I haven't used the Shield Tablet so I'm not sure (I only have a Jetson TK1, that uses CUDA on Linux). It isn't possible to directly compare CUDA on Tegra K1 versus CUDA on any other mobile device, whereas there are obviously many competing devices supporting OpenCL, so the linked benchmark results are probably the closest thing there is to an apples-to-apples comparison of GPGPU compute performance on Tegra K1 versus all the competition. If a marketing slide on the NVIDIA website said Tegra K1 gets 5x boost in OpenCL compared to the next contendor then I'd understand the skepticism, but the linked results are for the industry-wide standard benchmark used for GPGPU comparisons.
Benchmarks should always be taken with some skepticism, but as an Engineer (not a Marketer) I know that Tegra K1's GPU really does optimize many whole algorithms by around 5x compared to competing mobile devices, so this isn't BS.
"Conformant" does not mean it is actually running faster due to GPU hardware acceleration. On mobile, "conformant" OpenCL usually just means it runs on the CPU or runs on the GPU at a very slow speed, as opposed to Tega K1 that is giving clear GPU speedups.
OpenCL isn't supported by Tegra K1 on Linux4Tegra (ie: Jetson TK1 embedded board) but it is clearly supported on Android, hence why this benchmark was able to execute and post results.
Recent iPhones certainly have very powerful PowerVR GPUs, and Apple was the original creator of OpenCL. But can you show any benchmarks or proof that iPhone 5S allows GPU-accelerated OpenCL apps?
The Anandtech article clearly mentions that OpenCL in Nexus 10 was an unsupported feature that hackers figured out how to use but it wasn't actually intended for developers to use officially, hence why it disappeared soon after an update. So I guess you are right that Tegra K1 is perhaps not the first mobile chip to do GPU accelerated OpenCL, but it is the first one to officially offer it and provide full support to ensure it runs well without bugs and without high power draw, etc.
And yes it's true that the Jetson TK1 embedded Linux board doesn't support OpenCL at all, but that is due to Linux OS related issues. There are only OpenCL drivers for Tegra K1 on Android, not Linux (unfortunately!). That doesn't change the fact that the Tegra K1 chip supports hardware-accelerated OpenCL on Android.
So I don't see either of those 2 points as being false information or dishonest.
Actually I worked in a university in United Arab Emirates near the border of Oman, and there were about 10x more females studying IT & post-grad in computer or engineering related fields than male students!
We assume it is because the local Emirati's in UAE are so rich that they don't need to study or work, but since females are expected to marry and become a house-wife & mother, the most obvious way for them to not have that way of life is if they become a professional. so we believe this explained why UAE has so many female IT & Engineering students of high quality, but almost no male IT or Engineering students and most of the males there are of very low quality since they are just doing it for fun.
The only real difference between the US and other first world countries is that in the US what you get for your tax dollar is a military larger than that of the next 17 countries put together; what we get in the rest of the world is free healthcare and free or heavily subsidized tertiary education.
Nice summary of USA:-) I also prefer the free healthcare & education rather than a powerful military, but I guess that's because I'm not American!
But phones don't have to be so small that they can't fit proper camera hardware. I already carry a 1/2" digicam in my pocket, so I'd be happy to replace it with a 1/2" phone that includes a digicam-level lens & camera sensor.
Sure there is always going to be a market for people that want smaller & smaller phones, but also there is a market for people that want phones as large as a digicam, if it means much better camera system, much bigger battery, potentially a larger keypad, and potentially shutter & zoom buttons on the side of the phone so you can actually use it like a digicam.
From the article & video, all I can see is a higher-resolution version of an Omnidirectional camera, which is very common in mobile robots. Such as this list of about 50 different types! "http://www.cis.upenn.edu/~kostas/omni.html"
They keep referring to the notion of depth being used, but unless there is some big technology that they completely forgot to mention in the article & video, it just does the equivalent of pointing a camera into a bowl shaped mirror, allowing you to see in all 360 degrees at once. eg: "http://en.wikipedia.org/wiki/Omnidirectional_camera"
That is quite different to say it is truly 3D, since it is still a 2D image without depth, just that its wrapped around a circle shape instead of rectangle shape.
