In 2017 they have so far launched 14 vehicles, which is a lot better than the 8 successful and one failure they had in 2016. The R-7 (Russia) has launched 11 and the Ariane (Europe) 5, so they are definitely competitive. At the same time there were 60 successful (and no failure) R-7 launches in communist USSR in 1974.
Following the Tsiolkovsky equation, rockets do most of their acceleration when they are almost empty, see this graphics for the Saturn V. NASA actually had to turn off one of the Saturn V engine towards the end of the first stage burn otherwise acceleration would have been too great for the astronauts. So the last few minutes of the burn are important.
Basically the Musk approach is that you have to lift some of the fuel, which is not used for lifting but for landing. It may be more costly than it looks. Science is not alway "obvious", you have to crunch the numbers and see.
Now the Musk approach may work because his rocket first stages have many small engines and using only one for landing is enough. Most other rockets have a small number of large engines. It may be that other agencies did they math correctly but for their own rockets, which are not as well suited to returning and landing than Musk's.
But I'm pretty sure staying there will not be economical perhaps for centuries. Even "The case for Mars" says so. Humans are not very good at planning for centuries.
one person eats about 2000 pounds or about a metric ton of food a year. Water can presumably be largely recycled and some can be found on Mars, so we can ignore this. So you are right, this is probably adequate for a while.
Large companies waste money too. If you don't believe me, you should try working for one for a few months, look for waste and inefficiencies and report the results. The reality is that humans are not very good at running complex shows.
That system may be capable of running a level 5 autonomous vehicle one day, but no now out of the box. Since Nvidia only makes the hardware and associated libraries, the hard work of setting up systems with cameras and other sensors, training and certifying these systems still remains to be done.
Free speech does not mean that anyone can say anything at anytime. It means that the government cannot suppress some forms of speech sometimes. Some forms of speech are definitely banned like hate speech, incitation to crime, divulgation of intellectual property, and many others. Peddlers of false cures are not protected by free speech but could be brought to justice under the heading of Truth in Advertising. See Tina.org
Overhead cables are often uninsulated because insulation is heavy. This is definitely the case for very high voltage lines. Insulation is provided by the distance to the powerline.
Birds don't get fried because both their legs rest on the same cable. There is no difference of potential between them, hence no current.
It is not an unfair comparison, since all we can do today is to try to simulate the brain and see what happens.
The Snapdragon CPU does not make 10^18 switches per seconds. An easy calculation shows that if it did that it would instantly vaporise itself. The CPU is all synchronized and most of the hardware is memory, that changes very little every cycle.
At this point we know that there exists this phenomenon called "intelligence" and that most of it seems to takes place in the human brain. We know a great deal of details on how the human brain is spatially organized but not how the various parts fit and what they do exactly.
Most of the AI-based fear is based on the assumption that somehow, pure intelligence above some level can somehow move mountains. However, all go ahead and read Kant's "A Critic of Pure Reason" to see how perfectly sane and intelligent minds can waste their precious time on pointless pursuits.
A superhuman intelligence bred by humans will definitely need to develop new physics to grow by itself.
However, you cannot just *invent* new physics to exploit it. We already know a great deal about the physical world and knowing more involves a great deal of investment and time. Think CERN and its thousands of top-notch physicists who spent years discovery a single new particle they already knew must exist. Think nuclear fusion, think gravitational waves. In each case the physics was already known and proved to be correct, but the experimental validation, without which there would be nothing but wasteful pure reason (think string theory, with apologies) took a great deal of effort.
So yeah, a superhuman superintelligence could probably come up with plenty of fancy new theories that it could exploit for it to grow, but it will need to experiment to see which one is correct and to go in the right direction. This is not going to take mere seconds but potentially decades.
This is not quite correct. For a start, the Apple 3TB fusion drive comes paired with a 128GB SSD. This is much larger than the 8GB cache of typical SSHD, which takes care of your slow read performance. Second according to Apple, their system is not just a cache, it "intelligently" stores most used as well as most recent data and applications. As such it is closer to a tiered access solution than a cache.
