In case anyone is interested, here's the link to the NPR review of 'Into Darkness'. That reviewer seems to echo this review's sentiments pretty closely; it's different from the old Star Trek, and doesn't have much thought to it, but is entertaining nonetheless.
the admin who wont call a terrorist attack a terrorist attack simply because it goes against his political agenda?
You are likely referring to the Boston bombings; as I understood it, Obama didn't use the term "terrorist" specifically ON the day of the bombings, and has ever after. I'd say this is simply him doing his due-diligence in not jumping to conclusions, as at the time no one knew if the explosions weren't simply a gas line exploding. If anything I'd want more of politicians and news stations taking a deliberate and thorough approach to things, rather than going all reddit on us and pointing fingers and making sensationalist claims. Each to their own eh?
Students have it good when it comes to matlab -- you can get a student version of matlab + simulink (with 10 or so toolboxes) for $99. The people who are really hurt by matlab's pricing schemes are the hobbyists who don't qualify for a student copy. There's this huge price dichotomy; when you're a student it's $99, after you graduate it's $5000+, and that's without any toolboxes.
However, for academic use it makes perfect sense for scientists to use matlab over the alternatives. At least in the UC (university of california) system, a department will have some (large) number of licences for it's faculty to use, and so professors and their student (who commonly won't have much coding experience) will have what's essentially free access matlab and it's associated toolboxes. From their perspective, they want to run their experiments and write their papers, not learn how to code and be a pseudo-sysadmin. They want to use the simplest environment that stays out of their way, and not have to deal with installing various libraries. Say what you will about matlab, its support and documentation is very good.
This is a major improvement for GPGPU, not game playing. Memory throughput is often the bottleneck in applications, as computational throughput improvements has greatly outstripped memory throughput improvements. To give you an idea about the importance of memory bandwidth, if you have a GPU with a peak arithmetic throughput of 1170 GFLOPS (this is how much a Tesla K20 gets for double precision floating point) performing FMA (fused multiply add, so 2 floating point operations for 3 operands), then to sustain that level of throughput, you would need roughly 13 TB/s*** memory throughput (this is assuming 8 byte operands and that each of the 3 operands of the FMA are unique). Of course you can't reach those levels with global memory, but any sort of improvement helps.
This Volta sounds pretty exciting, DRAM bandwidth is commonly a limiting factor in GPGPU applications, so if it can get 1TB/s, it'll be more than 3x faster for memory bound kernels than the current high-end scientific computing cards (e.g. the Tesla K20). With that said, I'm a bit apprehensive about how much it'll cost; Tesla K20's currently cost over $2k per card...
seriously? The getting started guide is really straightforward, and even has examples embedded into it to let you know if you've set everything up correctly. I can understand someone griping about the build process, but that's hardly specific to boost. Your inability to get started with it might reflect more on you than boost....
By the time remote controlled robots would be usable enough to carry around and install office equipment it won't be long before we have robots that can do it without any remote control.
There's actually a huge gap here. Having robots perform tasks autonomously in anything other than a very narrow, constrained environment would require semantic understanding. We've had robots that can go through the motions while being controlled by humans for decades (telerobotics), but developing machines with deep understanding on a semantic level has been something of a holy grail in AI, and is still as far-off as ever. While I agree with you that we'll have to undergo a dramatic cultural shift if we ever create a legitimate AI, that is still many, many years off.
CUDA has been out for a couple of years now, and has gotten to be quite stable and has grown a good ecosystem of tools. As for python? Knock yourself out
I doubt he has a chance in hell of winning his suit.
I wouldn't be too sure; from the article "Lecerf said that it wasn't the first time his speed dial had jammed but that Renault had looked at the car and assured him that it was fine." (emphasis mine)
whether or not this makes them liable is another matter, but they apparently did have some involvement with his vehicle
From the article:
"A Renault technician had been on the phone with police throughout the chase trying to help but couldn't come up with a solution."
Of course I can't say with 100% certainty, but I'm guessing the Renault technician would have thought of all of your proposed solutions and more. It's important to note that his car had been "adapted for disabled drivers", which likely played some role in its malfunction, so conventional wisdom about cars may not be as applicable, depending on the modifications made. Also, he likely has various disabilities, given that his car is for disabled drivers, and that he "had two epileptic seizures" during the drive, so it's likely not necessarily a matter of him failing "to understand his vehicle's operation" as you say, so much as him being physically and/or mentally unable to take action.
One last interesting note from the article: "it wasn't the first time his speed dial had jammed but that Renault had looked at the car and assured him that it was fine." That's probably where the legal complaint comes into play
Then I'm glad to report that things are different nowadays (at least at my university). I went to a large public university with a sizable foreign student contingent, and saw none of this. There were ethnic cliques, but if you tried talking to them they would be very accommodating. In fact, I found the foreign students to be among the most friendly, at least when the language barrier wasn't an issue.
