Slashdot Mirror

Pawan Sinha runs Project Prakash which goes into rural areas of India where treatment for congenital cataracts is not generally available. They do the surgery, for free, and in some cases ask the recipient whether they would like to contribute to the research program, which tracks how patients learn to see after the surgery [pdf]. The oldest person to receive the surgery was 29, and has had limited recovery of visual acuity. Children under the age of 6 typically have excellent prognoses following the surgery. See Pawan's TED talk here.

Pawan Sinha runs Project Prakash which goes into rural areas of India where treatment for congenital cataracts is not generally available. They do the surgery, for free, and in some cases ask the recipient whether they would like to contribute to the research program, which tracks how patients learn to see after the surgery [pdf]. The oldest person to receive the surgery was 29, and has had limited recovery of visual acuity. Children under the age of 6 typically have excellent prognoses following the surgery. See Pawan's TED talk here.

If it's good enough for MIT, ought to be good enough for everyone.
http://ocw.mit.edu/about/
http://www.opensourcetext.org/
(and many other references)

I know you wouldn't be arrogant enough to try to invent your own encryption algorithm, so don't be dumb enough to try to invent your own authentication protocol.

Follow Microsoft and steal it from the best: http://web.mit.edu/Kerberos/#what_is

Invented by the professors at MIT to secure their own logins from the students at MIT. A.k.a. "forged in the fires of Mount Doom" and so safe from most threats.

No! I noticed that the New York Times is saying that these documents show that the death toll is around 100,000. They do no such thing. They simply report this number of deaths. This is not the same as determining that this number has been killed, since these documents didn't set out to report all deaths. There have been very few comprehensive studies on the actual death toll, but it is much higher (see http://web.mit.edu/CIS/pdf/Human_Cost_of_War.pdf).

I can tell it's truly News for Nerds because I can barely understand what it's saying and it drops causal references to advanced mathematics--the stuff I only wish I'd had the fortitude to study in college.

You more than likely did study this in college. This involves linear algebra, specifically the inversion of matrices and/or solving linear system. Ax=b, where A is an mxn matrix, and x and b are nx1 and mx1 vectors respectively. What we'd like to do is solve for x using x=A^-1 b, where A^-1 is an inverse matrix of some kind. But getting the inverse is a notoriously difficult problem.

In fact, a large reason digital computers were invented at all was so that "large" matrices could be inverted. And by large, I mean 12x12 matrices (keep in mind this was the 1950s). Computer games, mp3 players, spreadsheets and the internet are all quite incidental byproducts of humanities quest to invert ever larger matrices, in less time. Though just about any half way sophisticated piece of software will invert a matrix at some point.

The reason for this is as follows: When human beings want to solve any difficult problem, they approximate it with a linear model and solve the associated linear system. Hence the ability to solve linear systems quickly and accurately is of fundamental importance. This goes double in the modern world as we increasingly rely on the ability of computers to handle ever larger linear system--that is, to invert ever larger matrices.

The biggest problem with matrices, say nxn matrices, is that the time taken for solving them by Gauss elimination goes up as 0(n^3). So while your desktop could probably invert a 1000x1000 matrix in a few seconds, it would take a year invert a million x million matrix, and would take around a million years to invert a billion x billion matrix. (Not that it could store either in memory). While Gauss elimination is a pretty poor algorithm efficiency-wise, this issues are unavoidable for general matrices.

However, most matrices we actually want to invert in practice are "sparse" or "diagonally dominant". The first property means that most of the elements are zero. So in a billion x billion matrix, instead of storing 10^18 numbers, you'd only have to store a few billion. Diagonally dominant means that largest entries are along the diagonal. Both of these mean you can take a lot of--often hueristic--shortcuts in your inversion algorithims to cut down on time.

These researchers claim O(s*log(s)^2) complexity for such systems. I suspect there are probably a lot of O(s^2*log(s)) system or the like anyway, but even still this is a nice improvement. I doubt it's as big a breakthrough--I suspect this is hyped anyway--but if it is an improvement you can kind of compare this to something like the Fast Fourier Transform speedup. Again, I doubt it's that large of an improvement.

