Slashdot ate my formatting but it was less than $100k where tape robots do not break even with offline Virtual Tape or disk drives.
And I'm not talking about a single tape drive to backup a few hundred gigabytes of data, I'm talking about systems that are several hundreds of terabytes, that requires multiple streams/heads. Last time I specced one out, even with LTO5 I was looking for at least 15 tapes to finish a single backup and it just couldn't keep up, by the time it was done, I had to restart it.
The price in blood is not due to weapons, it's due to a large portion of the population that chooses to remain undereducated and idolizes a culture of perpetual non-integration, guns and criminal activity.
Farmers and hunters across the world have guns and other weapons including the U.K., Australia and New Zealand yet don't have any problems with weapons in their cities.
Offline backups are fine, you don't have to do them on tape anymore. Most people however have never worked or cannot afford modern tape. A backup is better than a badly working, slow or intervention-prone backup which is synonymous to cheap tape system offers ($100k)
Says plenty of people that think their home insurance covers them against storms, hail and floods. Whoops, we just declared it an act of god, you're (not) in a flood plain or wind speeds above 60mph are tropical tornado's, not regular tornado's.
Nobody insures against petty crimes, that's why you have a deductible on your insurance in the first place. Mandating insurance just means people will go to the cheapest offers, which usually covers nothing with huge deductibles.
In that case insurances would simply declare everything 'not a crime' so they wouldn't have to pay out. At some point you'll need a form of regulation in those systems unless you're advocating anarchy and you would have to defend yourself against bands of roving bandits by force.
Bail is posted from police stations, not state prisons. You can make a reasonable set of calls free of charge while you're being arraigned and before you go to an actual prison, such as to call for an attorney. They can't charge you money for anything before you are convicted.
Encoding something with that method works well enough for very short, well-defined animations. Even that would require more than a few kB but it's technically feasible to procedurally generate something using external sources. But even then, the definitions themselves would be both very precise and vary from system to system (eg. like MIDI or MOD files depend on what you have in your local instrument bank). You'd also have to rely on downloading immense amounts of data from various sources, so you're not really compressing anything, you're just re-distributing the final delivery.
It indeed would not work on all possible inputs unless your search space is capable of encompassing every single piece of data ever collected. Which, as above, just moves the "compression" job somewhere else and highly depends on the availability of resources. Both 1 and 2 are results from the exact same argument.
With 3, the problem is any pointer to the data you're trying to "encode" will be equal to or greater than the data you're trying to compress. Even with the ~2000 exabytes of data we currently have "stored" in the world, the pointer to any point in it would be immensely huge. A single pointer to any segment of it would be several terabytes large.
Sloot claimed a lossless algorithm. But even lossy algorithms have their limits, as I said before there are limits and there is a point where you'll compress everything so much, it turns everything into noise. As you encode something using lossy algorithms it's entropy increases (I think it's called Shannon entropy). You can basically encode but as you do, you add noise your ideal algorithm has to specify how "much" noise can be added, for humans it doesn't take long before things start looking weird and different or remove key elements of the image.
And if you fed that blue screen through a DAC, a few noise generators and bandwidth filters and then through an ADC again, would it still compress to that?
But it's not the movie then, you could encode a lot in the movie, eg. Google Street View locations etc to save on space but it still wouldn't be the movie as intended. Good enough for things like demonstrations (eg. Flash used similar methods) and enhanced reality but we're talking about true compression of random video into the format.
You're going in the wrong direction though. Music you can encode in MIDI/MOD but it's not really the song, it's just the "score" and then your computer fills it in with instruments in it's local instrument bank (which is why your MIDI may sound different than mine). You're not really encoding the song output of eg. an orchestra where someone in the audience could cough or clap into a MIDI file.
To do something similar for video, you're talking about giving someone the storybook to the movie and the computer would generate the entire movie, using pre-programmed likenesses of eg. Leonardo DiCaprio and a handful of other actors to regenerate the movie. It's possible for eg. video games but to encode a movie from film (where every frame is nearly completely noise to a computer) you can't use this method, modern methods track the movements of blocks of colors and with various other tricks try to encode that and with that we can get very decent lossy compression (300:1 or more, you could even go 1000:1 in some cases before it becomes unrecognizable). The compression talked about here is 40M:1 or better, it's not just a magnitude better, this was ~20 years ago so we're talking about competing with Ping and MPEG2 compression.
