In order really pull off human-level AI, I think we'll need to model the brain at the subatomic level, because we probably aren't capable of understanding, truly understanding, how our intelligence works to the point that we could code it at a higher level than that. Basically you will have a complete human brain AND body, totally simulated. It will be just like a human - no smarter, no dumber.
So why not just have sex and make a real human?
Well, the advantage would (presumably) be that the simulated human has no rights, so you can make it do all sorts of tedious tasks that require human-level intelligence (such as sorting and classifying images based on content). And with enough CPU power, you can make the simulated human run at a faster speed, so it can complete complicated tasks more quickly. With lots of these things you could get quite a lot done.
Jack Nicklaus was 10 when he played his first nine holes of golf. He shot a 51, and by age 13 he broke 70 on 18 holes. Many people play for years and still regularly shoot 51 on nine holes. I started playing at about twelve and it took me several years before I could shoot below 50 on nine holes. Sometimes it really is talent!
Technically you are correct that the rovers' main camera is 1.3 megapixel...however, the camera takes color images by snapping 3 different pictures with 3 different filters. This corresponds more accurately to a 3.9 megapixel consumer camera, since all but a few consumer digital camera have one third of their pixels for red, one third for green, and one third for blue; the final full RGB image is created by interpolating the missing colors, so the true resolution of a 3.9 megapixel consumer camera is probably about the same as the rovers' 1.3 megapixel camera, assuming identical optics.
Another poster made the point that most of the images you see from the rovers are actually multiple images stitched together, resulting in even more 'megapixels' per image.
On a separate point, what distinguishes good from great cameras is not megapixels but optical quality. A terapixel camera with a pinhole lens would produce much lower resolution photos than a 6 megapixel Nikon with mulithousand dollar glass attached.
...if Leopard includes a virtualization environment, or if a Wine-like solution becomes popular, it would be so easy to run Windows software that users might hardly notice when an app is for Windows or OS X. In that environment, why would software vendors spend any resources on a special Mac port?
I hope to God/Jobs you aren't right, but I fear you might be. On the other hand, Apple appears to be betting that people like OS X more than Windows, and that by doing this they can get more people to buy Macs, increase their market share, and maybe even turn the tables on the application porting equation.
This is nice and all, but Apple just killed off any remaining incentive to develop or port native OS X games. What game developer in their right mind would bother, knowing that two or three years from now every Mac with up-to-date graphics hardware will also be able to run Windows?
Actually, switching to x86 forces all developers to write code that can be compiled under Xcode (and therefore strictly conforms to Cocoa standards).
Remember, Mac OS X is based on Nextstep/Openstep, which is partly why it was not THAT big of a deal to have OS X running on x86 this whole time.
When Apple first bought Next, there was a product called 'Openstep for Windows.' It was basically the whole Openstep framework, running on top of Windows. So you could write your Openstep code, and with a simple recompile it would run on a Windows machine, or any of the various machines running Nextstep.
So it isn't out of the question that Apple could switch to some sort of customized Windows with a rebadged Openstep for Windows running on top of it; and the recent switch to x86 would ensure that all Universal apps would be able to run on this version of Windows. Heck, if Apple is really clever, it might even be that you don't have to recompile your Universal Binary apps - they will just run as is.
If I own a business and that business brings in $100k in gross profit, without deductions, I pay tax on $100k. However, looking at the bigger picture, If my business is anything like the norm, only about 30% of that gross stays in my pocket. That means, I had to pay employees (who are taxed on that pay), advertising (which is taxed on the service provider), and office supplies (which were already taxed at the OfficeMax counter). I have to be able to deduct business expenses otherwise the remainder of the gross that I hold in my hand after business expenses will go, in total, the IRS and I end up having run a business that did $100k in profit and I, as the owner, have exact $0 to show for it (if I don't end up oweing.
Oops! You are confusing "personal deductions" and "business expenses." The business expenses would be subtracted from your business income and wouldn't even show up as "income" on your taxes.