I'll finish by mentioning that this all falls under the umbrella of Linear Algebra, which is an absolutely essential skill of higher mathematics, and computational mathematics. Any geeks who prides themselves on their mathematical ability, but doesn't know their linear algebra (linear systems, the four fundamental subspaces, eigenvectors/values, rotation/transformation matrices, etc) shouldn't be holding their skills in such high regard. If you don't know your LA, or have forgotten, one of the best places to start is to watch Gilbert Strang's online lecture series provided by MIT. These lectures are indispensable to anyone working with linear systems. After this, those who need in depth analysis of matrix algorithms should turn to something like "Matrix Algorithms" by G. W. Stewart which goes into more detail and shows the more canonical algorithms. A little knowledge of linear algebra goes a long way in mathematics.

...Or failing all that, you could just open up MATLAB... or Octave. I recommend the educational route first though.

Why do you need to go to college? Here, watch Linear Algebra and Algorithms.

Also there is no remotely advanced math dropped in TFA, though the actual paper of course is cutting edge research.

Why do you need to go to college? Here, watch Linear Algebra and Algorithms.

Also there is no remotely advanced math dropped in TFA, though the actual paper of course is cutting edge research.

Accroding to http://en.wikipedia.org/wiki/TILE64 the Tilera CTO and co-founder is Anant Agarwal. According to http://www.csail.mit.edu/user/723 he is from Madras, India.

Imagine all Indian computer gurus moving back to India, backed by the wealth of Tata (www.tata.com) or the like.

Do you think China and other high focus companies have the right to be scared?

Yet, before then, show me the benchmarks

You might be tickled to learn that there are some (wild-ish) theories that posit "every mathematical abstraction exists", as in, for every concept you can derive from mathematics, it actually exists "somewhere". Look at "mathematical multiverse" here http://space.mit.edu/home/tegmark/crazy.html And Tegmark is not actually a crackpot, just fanciful. :)

Paraphrasing ontologist Bill Clinton: "It depends on your definition of 'exists'". For epistemological questions I refer you to Donald Rumsfeld.

You might be tickled to learn that there are some (wild-ish) theories that posit "every mathematical abstraction exists", as in, for every concept you can derive from mathematics, it actually exists "somewhere". Look at "mathematical multiverse" here http://space.mit.edu/home/tegmark/crazy.html And Tegmark is not actually a crackpot, just fanciful. :)

Actually the project is a testbed for some software algorithms for optimal control of systems in the context of variable power availability (as is the case with solar). Presumably this "smart" controller can achieve significantly higher throughput than a naive approach, for example it can probably optimize the process so that the power-consuming components are operating in their most efficient range over a wide range of input power availability.

The components of the device are all off the shelf items, the component engineering and related issues are not part of the research scope.

The project is better described on the research group's website: http://robots.mit.edu/projects/KFUPM/index.html

Ahem! This was already done in 2009 Myth Busters 2.009

Pump-fed nanofilters are sort of an old idea at this point. The summary leaves off some critical points like how much it costs and how long the filter lasts.

According to the article, it costs $8000, which is a lot for some things but probably accessible for others. Let's just say it's not going to solve the world's water problem overnight, but it might be handy for relief efforts.

Surfing through to the parent MITnews article, we get a bit more information, but it's still lacking anything about how long the system can operate or what its maintenance costs and requirements are. Does it last a week then you're out most of another $8000? Does it require a lot of technical expertise to maintain? It doesn't say...

Most people forget about the basic research performed by most universities, which is absolutely necessary to the academic industry and flows into every aspect of the rest of the world (especially including tech, medical, and military). A good deal of the criticism on the current system comes from a lack of understanding of basic research and its part in academia. While the Wikipedia-style likely has merits for far more than we currently expect (it was equally ill-received when proposed for encyclopedias!), it can't fit into our current paradigm of research universities while retaining the current organization of journals and how they handle submissions (which is another point of contention that needs a serious upgrade of its own).

Therefore, perhaps the part-time lecturer model is preferable as a starting-point. However, due to its for-profit (not to mention anticompetitive and controversial) nature, Phoenix is not an appropriate role model.

Take a look at examples that are already far closer to Wikipedia, like MIT's ESP and Tuft's Experimental College.

http://web.mit.edu/2.009/www/experiments/deathray/10_Mythbusters.html

All this will take up a massive amount of space compared to LFTR and comes with problems of its own.