You guys think it is ridiculous that these websites often only pay about $0.60/hr, but thats because you guys are all living in USA or the West, where $5/hr is low. But most people live in 3rd world countries, where $0.60/hr is considered a reasonable income. That's why nearly all outsourcing is done in India or other poor countries, because $0.60/hr through Internet (with a possibility of earning several dollars per hour) seems as good as working a typical fulltime job that may also pay just $0.60/hr.
So the crazy thing is not that these websites pay less than $1/hr, its that the same job in the West will pay 10 or 20 times higher than in the rest of the world!
PS: I currently live in a farming town in Philippines where the average fulltime income is about $0.50/hr, so the prospect of earning that much through Internet seems pretty good.
It was already free for non-commercial use, so I guess Microsoft finally realised that not many companies are making money from robots (besides military robots & industrial robots), so it wasn't really earning them much money anyway.
Personally, I hated learning Microsoft Robotics Studio, it is great for doing a few basic things, but doing anything different than that requires a lot of things that must be learnt that have very little to do with your robotics project. The whole purpose of a robotics environment like this is so that robotics builders can concentrate on actual robot stuff, whereas MSRDS makes you waste so much time learning their proprietary formats & techniques that you could easily loose the fun in learning robots!
Thats my opinion anyway, as a someone with a Masters degree in Robotics Engineering that still has trouble using MS Robotics Design Studio!
Yeah for example in Dubai & UAE where about 80% of the people in the country are temporary hired workers from overseas, its common that even for the "blue-collar" jobs, your employer keeps your passport during your 1 or 2 yr contract to make sure you can't leave the job or country!
The PR2 robot has two eight-core i7 Xeon system servers on-board, each with 24 GB of RAM..... over a 16-port gigabit Ethernet hub
Thats what happens when Google funds a robot: it becomes a mobile server rack!
But I must admit that its pretty impressive computing power to have on a mobile robot! Most research robots have relatively outdated embedded computing hardware compared to consumer laptops, so this will allow much more detailed algorithms to run.
Its true that military robots & technology are trying to be used to make warfare more "clean", so that the desired targets can be bombed with the least damage to civilians. That is good, atleast in theory, BUT at the end of the day, it is still all going to be controlled by the military leaders that don't necessarily know who are the civilians and who are the targets.
I have actually worked on a military robot for intended deployment in Iraq, and our military officer explained that when you are in a place like Iraq, you don't know who the enemy is and who the civilians are, because even if you see a 5yr old girl with her innocent looking grandmother and you ignore or help them, they are just as likely to try to secretly attack you as someone dressed in military uniform. So the US military in Iraq has to basically assume everyone that isn't a US soldier might be the enemy and therefore they can convince themselves that the ethical thing to do is kill anyone they see that they aren't completely sure is on their side.
So it doesn't matter whether the soldiers have basic weapons or latest military robots, they are still in the mind-set that any civilian can be considered part of the enemy's military.
The main advantage of military robots to the USA is that the countries that USA invades will be much poorer & less advanced countries than USA, so the enemy wont be able to make use of cutting-edge military technology compared to America.
If you don't believe me, put it this way: if Iraq had just as many military soldiers & robots fighting in USA as USA has in Iraq, do you still think people would see this the same way? The "Iraq War" and the "Afghanistan War" aren't wars, they are one-sided invasions, so its very different than if those countries were actually bombing America on a daily basis.
Actually nearly every time someone uses neural networks for image processing, they use the same approach, of making a long vector from the image pixels and feeding it into an ANN. And you are right that it has a lot of associated problems, but the ones you gave aren't really a problem with their software.