... when they build in a flood plane but don't want to bother to buy flood insurance. After all, they needed to money that they could have spent on flood insurance to pay for their fancy smart phone to run the app...
That's a very cynical view; insurance policies are much trickier than that. For example, the "anti-concurrent causation clause", which states that "if you have two events happen at the same time, one that is covered and the other that is not covered by the policy, the insurance company doesn't have to pay for either. So if your house has damage from wind, which is covered by your policy, and it also has damage from a flood, which is not covered, the insurance company doesn't have to pay for either the wind or flood damage."
Taken from here
So you can have full flood insurance, but if your property received damage due to something else, you'll still be screwed. If you're in say, a hurricane, odds are you'll be getting more than just flooding. Try having a little compassion for your fellow man, although with sentiment like
... at least we will not be pestered by the truly needy
From wikipedia:
FLOPS stands for "FLoating-point Operations Per Second, also flop/s". The/sec is part of the acronym. Hence "FLOPS/sec" would be floating point operations per second per second, which probably isn't what you meant. Likewise, saying "FLOP" denotes the execution of one floating pointer operation. There's no time metric, and is equally meaningless.
According to NVIDIA, the peak single precision floating point throughput is 3.95 TFLOPS for a Tesla K20X. If, for simplicity, we round up to 4TFLOPS and assume that it operates at peak efficiency (which is unlikely), you'd need 250 of these. Of course that also assumes linear scaling with the number of processing units (which is also unlikely). This would be for 1 PFLOPS mind you; the DOE wants 10 ~ 30.
So how many GPU for a Peta Flop ?
Not to be overly pedantic, but PFLOPS --> 10^15 FLOP/sec, so saying "Peta Flop" doesn't make any sense - that would just be 10^15 floating point operations, which without a time-scale (seconds) is meaningless.
You might not want to invoke Mathworks as an example of competence. They just include binaries of whatever libraries they need, and maintain a different search path for libraries. This works, until you have to write a MEX file and want to use something like OpenCV. If your version is different from the one bundled with Matlab, you'll get runtime errors. You can either delete the library in the Matlab source directory (and symlink whatever version of the library you want to use), or be forced to use whatever version of the library that Matlab bundled with their software.
Honestly curious why this is set up this way, it seems so inefficient and insecure.
Hah, welcome to the internet. But seriously though, a lot of the protocols in use weren't designed with the current form of the internet in mind, so looking at them now it's almost amazing that the internet is as functional as it is. The web is built on trust, which made sense back in its infancy. Not quite as much anymore however.
for example, just a few months ago google was effectively inaccessible to a portion of the world, entirely by accident
On NPR a while back there was an interview with a NASA scientist about the doomsday predictions. He runs the Q/A column "Ask an Astrobiologist" responsible for answering questions posed to NASA, which as of late have mostly centered around Dec 21st. He was incredulous that anyone would believe these stories, but as the same time took it very seriously; he mentioned that many of the people writing to him were of the younger generations (i.e. schoolkids) genuinely concerned about whether the world was going to end. This was the demographic that concerned him. One such excerpt -- "Though some of the questions may seem frivolous and outlandish, Morrison receives queries from people who are legitimately concerned and contemplating suicide. "Another extreme one... I got was quite touching. It was: My only friend is my little dog. When should I put her to sleep so she won't suffer in the cataclysm?"
It's easy to dismiss the doomsday people as loons (and most are), but some of them are just kids so we should focus more on taking them seriously and helping educate them to understand that it's a myth rather than dismissing the entire thing offhand. This can be a very good opportunity to show the community that science > superstition.
Interview is at http://www.npr.org/2012/11/26/165928588/as-dec-21-draws-nigh-the-facts-about-doomsday
Agreed. Back at launch they were mentioning how surprised they were at the effect the auction house had. Internally they had something like a few dozen people playing it, so the auction house had a very limited pool of items, and as a result was fairly insignificant. Fast forward to a week after launch and they were completely blindsided at the massive effect the auction house had on itemization. On the one hand I feel like they should have seen it coming, but on the other hand, complex, dynamic systems aren't easy to predict.
The part where you then go on to state that that parallel programming is trivial (or at least, something beneath *serious* programmers), because you managed to code up an embarrassingly parallel application.
You're describing an embarrassingly parallel application, the kind where you can actually get linear speedups. Unfortunately many problems don't fit into this category, and once you start getting into inter-thread data dependencies, jagged load balancing, and debugging race conditions it's an entirely different ballgame, not to mention massively parallel stream processors (i.e. GPGPU). Easy threading is easy, but deriding it as something less than "serious" programming belies a degree of ignorance.