And we have plenty of space. The National Renewable Energy Lab's Wind Atlas details the wind potential of different regions of the US. The Rocky Mountains alone contain enough potential wind energy to supply all 48 contiguous states with electricity. However that's not all. On the Pacific Coast from British Colombia south through southern California then east to western Texas, there's more. Why during California's rolling blackouts in the early 2000s, there was an idle wind farm in the Mojave capable of generating 10 megawatts per hour. Over on the Atlantic Coast from Maine to Cape Hatteras off the North Carolina coast there are good sites for wind farms. As senator before his death Ed Kennedy was one of the NIMBYs opposing one such wind farm, on Cape Cod. On-shore through the Appalachian Mountains north from Georgia then into Pennsylvania's Poconos and New York's Catskills Mountains, hell all along the Appalachian Trail to Maine, there is good wind potential.

That's just wind, solar adds more. Again according to DOE, just 100 square miles of land in Nevada, that's an area of 10 miles by 10 miles, "could supply all U.S. electricity needs with current (~10%) commercial efficiency rates." But Nevada isn't the place with good solar potential. Now let's go back geothermal. According to an MIT led panel sponsored by DOE geothermal can be a "key U.S. energy source". Here's some info on geothermal in New York state, and more for Minnesota and Wisconsin. I've already mentioned California and Yellowstone, recently there was a discussion of how West Virginia Is Geothermically Active.

With today's technology solar and wind can provide the US's peak electricity, while geothermal and existing natural gas and nuclear power plants supply the baseload until more geothermal capacity and storage is developed.

Falcon

Here's a paper written by a fellow who's now a professor at U of I, Chicago which relates to the topic. The gist is that taxi's in a city were equipped with wifi and opportunistically connected to open access points as they traveled. The article won't revolutionize anything but it's certainly an interesting read and something worthy of building upon. One of the interesting parts is that the taxi-side wifi used a custom written utility to accelerate establishing a connection which didn't bother negotiating transmission speed but rather used a fixed 11Mbps as this was determined to be optimal for the setting.

Is study many languages other than C++. For example, learn Scheme through The Structure and Interpretation of Computer Programs.

As a "one language" developer you will always be fifth-rate, in tools, techniques, and thinking.

However at a gbps, you are really looking at being able to do just about everything someone wants to for any foreseeable future in realtime.

What about holographic video?

Ad that to a few dozen banner ads and we could spam up a gigabit link pretty fast.

At SciFoo 2010, Yoshiuki Sankai of Tsukuba University gave a talk with videos of the varied robotic exoskeleton walking-prosthetics available from his company. The film included many examples of people who had not walked for years standing up and walking with these "legs". You could hear the doctors and nurses watching exclaiming their amazement and sometimes crying. Here is a 2006 biography of Sankai already discussing his exoskeletal robot, first demoed in 2005: http://web.mit.edu/invent/iow/sankai.html Their company page also seems to have more information in English: http://www.cyberdyne.jp/english/index.html

Guy Ben-Ary is an artist who did a residency at the SymbioticA Research Lab at the University of Western Australia and then at the Potter Lab at Georgia Tech. During that time he created a system where a culture of rat brain neurons controlled a robotic pen controller to draw "art". Further, the two components (brain and arm) were geographically separated and communicated across the internet.

MEART: The Semi Living Artist

http://web.mit.edu/shkolnik/www/meart/

http://www.fishandchips.uwa.edu.au/

There is a spiffy ppt presentation spelling all this out.

http://web.mit.edu/22.012/www/presentations/Helium-3%20version%202.ppt

While you're at, maybe just cut the Pell Grants, too.

Read Andrew Hacker on the sorry state of higher education.

http://www.google.com/search?q=andrew+hacker+higher+education

All the grants, loans, and whatnot seem to have done is a 439% increase in higher ed costs, 3x the median family income's increase.

Lifelong learning is an admirable goal which should be opensourced, crowdsourced, and meetup'd.

Unfortunately, the summary as well as the short articles on the web were more or less completely missing the point. The actual paper ( http://pdos.csail.mit.edu/papers/linux:osdi10.pdf ) explains what was done.

Essentially they benchmarked a number of applications, figured out where the bottlenecks were, and fixed them. Some of the things they fixed where done by introducing "sloppy counters" in order to avoid updating a global counter. Others were to switch to more fine-grained locking, switching to per-cpu data structures, and so forth. In other words, pretty standard kernel scalability work. As an aside, a lot of the VFS scalability work seems to clash with the VFS scalability patches by Nick Piggin that are in the process of being integrated into the mainline kernel.

