, it's not really surprising that the Arizona lab did find a (near) match. Still, the chances of a false match for an individual are 0.00005309, so the question about its effectiveness/usefulness is: how many profiles are compared against the database per year?
Hold it right there, cowboy. This is a perfectly legitimate study, and not the first one either (although the first one on numbers in Piraha). I know Ted Gibson and I can assure you he's a respectable scientist. Do you really think the reviewers of the article (it has been published in a very decent journal, actually) would not have caught an obvious fraud?
I have a Asus WL530g which plays badly with my wife's Acer laptop. Rebooting the router seemed to help sometimes. However, I work on a Mac laptop, and it never has had this problem. Even when the Acer loses the connection, the Mac keeps working. So it's obvious that the error is on the Acer's side. I'm pretty sure it's something called XPCOM, but it's hard to be sure (XPCOM might lock up because of a deeper problem).
Anyway, my point is: router problems are sometimes not in the router, but in the client...
Diffusion imaging is not new and the problems are well-known. Basically, you try to estimate a flow by sampling a lot of points and connect them if they go in (more or less) the same direction. If a flow (in this case a fiber) changes direction too much between sample points, you make a mistake. Also, averaging over 5 people can lead to strange errors, but I guess the authors are competent enough to avoid those pitfalls.
The thing about the hub isn't that interesting: don't think all traffic passes through it. And these fiber tracts are not supposed to do much processing anyway. It does strike me that the map is asymmetrical.
One of the authors is quoted as saying: "This means that if we know how the brain is connected we can predict what the brain will do." That should probably be: from knowing the structure we can partially predict the BOLD response (what you measure in fMRI). So much for journalism.
Ok, I'll byte to. I did my degree in theoretical CS (automata, power of formalisms, complexity, etc.). Great topics. But that can get you a research job. If you want to design, do DBA, or manage programmers, you're going to have a hell of a time if you can't actually program. You really need to have felt the difficulties of building a large scale system before you can be successful at it, and preferably more than once.
You're right: a computer scientist is not a programmer. Instead, he (or rarer: she) should be way above a programmer, but should none the less understand programming and the ridiculous problems that arise in real life programming. Otherwise you're just a side-line theoretic.
About monkeys programming: I think you'll find that their mental skills are not really apt. And teaching fundamentals is not the same as experience. You don't know portability until you've hit your head a few times, to name just one practical example. To write efficient programs, expandable architectures, etc., that's also you can only appreciate after having done it a few times. Simple example: there's no point in implementing the most efficient search algorithm if you're only going to sort 20 data points or so. You have to know when your design is good enough. As someone whom I hold in high esteem said: the perfect is the enemy of the good.
Brain imaging, huh? With one of those nice, big, expensive MRI scanners? All it tells you that the BOLD response in a certain area reduces. This is not always associated with less "attention" or "processing". It is well known that if you repeat a picture, a word or a sound (correct term: stimulus), the BOLD response decreases while reaction time is faster. By that account you could conclude that listening improves reaction time.
The truth is that we know very little about the meaning of the BOLD response. While they might inform us about theories about brain structure, BOLD responses should not be taken as evidence for a theory of behaviour.