Slashdot Mirror
In case anyone is interested, here's the link to the NPR review of 'Into Darkness'. That reviewer seems to echo this review's sentiments pretty closely; it's different from the old Star Trek, and doesn't have much thought to it, but is entertaining nonetheless.
the admin who wont call a terrorist attack a terrorist attack simply because it goes against his political agenda?
You are likely referring to the Boston bombings; as I understood it, Obama didn't use the term "terrorist" specifically ON the day of the bombings, and has ever after. I'd say this is simply him doing his due-diligence in not jumping to conclusions, as at the time no one knew if the explosions weren't simply a gas line exploding. If anything I'd want more of politicians and news stations taking a deliberate and thorough approach to things, rather than going all reddit on us and pointing fingers and making sensationalist claims. Each to their own eh?
Students have it good when it comes to matlab -- you can get a student version of matlab + simulink (with 10 or so toolboxes) for $99. The people who are really hurt by matlab's pricing schemes are the hobbyists who don't qualify for a student copy. There's this huge price dichotomy; when you're a student it's $99, after you graduate it's $5000+, and that's without any toolboxes.
However, for academic use it makes perfect sense for scientists to use matlab over the alternatives. At least in the UC (university of california) system, a department will have some (large) number of licences for it's faculty to use, and so professors and their student (who commonly won't have much coding experience) will have what's essentially free access matlab and it's associated toolboxes. From their perspective, they want to run their experiments and write their papers, not learn how to code and be a pseudo-sysadmin. They want to use the simplest environment that stays out of their way, and not have to deal with installing various libraries. Say what you will about matlab, its support and documentation is very good.
I'm aware your comment was mean facetiously, but lambdas were also added to C++ in C++11
This is a major improvement for GPGPU, not game playing. Memory throughput is often the bottleneck in applications, as computational throughput improvements has greatly outstripped memory throughput improvements. To give you an idea about the importance of memory bandwidth, if you have a GPU with a peak arithmetic throughput of 1170 GFLOPS (this is how much a Tesla K20 gets for double precision floating point) performing FMA (fused multiply add, so 2 floating point operations for 3 operands), then to sustain that level of throughput, you would need roughly 13 TB/s*** memory throughput (this is assuming 8 byte operands and that each of the 3 operands of the FMA are unique). Of course you can't reach those levels with global memory, but any sort of improvement helps.
*** required memory throughput per second = 1170 * 10^9 ops * (24 bytes / 2 ops) = 14040 * 10^9 bytes ~= 13 TB
This Volta sounds pretty exciting, DRAM bandwidth is commonly a limiting factor in GPGPU applications, so if it can get 1TB/s, it'll be more than 3x faster for memory bound kernels than the current high-end scientific computing cards (e.g. the Tesla K20). With that said, I'm a bit apprehensive about how much it'll cost; Tesla K20's currently cost over $2k per card...
It's also possible to build boost using cmake
seriously? The getting started guide is really straightforward, and even has examples embedded into it to let you know if you've set everything up correctly. I can understand someone griping about the build process, but that's hardly specific to boost. Your inability to get started with it might reflect more on you than boost....
By the time remote controlled robots would be usable enough to carry around and install office equipment it won't be long before we have robots that can do it without any remote control.
There's actually a huge gap here. Having robots perform tasks autonomously in anything other than a very narrow, constrained environment would require semantic understanding. We've had robots that can go through the motions while being controlled by humans for decades (telerobotics), but developing machines with deep understanding on a semantic level has been something of a holy grail in AI, and is still as far-off as ever. While I agree with you that we'll have to undergo a dramatic cultural shift if we ever create a legitimate AI, that is still many, many years off.
CUDA has been out for a couple of years now, and has gotten to be quite stable and has grown a good ecosystem of tools. As for python? Knock yourself out
I doubt he has a chance in hell of winning his suit.
I wouldn't be too sure; from the article "Lecerf said that it wasn't the first time his speed dial had jammed but that Renault had looked at the car and assured him that it was fine." (emphasis mine) whether or not this makes them liable is another matter, but they apparently did have some involvement with his vehicle
From the article: "A Renault technician had been on the phone with police throughout the chase trying to help but couldn't come up with a solution." Of course I can't say with 100% certainty, but I'm guessing the Renault technician would have thought of all of your proposed solutions and more. It's important to note that his car had been "adapted for disabled drivers", which likely played some role in its malfunction, so conventional wisdom about cars may not be as applicable, depending on the modifications made. Also, he likely has various disabilities, given that his car is for disabled drivers, and that he "had two epileptic seizures" during the drive, so it's likely not necessarily a matter of him failing "to understand his vehicle's operation" as you say, so much as him being physically and/or mentally unable to take action. One last interesting note from the article: "it wasn't the first time his speed dial had jammed but that Renault had looked at the car and assured him that it was fine." That's probably where the legal complaint comes into play
Then I'm glad to report that things are different nowadays (at least at my university). I went to a large public university with a sizable foreign student contingent, and saw none of this. There were ethnic cliques, but if you tried talking to them they would be very accommodating. In fact, I found the foreign students to be among the most friendly, at least when the language barrier wasn't an issue.