Like you said, you might want to use images of different sizes, or images that are in different pixel locations, and like you said, you need to pre-process the images so that they are of the exact same size and generally the desired features are in the exact same position in all images. Any image recognition system needs to do this, because it definitely is the most important part of image recognition, but after you have done that, then I don't see why you can't use the ANN library mentioned in this article to do the actual image recognition part.
In other words, any software that can handle images of different sizes or features in different locations is likely to include the same sort of preprocessing steps as you would need to do when using this ANN software.
OpenCV does provide very basic image recognition abilities but I definitely welcome any new open-sourced libraries that tackle image recognition too.
In 2006 & 2007 I was working at the American company that was working on this. I was working on a different project so I don't know the exact details, but I know they were using the BrainPort to let blind people see things (sending electrical spikes onto the tongue with a resolution of 64 points on the tongue). But it was actually funded by the military to allow hi-tech soldiers to get extra information through their tongue.
The demo they were working on at the time was to allow the soldier to get an idea of where their fellow soldiers where, through the BrainPort. After some practise, they can tell through their tongue that they have a soldier behind them on their left and 2 more coming on their right, for example.
Here is a little description of the project:
http://www.ihmc.us/research/projects/SensorySubstitution/
Yeah actually I read quickly through that already:-) But I was still confused because I know that in Australia we had a vote across the whole country, trying to decide whether Australia should become a Republic (having a President) or stay as a Monarch. It turns out that there's 2 meanings of the word "Republic", and so by some definitions, Australia is a Republic, and by other definitions, Australia is NOT a Republic! "http://www.ozpolitics.info/guide/topics/republic/"
I'm no expert, but doesn't "Republic" just mean there is a President in the federal government? For example, Australia has a Prime Minister and no President, so its not called a Republic, whereas somewhere like Iran or America has a President, hence why they're called a Republic?
Slashdot Mirror
I haven't used the Shield Tablet so I'm not sure (I only have a Jetson TK1, that uses CUDA on Linux). It isn't possible to directly compare CUDA on Tegra K1 versus CUDA on any other mobile device, whereas there are obviously many competing devices supporting OpenCL, so the linked benchmark results are probably the closest thing there is to an apples-to-apples comparison of GPGPU compute performance on Tegra K1 versus all the competition. If a marketing slide on the NVIDIA website said Tegra K1 gets 5x boost in OpenCL compared to the next contendor then I'd understand the skepticism, but the linked results are for the industry-wide standard benchmark used for GPGPU comparisons.
Benchmarks should always be taken with some skepticism, but as an Engineer (not a Marketer) I know that Tegra K1's GPU really does optimize many whole algorithms by around 5x compared to competing mobile devices, so this isn't BS.
"Conformant" does not mean it is actually running faster due to GPU hardware acceleration. On mobile, "conformant" OpenCL usually just means it runs on the CPU or runs on the GPU at a very slow speed, as opposed to Tega K1 that is giving clear GPU speedups.
OpenCL isn't supported by Tegra K1 on Linux4Tegra (ie: Jetson TK1 embedded board) but it is clearly supported on Android, hence why this benchmark was able to execute and post results.
Recent iPhones certainly have very powerful PowerVR GPUs, and Apple was the original creator of OpenCL. But can you show any benchmarks or proof that iPhone 5S allows GPU-accelerated OpenCL apps?
The Anandtech article clearly mentions that OpenCL in Nexus 10 was an unsupported feature that hackers figured out how to use but it wasn't actually intended for developers to use officially, hence why it disappeared soon after an update. So I guess you are right that Tegra K1 is perhaps not the first mobile chip to do GPU accelerated OpenCL, but it is the first one to officially offer it and provide full support to ensure it runs well without bugs and without high power draw, etc.
And yes it's true that the Jetson TK1 embedded Linux board doesn't support OpenCL at all, but that is due to Linux OS related issues. There are only OpenCL drivers for Tegra K1 on Android, not Linux (unfortunately!). That doesn't change the fact that the Tegra K1 chip supports hardware-accelerated OpenCL on Android.