Yes you can do something similar with a HD + a SSD, but AFAIK not trivially. The Intel SSD SRT caching solution is only that, a cache, and I believe it is limited to 64GB of SSD.
Note that using a SSH as a straight read/write cache can wear it out quickly.
Nobody outside of Apple really knows what the fusion drive mechanism does because Apple does not open-source the technology. What you wrote is basically straight from Apple's marketing department. We know that they are very good, this does not mean the technology is that great. Early reports were saying that the tech was basically on par with SSHD (a HDD + some amount of Flash as cache) or a bit better, but not a game changer.
1- Nobody sells large flash drives at a consumer-friendly price point. 2- OS X allows a smaller flash drive and a larger HDD to appear as a single drive, where the flash drive acts as a large cache. They offer this feature for free.
The hippocampal prosthesis is a proof of concept in rats. Have you actually read the wikipedia page? It is full of "must", "should", "will", "may" and so on. Not exactly as if it were working.
So far we've been able to semi-conclusively simulate the brain of C. Elegans, a brain with 302 neurons. This is debated, by the way. The Human brain has about 10^11 neurons. That is approximately 8-9 orders of magnitude more. That represents 2^30 or 30 doublings, or another 50-60 years of "Moore's law", which is already flattening out.
I'm not exactly optimistic that this line of research will work in the near future. In a century, if we manage to survive as a species and continue to perform scientific research at a good clip, maybe.
Slashdot Mirror
Not so much the seal, the cost of actually pumping that air out of a huge, long tunnel.
In 2017 they have so far launched 14 vehicles, which is a lot better than the 8 successful and one failure they had in 2016. The R-7 (Russia) has launched 11 and the Ariane (Europe) 5, so they are definitely competitive. At the same time there were 60 successful (and no failure) R-7 launches in communist USSR in 1974.
Following the Tsiolkovsky equation, rockets do most of their acceleration when they are almost empty, see this graphics for the Saturn V. NASA actually had to turn off one of the Saturn V engine towards the end of the first stage burn otherwise acceleration would have been too great for the astronauts. So the last few minutes of the burn are important.
Basically the Musk approach is that you have to lift some of the fuel, which is not used for lifting but for landing. It may be more costly than it looks. Science is not alway "obvious", you have to crunch the numbers and see.
Now the Musk approach may work because his rocket first stages have many small engines and using only one for landing is enough. Most other rockets have a small number of large engines. It may be that other agencies did they math correctly but for their own rockets, which are not as well suited to returning and landing than Musk's.
... the answer is no, and the author agrees with me.
So far DL is great progress but still statistical methods.
But I'm pretty sure staying there will not be economical perhaps for centuries. Even "The case for Mars" says so. Humans are not very good at planning for centuries.
Or Musk wants to be in the news and seen as a shaker and mover (which he is).
one person eats about 2000 pounds or about a metric ton of food a year. Water can presumably be largely recycled and some can be found on Mars, so we can ignore this. So you are right, this is probably adequate for a while.
Large companies waste money too. If you don't believe me, you should try working for one for a few months, look for waste and inefficiencies and report the results. The reality is that humans are not very good at running complex shows.
That system may be capable of running a level 5 autonomous vehicle one day, but no now out of the box. Since Nvidia only makes the hardware and associated libraries, the hard work of setting up systems with cameras and other sensors, training and certifying these systems still remains to be done.
Once it is trained, maybe you don't. Training a deep CNN is computationally expensive.
Not general purpose teraflops. Useful if you want to do lots of fixed-sized matrix multiplications but not much else.
Free speech does not mean that anyone can say anything at anytime. It means that the government cannot suppress some forms of speech sometimes. Some forms of speech are definitely banned like hate speech, incitation to crime, divulgation of intellectual property, and many others. Peddlers of false cures are not protected by free speech but could be brought to justice under the heading of Truth in Advertising. See Tina.org
No, it disproved the existence of aether.