To put it another way: If you run a business, you don't get taxed on revenue - you get taxed on profit.
Personal deductions come into play on your personal income, which can include business profits, salary from an employer, investment income, etc. Congress could theoretically eliminate personal deductions while still only taxing your profit.
As I said, later versions of the OS brought it up to 100%.
Mac OS X makes use of the 64-bit architecture in the only way they can, this is still a 32-bit world. Very few mainstream companies have yet to port their software because the 64-bit core is still fairly untested in mainstream computing.
WTF are you talking about? Have you ever heard of this little company called IBM? Yeah, they make and use the PowerPC 970 (aka the G5) in all sorts of mainstream, mission-critical, 64-bit environments.
The reason there aren't a lot of 64 bit apps for OS X is that the vast majority of the Cocoa frameworks are still 32 bit only. Apple has to make them 64 bit compatible before anyone else can write 64 bit apps that use them. So for now, the only 64 bit apps you can find are background/command-line only apps (Mathematica, for example, is technically two programs: the kernel, which runs in the background and is 64 bit, and the front end, which uses Cocoa and is 32 bit).
8 bits of brightness per subpixel which is fine for most situations... but duplicating nature can require about 20 bits
I'm not buying it. Where's the proof?
Check out http://kenrockwell.com/tech/ev.htm...the Light Value scale is logarithmic, each additional Light Value is twice the intensity of light of the previous value. The highest LV you'd see is about 20, the lowest about -15, but those wouldn't be in the same scene. Since LV 20 is basically looking straight into the sun, and LV 1 is typical outdoor scene at night, probably 20 bits is an exaggeration.
On the other hand, my Nikon D50 has 12 bits of dynamic range and it certainly is possible to both blow out highlights and underexpose shadows in the same scene.
Of course, this all gets more complicated by the fact that one usually applies some sort of non-linear gamma curve to remap the intensity levels to something more closely approximating the human eye's response, which means that 8-bit intensity coding using a gamma curve can almost capture the 12 bits of "linear" response of the Nikon sensor (I put "linear" in quotes because it seems more like an exponential to me, but whatever).
I am a freelancer mathematician (see http://www.northcountrynumerics.com/) . I work in seismic exploration, and also defense-related industries. I don't think it is possible to do this kind off work without having lots of personal connections, though; clients don't want to entrust some random person they've met once with a difficult and important mathematics problem. My work with my clients is much more like an academic collaboration (without the annoying emphasis on publications, though ironically I have more time for publication now than I did when I was in academia) than it is like an engineering or software development task.
The projects are also usually quite specialized, so you can't really walk in and solve someone's problem unless you aren't already quite knowledgeable in that particular sub-field of mathematics, and have a proven record of solving problems in that area.
> I drive a car without abs and tracking direccional system, it isn't that hard, btw in europe people tell you actually how to drive at classes:)
So do I, my wife's car has all the gadgets. I never get stuck and have never been in an accident that was my fault. I've driven cars with no ABS or traction control on lakes frozen like glass, did they teach you how to do that in your expensive european driving class?
My point was simply that just because the ABS or traction control engages, doesn't mean you are doing something wrong...
A good way to tell which of these things actually help is to look at which ones will get a discount on your insurance...ABS will do that, for example. The big insurance companies wouldn't do that if ABS didn't make you safer, because they only really care about their bottom line.
> If the ABS comes on then you are driving in a manner likely to endanger life. Most probably your own!
Obviously you've never driven on an icy road...on slick ice, if you just touch the brakes, even at 5 mph, the ABS will come on.
> Off the race track, hardly anyone would ever encounter a situation where traction control is needed.
Again I would refer to the aforementioned icy road...getting started on a road with snow on top of ice can be quite dicey without traction control...with traction control, just hit the gas and the car figures out how to get going.