... when they build in a flood plane but don't want to bother to buy flood insurance. After all, they needed to money that they could have spent on flood insurance to pay for their fancy smart phone to run the app...
That's a very cynical view; insurance policies are much trickier than that. For example, the "anti-concurrent causation clause", which states that "if you have two events happen at the same time, one that is covered and the other that is not covered by the policy, the insurance company doesn't have to pay for either. So if your house has damage from wind, which is covered by your policy, and it also has damage from a flood, which is not covered, the insurance company doesn't have to pay for either the wind or flood damage." Taken from here
So you can have full flood insurance, but if your property received damage due to something else, you'll still be screwed. If you're in say, a hurricane, odds are you'll be getting more than just flooding. Try having a little compassion for your fellow man, although with sentiment like
I suppose that might be asking for too much
NPR has an article about this as well, apparently it's affecting more than 30 cities in China
From wikipedia: FLOPS stands for "FLoating-point Operations Per Second, also flop/s". The /sec is part of the acronym. Hence "FLOPS/sec" would be floating point operations per second per second, which probably isn't what you meant. Likewise, saying "FLOP" denotes the execution of one floating pointer operation. There's no time metric, and is equally meaningless.
So how many GPU for a Peta Flop ?
Not to be overly pedantic, but PFLOPS --> 10^15 FLOP/sec, so saying "Peta Flop" doesn't make any sense - that would just be 10^15 floating point operations, which without a time-scale (seconds) is meaningless.
You might not want to invoke Mathworks as an example of competence. They just include binaries of whatever libraries they need, and maintain a different search path for libraries. This works, until you have to write a MEX file and want to use something like OpenCV. If your version is different from the one bundled with Matlab, you'll get runtime errors. You can either delete the library in the Matlab source directory (and symlink whatever version of the library you want to use), or be forced to use whatever version of the library that Matlab bundled with their software.
Honestly curious why this is set up this way, it seems so inefficient and insecure.
Hah, welcome to the internet. But seriously though, a lot of the protocols in use weren't designed with the current form of the internet in mind, so looking at them now it's almost amazing that the internet is as functional as it is. The web is built on trust, which made sense back in its infancy. Not quite as much anymore however. for example, just a few months ago google was effectively inaccessible to a portion of the world, entirely by accident
Didn't do much in 2012, but I make it something of an annual tradition to read through the archive. http://achewood.com/
On NPR a while back there was an interview with a NASA scientist about the doomsday predictions. He runs the Q/A column "Ask an Astrobiologist" responsible for answering questions posed to NASA, which as of late have mostly centered around Dec 21st. He was incredulous that anyone would believe these stories, but as the same time took it very seriously; he mentioned that many of the people writing to him were of the younger generations (i.e. schoolkids) genuinely concerned about whether the world was going to end. This was the demographic that concerned him. One such excerpt -- "Though some of the questions may seem frivolous and outlandish, Morrison receives queries from people who are legitimately concerned and contemplating suicide. "Another extreme one ... I got was quite touching. It was: My only friend is my little dog. When should I put her to sleep so she won't suffer in the cataclysm?"
It's easy to dismiss the doomsday people as loons (and most are), but some of them are just kids so we should focus more on taking them seriously and helping educate them to understand that it's a myth rather than dismissing the entire thing offhand. This can be a very good opportunity to show the community that science > superstition.
Interview is at http://www.npr.org/2012/11/26/165928588/as-dec-21-draws-nigh-the-facts-about-doomsday
Agreed. Back at launch they were mentioning how surprised they were at the effect the auction house had. Internally they had something like a few dozen people playing it, so the auction house had a very limited pool of items, and as a result was fairly insignificant. Fast forward to a week after launch and they were completely blindsided at the massive effect the auction house had on itemization. On the one hand I feel like they should have seen it coming, but on the other hand, complex, dynamic systems aren't easy to predict.
The part where you then go on to state that that parallel programming is trivial (or at least, something beneath *serious* programmers), because you managed to code up an embarrassingly parallel application.
You're describing an embarrassingly parallel application, the kind where you can actually get linear speedups. Unfortunately many problems don't fit into this category, and once you start getting into inter-thread data dependencies, jagged load balancing, and debugging race conditions it's an entirely different ballgame, not to mention massively parallel stream processors (i.e. GPGPU). Easy threading is easy, but deriding it as something less than "serious" programming belies a degree of ignorance.
Just as an aside, they have a GoG Downloader that you can install (for Windows and OS X), then use to download your game(s)