And yes, as the PDF article explains, the Linux cpu scheduler mostly works per-core, with only occasional communication with schedulers on other cores.

So, they found scalability problems in some microbenchmarks. Well, some of the scalability paths cited in the paper will be fixed when Nick Piggin's VFS scalability patchset gets merged. But it's not like you need to rewrite every operative system to scale beyond 48 cores, it's just the typical scalability stuff, and the kind of scalability issues found these days are mostly corner cases (Piggin's VFS being an exception).

It appears that the problem, that affect the available memory in a chip when multiple cores are working on the same chunks of data, is getting worse and may be hitting a peak somewhere in the neighborhood of 48 cores, when entirely new operating systems will be needed, the report says.

Seriously? You picked that over my submission?

I submitted this earlier this morning I guess my submission was lacking. But if you're interested in the original MIT article and the actual paper (PDF):

eldavojohn writes "Multicore (think tens or hundreds of cores) will come at a price for current operating systems. A team at MIT found that as they approached 48 cores their operating system slowed down. After activating more and more cores in their simulation, a sort of memory leak occurred whereby data had to remain in memory as long as a core might need it in its calculations. But the good news is that in their paper (PDF), they showed that for at least several years Linux should be able to keep up with chip enhancements in the multicore realm. To handle multiple cores, Linux keeps a counter of which cores are working on the data. As a core starts to work on a piece of data, Linux increments the number. When the core is done, Linux decrements the number. As the core count approached 48, the amount of actual work decreased and Linux spent more time managing counters. But the team found that 'Slightly rewriting the Linux code so that each core kept a local count, which was only occasionally synchronized with those of the other cores, greatly improved the system's overall performance.' The researchers caution that as the number of cores skyrockets, operating systems will have to be completely redesigned to handle managing these cores and SMP. After reviewing the paper, one researcher is confident Linux will remain viable for five to eight years without need for a major redesign."

I don't know, guess I picked a bad title or something?

Luckily we aren't anywhere near 48 cores and there is some time left to come up with a new Linux (Windows?).

Again, seriously? What does "(Windows?)" even mean? As you pass a certain number of cores, modern operating systems will need to be redesigned to handle extreme SMP. It's going to differ from OS to OS but we won't know about Windows until somebody takes the time to test it.

It appears that the problem, that affect the available memory in a chip when multiple cores are working on the same chunks of data, is getting worse and may be hitting a peak somewhere in the neighborhood of 48 cores, when entirely new operating systems will be needed, the report says.

Seriously? You picked that over my submission?

I submitted this earlier this morning I guess my submission was lacking. But if you're interested in the original MIT article and the actual paper (PDF):

eldavojohn writes "Multicore (think tens or hundreds of cores) will come at a price for current operating systems. A team at MIT found that as they approached 48 cores their operating system slowed down. After activating more and more cores in their simulation, a sort of memory leak occurred whereby data had to remain in memory as long as a core might need it in its calculations. But the good news is that in their paper (PDF), they showed that for at least several years Linux should be able to keep up with chip enhancements in the multicore realm. To handle multiple cores, Linux keeps a counter of which cores are working on the data. As a core starts to work on a piece of data, Linux increments the number. When the core is done, Linux decrements the number. As the core count approached 48, the amount of actual work decreased and Linux spent more time managing counters. But the team found that 'Slightly rewriting the Linux code so that each core kept a local count, which was only occasionally synchronized with those of the other cores, greatly improved the system's overall performance.' The researchers caution that as the number of cores skyrockets, operating systems will have to be completely redesigned to handle managing these cores and SMP. After reviewing the paper, one researcher is confident Linux will remain viable for five to eight years without need for a major redesign."

I don't know, guess I picked a bad title or something?

Luckily we aren't anywhere near 48 cores and there is some time left to come up with a new Linux (Windows?).

Again, seriously? What does "(Windows?)" even mean? As you pass a certain number of cores, modern operating systems will need to be redesigned to handle extreme SMP. It's going to differ from OS to OS but we won't know about Windows until somebody takes the time to test it.

It appears that the problem, that affect the available memory in a chip when multiple cores are working on the same chunks of data, is getting worse and may be hitting a peak somewhere in the neighborhood of 48 cores, when entirely new operating systems will be needed, the report says.

Seriously? You picked that over my submission?