You haven't even heard of how or why HVDC works so no, I'm not taking your opinion on the matter seriously. HVDC is not viable for the few hundred volt you would get out of solar panels.
WINE supports running 16-bit Windows apps on 64-bit Linux. The limitation is in the Windows OS because they chose to remove the necessary code so you'd buy into their virtualization platform.
Exactly my point. Conversion from 12VDC to 100VDC (or whatever it is, but I would assume somewhere around US net-voltage for simplicity sake) requires a pass through AC (DC-DC conversion is really DC -> ~30-100kHz AC -> DC). Also transporting the current from the panels to wherever it needs to go over DC incurs significant losses unless you are capable of getting above ~10kVDC where transmission over DC makes sense again.
Actually it's a good thing because carbon credits are exactly that, credits you can exchange for carbon fuels. The goal was to have richer nation pay for renewables and the money they save on carbon fuels would go to poor nations so they can better themselves by paying for the cheaper carbon fuels and get themselves (hopefully) out of poverty, just like previous aid to Africa has helped them not be part of the third world anymore.
If you want to be at the forefront of AI you should first invent it and to do that, you need to define intelligence and individualism. I think brain research would be your field to try.
You seem to want to go into statistical classification and database systems, neither of which are new and have well defined paths via mathematical statistics, programming and systems design. The fact that you think learning a particular language will help shows that you are missing the point of modern IT systems but if anything, if you want to go into this field and do research, that is done in R, MATLAB, some C, Python and Java.
If you want to go into commercial AI, make an app, some news releases and a site, get a bunch of investors and hope Google or Apple buys you out before the jig is up.
1) still need water, sewage and permits and property taxes. This also makes the local grid less stable and diverse. 2) still need to burn carbon in order to mine heavy metals and transport them batteries, the electronics and steel from China etc without much though to eventual disposal. 3) and neither is a utility company beholden to them, what if future governments follow EU lead and mandate dumping excess energy from solar plants into the market. Also, any conversion of power involves AC, even converting the 8-15V'DC' from the panels. And transporting that DC any distance longer than a few meters will almost always involve AC, 12V/100A DC over the length of the football field size panel installation will require some REALLY thick cabling or more like a rail.
Not necessarily, I still have to keep PPC systems around for some legacy stuff. The nice thing is that Apple has been pushing for developers for cross platform because they tend to switch (from Motorola to IBM to Intel to Intel64 to ARM). Also a lot of it is Unix, cross platform is sort of built into the C/C++ language while Microsoft has always been pushing for its own lock-in so now you have things that are written for 16 bit API which cannot be compiled to a 32 bit API (true for anything that starts with Visual or ends with.NET)
Usually it's the other way around. Hardware is easy to replace. Even the latest Intels will run in real or protected mode if you need them to. It's the software that doesn't always get updates or if you got suckered into closed source software, your upgrade path is expensive.
Siemens MRI systems run on Windows XP 64 bit and the imager actually runs Linux (actually Linux controls a bunch of DSP programmed in multiple full length FPGA array cards).
As far as I'm aware, some GE MRI also run Linux although they used to be a Sun shop (Solaris) and still some older 1.5T systems which they still sell runs on Octanes (SGI).
MRIs were mostly developed through the late 90s through 2000s and the OS choices reflect it, Microsofts offerings weren't nearly stable enough for cutting edge technologies. You basically see a huge gap between DOS and Windows XP SP2 although many DOS systems eventually got converted to have the Windows shell, true 95-ME apps are nearly absent.
64 bit ms office doesn't work as well as 32 bit. This has been known since at least 2 office versions ago. Also Office plugins don't work because they plugin directly to Office memory space, not something like sockets or other common protocol.
Slashdot Mirror
Slashdot ate my formatting but it was less than $100k where tape robots do not break even with offline Virtual Tape or disk drives.