Of course they are maximizing profits...what about your description is not maximizing profits? They are simply choosing to use different model for their profit than just "next quarter" or "next year," and choosing to maximize total profit rather than profit margin as a percentage of cost. You seem to think that maximizing profit means maximizing profit margin as a percentage, which of course it does not.
matlab is the only real option for numerical stuff
Actually, it isn't a real option for numerical stuff, unless you mean playing around/prototyping. For production-quality industrial numerical work, you have to write it yourself. I've had a number of consulting jobs where my task was simply to efficiently duplicate some matlab code in C/C++/Fortran..with the end result that it runs _hundreds_ of times faster, and can be modified/enhanced without relying on Matlab's proprietary toolboxes.
It is somewhat disconcerting to be scrolling around the area where I grew up, and see one half of a lake in full summer splendor, with boats frolicing and surrounded by green hills...and the other half of the lake is frozen solid.
The real problem with using 4*(1 - 1/3 + 1/5 - 1/7...) to approximate Pi, as any Calc II student knows, is not exactly rounding error but rather that it takes a freakin' long time to converge! The error when using n terms in this series is approximately 1/(2*n+1), which means to get 14 digits correct you need 5*10^13 terms. Rounding error comes into play only because you start to accumulate significant error when doing that many additions; almost every floating-point computation has rounding error though.
Price of stock is only meaningful when measured relative to earnings or dividends or something like that. $100/share earning -5 cents a share is different than $100/share earning $6/share. Hence the common use of P/E ratio in determining if a stock is over or underpriced.
Most of you are thinking of today's applications...but what about things like eye/head tracking, voice recognition, face recognition, telepresence, real-time cinema-quality CGI, etc...those are tasks requiring large-scale numerical computation, and they all might appear on your desktop in the not-too-distant future thanks to chips like CELL and its future ancestors.
Slashdot Mirror
In order really pull off human-level AI, I think we'll need to model the brain at the subatomic level, because we probably aren't capable of understanding, truly understanding, how our intelligence works to the point that we could code it at a higher level than that. Basically you will have a complete human brain AND body, totally simulated. It will be just like a human - no smarter, no dumber.
So why not just have sex and make a real human?
Well, the advantage would (presumably) be that the simulated human has no rights, so you can make it do all sorts of tedious tasks that require human-level intelligence (such as sorting and classifying images based on content). And with enough CPU power, you can make the simulated human run at a faster speed, so it can complete complicated tasks more quickly. With lots of these things you could get quite a lot done.
Jack Nicklaus was 10 when he played his first nine holes of golf. He shot a 51, and by age 13 he broke 70 on 18 holes. Many people play for years and still regularly shoot 51 on nine holes. I started playing at about twelve and it took me several years before I could shoot below 50 on nine holes. Sometimes it really is talent!
http://en.wikipedia.org/wiki/Jack_Nicklaus
And in Germany, people were mocking black soccer players by making monkey sounds and other racial epithets...oh wait that's today!
4 racism.html
http://www.nytimes.com/2006/06/04/sports/soccer/0
>Nope. It's 1/4 red, 2/4 green and 1/4 blue.
Crap! Well, forget everything I said then.
Technically you are correct that the rovers' main camera is 1.3 megapixel...however, the camera takes color images by snapping 3 different pictures with 3 different filters. This corresponds more accurately to a 3.9 megapixel consumer camera, since all but a few consumer digital camera have one third of their pixels for red, one third for green, and one third for blue; the final full RGB image is created by interpolating the missing colors, so the true resolution of a 3.9 megapixel consumer camera is probably about the same as the rovers' 1.3 megapixel camera, assuming identical optics.
Another poster made the point that most of the images you see from the rovers are actually multiple images stitched together, resulting in even more 'megapixels' per image.
On a separate point, what distinguishes good from great cameras is not megapixels but optical quality. A terapixel camera with a pinhole lens would produce much lower resolution photos than a 6 megapixel Nikon with mulithousand dollar glass attached.