I submitted this earlier this morning I guess my submission was lacking. But if you're interested in the original MIT article and the actual paper (PDF):

eldavojohn writes "Multicore (think tens or hundreds of cores) will come at a price for current operating systems. A team at MIT found that as they approached 48 cores their operating system slowed down. After activating more and more cores in their simulation, a sort of memory leak occurred whereby data had to remain in memory as long as a core might need it in its calculations. But the good news is that in their paper (PDF), they showed that for at least several years Linux should be able to keep up with chip enhancements in the multicore realm. To handle multiple cores, Linux keeps a counter of which cores are working on the data. As a core starts to work on a piece of data, Linux increments the number. When the core is done, Linux decrements the number. As the core count approached 48, the amount of actual work decreased and Linux spent more time managing counters. But the team found that 'Slightly rewriting the Linux code so that each core kept a local count, which was only occasionally synchronized with those of the other cores, greatly improved the system's overall performance.' The researchers caution that as the number of cores skyrockets, operating systems will have to be completely redesigned to handle managing these cores and SMP. After reviewing the paper, one researcher is confident Linux will remain viable for five to eight years without need for a major redesign."

I don't know, guess I picked a bad title or something?

Luckily we aren't anywhere near 48 cores and there is some time left to come up with a new Linux (Windows?).

Again, seriously? What does "(Windows?)" even mean? As you pass a certain number of cores, modern operating systems will need to be redesigned to handle extreme SMP. It's going to differ from OS to OS but we won't know about Windows until somebody takes the time to test it.

It appears that the problem, that affect the available memory in a chip when multiple cores are working on the same chunks of data, is getting worse and may be hitting a peak somewhere in the neighborhood of 48 cores, when entirely new operating systems will be needed, the report says.

Seriously? You picked that over my submission?

I submitted this earlier this morning I guess my submission was lacking. But if you're interested in the original MIT article and the actual paper (PDF):

eldavojohn writes "Multicore (think tens or hundreds of cores) will come at a price for current operating systems. A team at MIT found that as they approached 48 cores their operating system slowed down. After activating more and more cores in their simulation, a sort of memory leak occurred whereby data had to remain in memory as long as a core might need it in its calculations. But the good news is that in their paper (PDF), they showed that for at least several years Linux should be able to keep up with chip enhancements in the multicore realm. To handle multiple cores, Linux keeps a counter of which cores are working on the data. As a core starts to work on a piece of data, Linux increments the number. When the core is done, Linux decrements the number. As the core count approached 48, the amount of actual work decreased and Linux spent more time managing counters. But the team found that 'Slightly rewriting the Linux code so that each core kept a local count, which was only occasionally synchronized with those of the other cores, greatly improved the system's overall performance.' The researchers caution that as the number of cores skyrockets, operating systems will have to be completely redesigned to handle managing these cores and SMP. After reviewing the paper, one researcher is confident Linux will remain viable for five to eight years without need for a major redesign."

I don't know, guess I picked a bad title or something?

Luckily we aren't anywhere near 48 cores and there is some time left to come up with a new Linux (Windows?).

Again, seriously? What does "(Windows?)" even mean? As you pass a certain number of cores, modern operating systems will need to be redesigned to handle extreme SMP. It's going to differ from OS to OS but we won't know about Windows until somebody takes the time to test it.

The ALU is from the "Hack" CPU described in this book: The Elements of Computing Systems, Building a Modern Computer from First Principles

He now does commonsense-reasoning stuff at IBM Research using formal logic, but back in his grad-school days, Erik Mueller wrote a thesis on building a computational model of daydreaming.

http://cookies.lcs.mit.edu/encyclopedia.html

From alt.sysadmin.recovery FAQ v1.799999999999999998... (1 April 1974):

4.4) Revolvers, cyanide and high voltages: The pros and cons of various luser education strategies.

There has been a great deal of debate on ASR about the best way of dealing with lusers, and at this time no consensus has been reached. What we can suggest, however, is to be sure it is painful, clean, and doesn't harm the computer. That unfortunately leaves a lot of options out; you can't just throw a grenade at them; it will hurt the machine.

ftp://rtfm.mit.edu/pub/faqs/sysadmin-recovery

Not so fast. I see that he does not dispute global warming, merely the magnitude that is caused by man, which he places at 1/3rd that identified in the worst-case IPCC models ("Taking Global Warming Seriously" - you'll find it in his papers). Further, he does not dispute the positive feedback mechanisms or any of the other mechanisms described in global warming, he merely states that other effects also exist and that radiative models alone are inadequate to model the atmosphere as a whole.