Overhead cables are often uninsulated because insulation is heavy. This is definitely the case for very high voltage lines. Insulation is provided by the distance to the powerline.
Birds don't get fried because both their legs rest on the same cable. There is no difference of potential between them, hence no current.
It is not an unfair comparison, since all we can do today is to try to simulate the brain and see what happens.
The Snapdragon CPU does not make 10^18 switches per seconds. An easy calculation shows that if it did that it would instantly vaporise itself. The CPU is all synchronized and most of the hardware is memory, that changes very little every cycle.
And you know this how, exactly ?
At this point we know that there exists this phenomenon called "intelligence" and that most of it seems to takes place in the human brain. We know a great deal of details on how the human brain is spatially organized but not how the various parts fit and what they do exactly.
Most of the AI-based fear is based on the assumption that somehow, pure intelligence above some level can somehow move mountains. However, all go ahead and read Kant's "A Critic of Pure Reason" to see how perfectly sane and intelligent minds can waste their precious time on pointless pursuits.
A superhuman intelligence bred by humans will definitely need to develop new physics to grow by itself.
However, you cannot just *invent* new physics to exploit it. We already know a great deal about the physical world and knowing more involves a great deal of investment and time. Think CERN and its thousands of top-notch physicists who spent years discovery a single new particle they already knew must exist. Think nuclear fusion, think gravitational waves. In each case the physics was already known and proved to be correct, but the experimental validation, without which there would be nothing but wasteful pure reason (think string theory, with apologies) took a great deal of effort.
So yeah, a superhuman superintelligence could probably come up with plenty of fancy new theories that it could exploit for it to grow, but it will need to experiment to see which one is correct and to go in the right direction. This is not going to take mere seconds but potentially decades.
More like 30,000 deaths annually in the US alone. Basically one 9/11-size event *every month*.
That's the (over) price of the SDD.
This is not quite correct. For a start, the Apple 3TB fusion drive comes paired with a 128GB SSD. This is much larger than the 8GB cache of typical SSHD, which takes care of your slow read performance. Second according to Apple, their system is not just a cache, it "intelligently" stores most used as well as most recent data and applications. As such it is closer to a tiered access solution than a cache.
Yes you can do something similar with a HD + a SSD, but AFAIK not trivially. The Intel SSD SRT caching solution is only that, a cache, and I believe it is limited to 64GB of SSD.
Note that using a SSH as a straight read/write cache can wear it out quickly.
Nobody outside of Apple really knows what the fusion drive mechanism does because Apple does not open-source the technology. What you wrote is basically straight from Apple's marketing department. We know that they are very good, this does not mean the technology is that great. Early reports were saying that the tech was basically on par with SSHD (a HDD + some amount of Flash as cache) or a bit better, but not a game changer.
The closest open-source equivalent would be the ZFS filesystem, which offers several levels of caching with its ZIL and L2ARC features. These do offer very good performance improvements.
1- Nobody sells large flash drives at a consumer-friendly price point.
2- OS X allows a smaller flash drive and a larger HDD to appear as a single drive, where the flash drive acts as a large cache. They offer this feature for free.
You can pair any two drives you want in this way. I'm not sure where you get the premium idea.
The hippocampal prosthesis is a proof of concept in rats. Have you actually read the wikipedia page? It is full of "must", "should", "will", "may" and so on. Not exactly as if it were working.
So far we've been able to semi-conclusively simulate the brain of C. Elegans, a brain with 302 neurons. This is debated, by the way. The Human brain has about 10^11 neurons. That is approximately 8-9 orders of magnitude more. That represents 2^30 or 30 doublings, or another 50-60 years of "Moore's law", which is already flattening out.
I'm not exactly optimistic that this line of research will work in the near future. In a century, if we manage to survive as a species and continue to perform scientific research at a good clip, maybe.
I think you mean always 50 years off.
That is not a generative model of intelligence, at best a critical description of some of its aspects.