I hope to God/Jobs you aren't right, but I fear you might be. On the other hand, Apple appears to be betting that people like OS X more than Windows, and that by doing this they can get more people to buy Macs, increase their market share, and maybe even turn the tables on the application porting equation.
This is nice and all, but Apple just killed off any remaining incentive to develop or port native OS X games. What game developer in their right mind would bother, knowing that two or three years from now every Mac with up-to-date graphics hardware will also be able to run Windows?
Actually, switching to x86 forces all developers to write code that can be compiled under Xcode (and therefore strictly conforms to Cocoa standards).
Remember, Mac OS X is based on Nextstep/Openstep, which is partly why it was not THAT big of a deal to have OS X running on x86 this whole time.
When Apple first bought Next, there was a product called 'Openstep for Windows.' It was basically the whole Openstep framework, running on top of Windows. So you could write your Openstep code, and with a simple recompile it would run on a Windows machine, or any of the various machines running Nextstep.
So it isn't out of the question that Apple could switch to some sort of customized Windows with a rebadged Openstep for Windows running on top of it; and the recent switch to x86 would ensure that all Universal apps would be able to run on this version of Windows. Heck, if Apple is really clever, it might even be that you don't have to recompile your Universal Binary apps - they will just run as is.
Oops, sorry, wrong story
If I own a business and that business brings in $100k in gross profit, without deductions, I pay tax on $100k. However, looking at the bigger picture, If my business is anything like the norm, only about 30% of that gross stays in my pocket. That means, I had to pay employees (who are taxed on that pay), advertising (which is taxed on the service provider), and office supplies (which were already taxed at the OfficeMax counter). I have to be able to deduct business expenses otherwise the remainder of the gross that I hold in my hand after business expenses will go, in total, the IRS and I end up having run a business that did $100k in profit and I, as the owner, have exact $0 to show for it (if I don't end up oweing.
Oops! You are confusing "personal deductions" and "business expenses." The business expenses would be subtracted from your business income and wouldn't even show up as "income" on your taxes.
To put it another way: If you run a business, you don't get taxed on revenue - you get taxed on profit.
Personal deductions come into play on your personal income, which can include business profits, salary from an employer, investment income, etc. Congress could theoretically eliminate personal deductions while still only taxing your profit.
As I said, later versions of the OS brought it up to 100%. Mac OS X makes use of the 64-bit architecture in the only way they can, this is still a 32-bit world. Very few mainstream companies have yet to port their software because the 64-bit core is still fairly untested in mainstream computing. WTF are you talking about? Have you ever heard of this little company called IBM? Yeah, they make and use the PowerPC 970 (aka the G5) in all sorts of mainstream, mission-critical, 64-bit environments. The reason there aren't a lot of 64 bit apps for OS X is that the vast majority of the Cocoa frameworks are still 32 bit only. Apple has to make them 64 bit compatible before anyone else can write 64 bit apps that use them. So for now, the only 64 bit apps you can find are background/command-line only apps (Mathematica, for example, is technically two programs: the kernel, which runs in the background and is 64 bit, and the front end, which uses Cocoa and is 32 bit).
8 bits of brightness per subpixel which is fine for most situations ... but duplicating nature can require about 20 bits
I'm not buying it. Where's the proof?
Check out http://kenrockwell.com/tech/ev.htm...the Light Value scale is logarithmic, each additional Light Value is twice the intensity of light of the previous value. The highest LV you'd see is about 20, the lowest about -15, but those wouldn't be in the same scene. Since LV 20 is basically looking straight into the sun, and LV 1 is typical outdoor scene at night, probably 20 bits is an exaggeration.
On the other hand, my Nikon D50 has 12 bits of dynamic range and it certainly is possible to both blow out highlights and underexpose shadows in the same scene.
Of course, this all gets more complicated by the fact that one usually applies some sort of non-linear gamma curve to remap the intensity levels to something more closely approximating the human eye's response, which means that 8-bit intensity coding using a gamma curve can almost capture the 12 bits of "linear" response of the Nikon sensor (I put "linear" in quotes because it seems more like an exponential to me, but whatever).