In short, his papers do NOT say what you (or Wikipedia) say they say. Unless you can get him to post here himself (verifiably, so no sock puppets), I will have to take what he has actually written over and above any other claim.

Somewhere around half a percent of the Earth itself and 1 percent of soil is titanium, so it isn't exactly rare. There's a large market for titanium dioxide in industrial quantities and it currently costs about $1.50 per pound.

I couldn't find any sources of GFP in industrial quantities (or any industrial uses of it), but looking at the production costs of other recombinant proteins is telling. In 1997, heparinase I production was estimated to cost around $250,000 per pound with capital costs in the tens of millions of dollars for an annual production of only 3 kg. On the other hand, bovine somatotropin is currently produced, and costs about $6.60 per 500 mg dose, which works out to about $6000 per pound.

I'm no expert, but the idea that GFP (a recombinant protein) is cheap and TiO2 (processed dirt) is expensive seems a bit strange to me. Can someone explain?

Without copyright, old son, you'd likely not have a textbook. People still generally have this whole "pay me for my knowledge" defect, you see. I know you think information "wants to be free", but it's actually "freeloaders want information to be free", and the producers of information have a lot more value to society than the freeloaders do.

The Internet, by which we participate in this conversation, was created by people willing to think and plan and code not for personal monetary gain but for the betterment of society. Fame and glory. There is no reason this will not work for textbooks. The Open Slate Project advocates fully integrating tablet computers into secondary education, and open source content ranging from Ebooks to apps. The content piece is called Chalk Dust.

If this sort of thing appeals to you, consider joining our SourceForge mailing list.

MIT has been doing great work with Open Courseware.

An MIT student did his thesis on Voice vs. Data lantency on cell phones, you might be interested in his methodology and results:
Quality of Service Analysis for Audio over Cellular Voice Networks

is at http://web.mit.edu/newsoffice/2010/self-healing-solar.html

Colorado has a similar law, but I have yet to see the highway patrol ticket the morons from Wyoming that think it's OK to just cruise I-25 in the left lane, and who have no idea what high-beams flashing in their rear view mirror means. Ticketing these people would be a great way to raise revenue in this state.

At least most Colorado drivers are finally getting the hang of it after 6+ years.

...people need to understand the concepts of bits, bytes, words, longwords, binary/octal/hex numbers, thinking sequentially and logically,
what an operating system actually does, what an IO system is and does, how a computer actually does math, etc., etc., etc.

I think there's a careful distinction to be made in such a discussion, between fundamental concepts and implementation details. For example, I think knowing binary is an important skill for a casual (non-professional) programmer, whilst hex isn't so much since it's just used as a more compact form to displaying binary in for humans*. Similarly bits are important, bytes not so much since the notion of 8 bits in a byte is just an arbitrary standard. Knowing things like the distinction between float and double, short and long, etc. isn't IMHO suitable for such an age group. This is because learning such things enough to be second-nature is often difficult since they're arbitrary and thus will probably rely on rote-learning (ie. they're boring) but more dangerously, since they would always be in the student's mind when they're taking the course it would distract them from the actual concepts being taught.

At University every computing course I took was based in Java (though I learned C and C++ from my Physics classes too), and it showed: Masters level students would struggle to grasp important steps in straightforward algorithms, yet their incorrect coursework implementations would show a clear appreciation for such irrelevant details as serialise versions (in code destined to never have a subsequent version), extraneous exception handler definitions (most of which would actually leave the state in a broken way) and elaborate layering of objects for streams, readers, writers, buffers, builders, factories, etc. to get data into and out of their wrong, one-method implementation of the algorithm, in a text-book-exact way (sometimes not even changing the variable names).