So I don't see either of those 2 points as being false information or dishonest.
Actually I worked in a university in United Arab Emirates near the border of Oman, and there were about 10x more females studying IT & post-grad in computer or engineering related fields than male students!
We assume it is because the local Emirati's in UAE are so rich that they don't need to study or work, but since females are expected to marry and become a house-wife & mother, the most obvious way for them to not have that way of life is if they become a professional. so we believe this explained why UAE has so many female IT & Engineering students of high quality, but almost no male IT or Engineering students and most of the males there are of very low quality since they are just doing it for fun.
The only real difference between the US and other first world countries is that in the US what you get for your tax dollar is a military larger than that of the next 17 countries put together; what we get in the rest of the world is free healthcare and free or heavily subsidized tertiary education.
Nice summary of USA :-) I also prefer the free healthcare & education rather than a powerful military, but I guess that's because I'm not American!
But phones don't have to be so small that they can't fit proper camera hardware. I already carry a 1/2" digicam in my pocket, so I'd be happy to replace it with a 1/2" phone that includes a digicam-level lens & camera sensor.
Sure there is always going to be a market for people that want smaller & smaller phones, but also there is a market for people that want phones as large as a digicam, if it means much better camera system, much bigger battery, potentially a larger keypad, and potentially shutter & zoom buttons on the side of the phone so you can actually use it like a digicam.
From the article & video, all I can see is a higher-resolution version of an Omnidirectional camera, which is very common in mobile robots. Such as this list of about 50 different types! "http://www.cis.upenn.edu/~kostas/omni.html"
They keep referring to the notion of depth being used, but unless there is some big technology that they completely forgot to mention in the article & video, it just does the equivalent of pointing a camera into a bowl shaped mirror, allowing you to see in all 360 degrees at once. eg: "http://en.wikipedia.org/wiki/Omnidirectional_camera"
That is quite different to say it is truly 3D, since it is still a 2D image without depth, just that its wrapped around a circle shape instead of rectangle shape.
You guys think it is ridiculous that these websites often only pay about $0.60/hr, but thats because you guys are all living in USA or the West, where $5/hr is low. But most people live in 3rd world countries, where $0.60/hr is considered a reasonable income. That's why nearly all outsourcing is done in India or other poor countries, because $0.60/hr through Internet (with a possibility of earning several dollars per hour) seems as good as working a typical fulltime job that may also pay just $0.60/hr.
So the crazy thing is not that these websites pay less than $1/hr, its that the same job in the West will pay 10 or 20 times higher than in the rest of the world!
PS: I currently live in a farming town in Philippines where the average fulltime income is about $0.50/hr, so the prospect of earning that much through Internet seems pretty good.
It was already free for non-commercial use, so I guess Microsoft finally realised that not many companies are making money from robots (besides military robots & industrial robots), so it wasn't really earning them much money anyway.
Personally, I hated learning Microsoft Robotics Studio, it is great for doing a few basic things, but doing anything different than that requires a lot of things that must be learnt that have very little to do with your robotics project. The whole purpose of a robotics environment like this is so that robotics builders can concentrate on actual robot stuff, whereas MSRDS makes you waste so much time learning their proprietary formats & techniques that you could easily loose the fun in learning robots!
Thats my opinion anyway, as a someone with a Masters degree in Robotics Engineering that still has trouble using MS Robotics Design Studio!
Yeah for example in Dubai & UAE where about 80% of the people in the country are temporary hired workers from overseas, its common that even for the "blue-collar" jobs, your employer keeps your passport during your 1 or 2 yr contract to make sure you can't leave the job or country!
The PR2 robot has two eight-core i7 Xeon system servers on-board, each with 24 GB of RAM ..... over a 16-port gigabit Ethernet hub
Thats what happens when Google funds a robot: it becomes a mobile server rack!
But I must admit that its pretty impressive computing power to have on a mobile robot! Most research robots have relatively outdated embedded computing hardware compared to consumer laptops, so this will allow much more detailed algorithms to run.