I am a freelancer mathematician (see http://www.northcountrynumerics.com/) . I work in seismic exploration, and also defense-related industries. I don't think it is possible to do this kind off work without having lots of personal connections, though; clients don't want to entrust some random person they've met once with a difficult and important mathematics problem. My work with my clients is much more like an academic collaboration (without the annoying emphasis on publications, though ironically I have more time for publication now than I did when I was in academia) than it is like an engineering or software development task.
The projects are also usually quite specialized, so you can't really walk in and solve someone's problem unless you aren't already quite knowledgeable in that particular sub-field of mathematics, and have a proven record of solving problems in that area.
> I drive a car without abs and tracking direccional system, it isn't that hard, btw in europe people tell you actually how to drive at classes :)
So do I, my wife's car has all the gadgets. I never get stuck and have never been in an accident that was my fault. I've driven cars with no ABS or traction control on lakes frozen like glass, did they teach you how to do that in your expensive european driving class?
My point was simply that just because the ABS or traction control engages, doesn't mean you are doing something wrong...
A good way to tell which of these things actually help is to look at which ones will get a discount on your insurance...ABS will do that, for example. The big insurance companies wouldn't do that if ABS didn't make you safer, because they only really care about their bottom line.
> If the ABS comes on then you are driving in a manner likely to endanger life. Most probably your own!
Obviously you've never driven on an icy road...on slick ice, if you just touch the brakes, even at 5 mph, the ABS will come on.
> Off the race track, hardly anyone would ever encounter a situation where traction control is needed.
Again I would refer to the aforementioned icy road...getting started on a road with snow on top of ice can be quite dicey without traction control...with traction control, just hit the gas and the car figures out how to get going.
Of course they are maximizing profits...what about your description is not maximizing profits? They are simply choosing to use different model for their profit than just "next quarter" or "next year," and choosing to maximize total profit rather than profit margin as a percentage of cost. You seem to think that maximizing profit means maximizing profit margin as a percentage, which of course it does not.
Crapity-crap, I just bought a dual 2.7 like a month ago! Oh well, maybe it will be worth something as a collectors item....
Actually, it isn't a real option for numerical stuff, unless you mean playing around/prototyping. For production-quality industrial numerical work, you have to write it yourself. I've had a number of consulting jobs where my task was simply to efficiently duplicate some matlab code in C/C++/Fortran..with the end result that it runs _hundreds_ of times faster, and can be modified/enhanced without relying on Matlab's proprietary toolboxes.
Yes, it is. Zoom out and it will become clear.
http://maps.google.com/maps?ll=44.962159395217896, -93.5434341430664&spn=0.008153915405273438,0.01179 0990829467773&t=k&hl=en
It is somewhat disconcerting to be scrolling around the area where I grew up, and see one half of a lake in full summer splendor, with boats frolicing and surrounded by green hills...and the other half of the lake is frozen solid.
The real problem with using 4*(1 - 1/3 + 1/5 - 1/7...) to approximate Pi, as any Calc II student knows, is not exactly rounding error but rather that it takes a freakin' long time to converge! The error when using n terms in this series is approximately 1/(2*n+1), which means to get 14 digits correct you need 5*10^13 terms. Rounding error comes into play only because you start to accumulate significant error when doing that many additions; almost every floating-point computation has rounding error though.
Price of stock is only meaningful when measured relative to earnings or dividends or something like that. $100/share earning -5 cents a share is different than $100/share earning $6/share. Hence the common use of P/E ratio in determining if a stock is over or underpriced.
Most of you are thinking of today's applications...but what about things like eye/head tracking, voice recognition, face recognition, telepresence, real-time cinema-quality CGI, etc...those are tasks requiring large-scale numerical computation, and they all might appear on your desktop in the not-too-distant future thanks to chips like CELL and its future ancestors.