My recommendations for things to include:

Message-passing Object Oriented programming; no throwbacks from structured programming like if/then/else, for/foreach/while/dowhile, etc. Stick to one concept that has no special-cases, and languages with as few reserved keywords as possible (since lots of keywords implies that some things need to be achieved via some non-pervasive concept, usually hard-coded into the compiler/interpreter). Here I would recommend Smalltalk ( http://www.smalltalk.org/main/ ), since it's been taught to children for years all over the world, so there's lots of experience to build on. It's based on objects with classes which send and receive messages, and essentially defined OO programming as it's known today. There's very little syntax to learn, if/then/else are messages sent to objects (eg. myCondition ifTrue: myTrueCode ifFalse: myFalseCode), loops are also messages (myListOfObjects do: myLoopBody) and so on. The distinction between classes and instances might be unnecessarily confusing, so you *may* want to look into languages like Self ( http://selflanguage.org/ ) which use prototypes in a similarly pervasive way. Smalltalk also has Etoys ( http://www.squeakland.org/ ) to play with, which is a prototype-based 'ultimate LOGO' and really makes the message-passing concept of OO explicit via the menu structure. DrGeo ( http://community.ofset.org/index.php/DrGeo ) runs in Smalltalk, which provides an incentive to kids to learn the system (since it would give those students who learn it an advantage in Maths classes, since they'd have access to a really sophisticated geometric calculator). Scratch ( http://scratch.mit.edu/ ) is built in Smalltalk too, but bears little relation to the underlying system (whereas with Etoys it is a simple step to pure Morphic, then to classes/instances, then t

"You don't have to be really patient... plates move at 2-10 cm/year so you'd start getting GPS data within 2-5 years"

It's faster than that. A few days/weeks of monitoring with the right equipment is sufficient at a given station (these are *not* handheld GPS units!), allowing the motion of entire regions to be studied from many points in a year or two of fieldwork moving the stations around. And many regions now have permanently mounted GPS networks to monitor continuously. A couple of years of continuous data is sufficient to get great detail and precision. That allows geologists to study not only the motion of entire plates, but the details of deformation of mountain ranges at the plate boundaries and the effects of individual earthquakes -- essentially real-time monitoring of the motion of the Earth's surface at millimetre precision. Here are a few papers [PDF].

If you want to know how fast you are moving at your own location with respect to a given reference frame try this, which is derived from current whole-Earth models of plate motion. Please note that it probably won't be accurate in areas with complex deformation near plate boundaries (it models the plates as rigid), but if you're within the plate somewhere it will be a reasonable approximation.

One of the coolest analogies of scale ever: the plates move at about the same rate that your fingernails grow.

Recently somebody pointed me to Scratch. Its a simple, fun, programming tool. It is aimed at kids eight years and above but I am sure it would be appropriate for some 14 year olds too. My dad is teaching computer skills at U3A. He is interested in using this tool in his classes.

By heating up the material, the oil can be removed and burnt locally and the nanofabric can be reused.

Notice the URL - it's MIT saying this, not someone mis-quoting them.

Also, good luck with that during hurricane season.

Additionally, bad math alert. To clean up 5 million barrels in 30 days with 5,000 units, each unit would have to pick up 33 barrels a day. 16'x7'= 112 square feet. A barrel is 42 gallons, and there are 7.5 gallons in a cubic foot. So, 33 barrels is 1,385 gallons, or 184.5 cubic feet. Your skimmer will be towing a chunk of oil-soaked nanofibres half a yard thick - you're not going to be making much headway dragging that with only 100 watts (1/8 horsepower).

It might start out okay, but as you collect oil, it will get worse, so take that 1 month and make it a year.

The way I see it, there are four main problems with holograms. First, they are static

There are several efforts underway to create video holograms using acousto-optical crystals or spatial light modulators, such as the Holovideo project at MIT.

First, they are static.

That's a limitation of the way most holograms have been produced, not a limitation of holography in general.

Secondly, they are not color. This is due to the nature of laser light. It is monochromatic.

So use three of them, like the people who have built colour vector display projectors using red, green, and blue lasers.

There is no holographic output device, like a monitor, on which to show holograms.

That's only because no one has come up with a mass-market device of that type. It's certainly possible to do. I feel like a broken record posting a link to the MIT Media Lab's historical page on the topic, but there it is again.

By Ben Vigoda, Co-Founder and CEO: http://phm.cba.mit.edu/theses/03.07.vigoda.pdf

Huh, I thought he was dead.

Maybe not.

By Ben Vigoda, Co-Founder and CEO: http://phm.cba.mit.edu/theses/03.07.vigoda.pdf

Huh, I thought he was dead.

By Ben Vigoda, Co-Founder and CEO: http://phm.cba.mit.edu/theses/03.07.vigoda.pdf

people have been trying to make this stuff cheaply and efficiently for a while now -> http://parts.mit.edu/igem07/index.php/Alberta was one such attempt that I heard about