And I'm not talking about a single tape drive to backup a few hundred gigabytes of data, I'm talking about systems that are several hundreds of terabytes, that requires multiple streams/heads. Last time I specced one out, even with LTO5 I was looking for at least 15 tapes to finish a single backup and it just couldn't keep up, by the time it was done, I had to restart it.
The price in blood is not due to weapons, it's due to a large portion of the population that chooses to remain undereducated and idolizes a culture of perpetual non-integration, guns and criminal activity.
Farmers and hunters across the world have guns and other weapons including the U.K., Australia and New Zealand yet don't have any problems with weapons in their cities.
But step 1 did?
I'd reply: My Solaris box says user "root" doesn't exist.
Offline backups are fine, you don't have to do them on tape anymore. Most people however have never worked or cannot afford modern tape. A backup is better than a badly working, slow or intervention-prone backup which is synonymous to cheap tape system offers ($100k)
Says plenty of people that think their home insurance covers them against storms, hail and floods. Whoops, we just declared it an act of god, you're (not) in a flood plain or wind speeds above 60mph are tropical tornado's, not regular tornado's.
Nobody insures against petty crimes, that's why you have a deductible on your insurance in the first place. Mandating insurance just means people will go to the cheapest offers, which usually covers nothing with huge deductibles.
In that case insurances would simply declare everything 'not a crime' so they wouldn't have to pay out. At some point you'll need a form of regulation in those systems unless you're advocating anarchy and you would have to defend yourself against bands of roving bandits by force.
Bail is posted from police stations, not state prisons. You can make a reasonable set of calls free of charge while you're being arraigned and before you go to an actual prison, such as to call for an attorney. They can't charge you money for anything before you are convicted.
Encoding something with that method works well enough for very short, well-defined animations. Even that would require more than a few kB but it's technically feasible to procedurally generate something using external sources. But even then, the definitions themselves would be both very precise and vary from system to system (eg. like MIDI or MOD files depend on what you have in your local instrument bank). You'd also have to rely on downloading immense amounts of data from various sources, so you're not really compressing anything, you're just re-distributing the final delivery.
It indeed would not work on all possible inputs unless your search space is capable of encompassing every single piece of data ever collected. Which, as above, just moves the "compression" job somewhere else and highly depends on the availability of resources. Both 1 and 2 are results from the exact same argument.
With 3, the problem is any pointer to the data you're trying to "encode" will be equal to or greater than the data you're trying to compress. Even with the ~2000 exabytes of data we currently have "stored" in the world, the pointer to any point in it would be immensely huge. A single pointer to any segment of it would be several terabytes large.
You can't make Python the dominant language because half of it has to be implemented in C in order to get it to work properly.
That and lack of clear program structure definition (braces) will make sure it will always be a scripting language like PHP.
Sloot claimed a lossless algorithm. But even lossy algorithms have their limits, as I said before there are limits and there is a point where you'll compress everything so much, it turns everything into noise. As you encode something using lossy algorithms it's entropy increases (I think it's called Shannon entropy). You can basically encode but as you do, you add noise your ideal algorithm has to specify how "much" noise can be added, for humans it doesn't take long before things start looking weird and different or remove key elements of the image.
And if you fed that blue screen through a DAC, a few noise generators and bandwidth filters and then through an ADC again, would it still compress to that?
But it's not the movie then, you could encode a lot in the movie, eg. Google Street View locations etc to save on space but it still wouldn't be the movie as intended. Good enough for things like demonstrations (eg. Flash used similar methods) and enhanced reality but we're talking about true compression of random video into the format.
You're going in the wrong direction though. Music you can encode in MIDI/MOD but it's not really the song, it's just the "score" and then your computer fills it in with instruments in it's local instrument bank (which is why your MIDI may sound different than mine). You're not really encoding the song output of eg. an orchestra where someone in the audience could cough or clap into a MIDI file.
To do something similar for video, you're talking about giving someone the storybook to the movie and the computer would generate the entire movie, using pre-programmed likenesses of eg. Leonardo DiCaprio and a handful of other actors to regenerate the movie. It's possible for eg. video games but to encode a movie from film (where every frame is nearly completely noise to a computer) you can't use this method, modern methods track the movements of blocks of colors and with various other tricks try to encode that and with that we can get very decent lossy compression (300:1 or more, you could even go 1000:1 in some cases before it becomes unrecognizable). The compression talked about here is 40M:1 or better, it's not just a magnitude better, this was ~20 years ago so we're talking about competing with Ping and MPEG2 compression.