Its true that military robots & technology are trying to be used to make warfare more "clean", so that the desired targets can be bombed with the least damage to civilians. That is good, atleast in theory, BUT at the end of the day, it is still all going to be controlled by the military leaders that don't necessarily know who are the civilians and who are the targets.
I have actually worked on a military robot for intended deployment in Iraq, and our military officer explained that when you are in a place like Iraq, you don't know who the enemy is and who the civilians are, because even if you see a 5yr old girl with her innocent looking grandmother and you ignore or help them, they are just as likely to try to secretly attack you as someone dressed in military uniform. So the US military in Iraq has to basically assume everyone that isn't a US soldier might be the enemy and therefore they can convince themselves that the ethical thing to do is kill anyone they see that they aren't completely sure is on their side.
So it doesn't matter whether the soldiers have basic weapons or latest military robots, they are still in the mind-set that any civilian can be considered part of the enemy's military.
The main advantage of military robots to the USA is that the countries that USA invades will be much poorer & less advanced countries than USA, so the enemy wont be able to make use of cutting-edge military technology compared to America.
If you don't believe me, put it this way: if Iraq had just as many military soldiers & robots fighting in USA as USA has in Iraq, do you still think people would see this the same way? The "Iraq War" and the "Afghanistan War" aren't wars, they are one-sided invasions, so its very different than if those countries were actually bombing America on a daily basis.
It has better dynamic range than anything your digital harddrives can offer
Actually it already exists: http://www.justin.tv/ (its called Lifecasting: http://en.wikipedia.org/wiki/Lifecasting_(video_stream))
Actually nearly every time someone uses neural networks for image processing, they use the same approach, of making a long vector from the image pixels and feeding it into an ANN. And you are right that it has a lot of associated problems, but the ones you gave aren't really a problem with their software.
Like you said, you might want to use images of different sizes, or images that are in different pixel locations, and like you said, you need to pre-process the images so that they are of the exact same size and generally the desired features are in the exact same position in all images. Any image recognition system needs to do this, because it definitely is the most important part of image recognition, but after you have done that, then I don't see why you can't use the ANN library mentioned in this article to do the actual image recognition part.
In other words, any software that can handle images of different sizes or features in different locations is likely to include the same sort of preprocessing steps as you would need to do when using this ANN software.
OpenCV does provide very basic image recognition abilities but I definitely welcome any new open-sourced libraries that tackle image recognition too.
and by then, your WPA Encryption would be the last of your worries!
so whats the square root of 5? i don't think the pedophile manual has that one!
Finally, some useful advice!
In 2006 & 2007 I was working at the American company that was working on this. I was working on a different project so I don't know the exact details, but I know they were using the BrainPort to let blind people see things (sending electrical spikes onto the tongue with a resolution of 64 points on the tongue). But it was actually funded by the military to allow hi-tech soldiers to get extra information through their tongue.
The demo they were working on at the time was to allow the soldier to get an idea of where their fellow soldiers where, through the BrainPort. After some practise, they can tell through their tongue that they have a soldier behind them on their left and 2 more coming on their right, for example. Here is a little description of the project: http://www.ihmc.us/research/projects/SensorySubstitution/
simple.
Yeah actually I read quickly through that already :-) But I was still confused because I know that in Australia we had a vote across the whole country, trying to decide whether Australia should become a Republic (having a President) or stay as a Monarch. It turns out that there's 2 meanings of the word "Republic", and so by some definitions, Australia is a Republic, and by other definitions, Australia is NOT a Republic! "http://www.ozpolitics.info/guide/topics/republic/"
No wonder I was confused!
Just make it a Google site for free and spend the rest on trips to Hawaii for "background research".
I'm no expert, but doesn't "Republic" just mean there is a President in the federal government? For example, Australia has a Prime Minister and no President, so its not called a Republic, whereas somewhere like Iran or America has a President, hence why they're called a Republic?