You haven't even heard of how or why HVDC works so no, I'm not taking your opinion on the matter seriously. HVDC is not viable for the few hundred volt you would get out of solar panels.
WINE supports running 16-bit Windows apps on 64-bit Linux. The limitation is in the Windows OS because they chose to remove the necessary code so you'd buy into their virtualization platform.
WINE supports running 16-bit Windows apps on 64-bit Linux
Exactly my point. Conversion from 12VDC to 100VDC (or whatever it is, but I would assume somewhere around US net-voltage for simplicity sake) requires a pass through AC (DC-DC conversion is really DC -> ~30-100kHz AC -> DC). Also transporting the current from the panels to wherever it needs to go over DC incurs significant losses unless you are capable of getting above ~10kVDC where transmission over DC makes sense again.
Actually it's a good thing because carbon credits are exactly that, credits you can exchange for carbon fuels. The goal was to have richer nation pay for renewables and the money they save on carbon fuels would go to poor nations so they can better themselves by paying for the cheaper carbon fuels and get themselves (hopefully) out of poverty, just like previous aid to Africa has helped them not be part of the third world anymore.
At least these two stories are from different perspectives: https://m.slashdot.org/story/3...
The story stays the same - don't fuck over your admins and have proper procedure and backup.
If you want to be at the forefront of AI you should first invent it and to do that, you need to define intelligence and individualism. I think brain research would be your field to try.
You seem to want to go into statistical classification and database systems, neither of which are new and have well defined paths via mathematical statistics, programming and systems design. The fact that you think learning a particular language will help shows that you are missing the point of modern IT systems but if anything, if you want to go into this field and do research, that is done in R, MATLAB, some C, Python and Java.
If you want to go into commercial AI, make an app, some news releases and a site, get a bunch of investors and hope Google or Apple buys you out before the jig is up.
1) still need water, sewage and permits and property taxes. This also makes the local grid less stable and diverse.
2) still need to burn carbon in order to mine heavy metals and transport them batteries, the electronics and steel from China etc without much though to eventual disposal.
3) and neither is a utility company beholden to them, what if future governments follow EU lead and mandate dumping excess energy from solar plants into the market. Also, any conversion of power involves AC, even converting the 8-15V'DC' from the panels. And transporting that DC any distance longer than a few meters will almost always involve AC, 12V/100A DC over the length of the football field size panel installation will require some REALLY thick cabling or more like a rail.
Not necessarily, I still have to keep PPC systems around for some legacy stuff. The nice thing is that Apple has been pushing for developers for cross platform because they tend to switch (from Motorola to IBM to Intel to Intel64 to ARM). Also a lot of it is Unix, cross platform is sort of built into the C/C++ language while Microsoft has always been pushing for its own lock-in so now you have things that are written for 16 bit API which cannot be compiled to a 32 bit API (true for anything that starts with Visual or ends with .NET)
Usually it's the other way around. Hardware is easy to replace. Even the latest Intels will run in real or protected mode if you need them to. It's the software that doesn't always get updates or if you got suckered into closed source software, your upgrade path is expensive.
Siemens MRI systems run on Windows XP 64 bit and the imager actually runs Linux (actually Linux controls a bunch of DSP programmed in multiple full length FPGA array cards).
As far as I'm aware, some GE MRI also run Linux although they used to be a Sun shop (Solaris) and still some older 1.5T systems which they still sell runs on Octanes (SGI).
MRIs were mostly developed through the late 90s through 2000s and the OS choices reflect it, Microsofts offerings weren't nearly stable enough for cutting edge technologies. You basically see a huge gap between DOS and Windows XP SP2 although many DOS systems eventually got converted to have the Windows shell, true 95-ME apps are nearly absent.
64 bit ms office doesn't work as well as 32 bit. This has been known since at least 2 office versions ago. Also Office plugins don't work because they plugin directly to Office memory space, not something like sockets or other common protocol.