My partner just replaced her Nokia Lumia 1020. I really liked the UI and it never crashed, but it did have an issue where the sound subsystem would fall over and then nothing would play sound, which meant that she had to test the alarm before she went to sleep each night or it have died during the day and be silent in the morning. She replaced it with an Android device because of the complete lack of third-party app support on Windows phone.
At 1g acceleration, it takes around 11 seconds to travel half the distance in this tube (assuming that the other half is spent decelerating). That's a peak speed of 110m/s, or around 245 miles per hour, so this train had less horizontal acceleration than humans experience vertically just by being on this planet. Give them a comfy chair and they'll happily manage that level of acceleration for 30 seconds to a minute. And you can always trade a little bit time for comfort. Half the acceleration and accelerate for two minutes instead of one and you'll add two minutes (one at each end) to the total travel time, which won't make much difference in a half-hour journey.
There's a story from the '70s about an artillery control system that used neural networks to classify enemy targets and civilians. At the first live-fire demonstration, it immediately targeted and destroyed the general's car. It turned out that they'd trained it to recognise things seen in daylight as civilian and things seen at night as enemy vehicles. The project was cancelled. Something similar happened when Google's face tagging software learned that any dark-coloured face was a gorilla, because someone had tagged a gorilla in their photo and that was the only dark face in their training set. I was recently at a presentation by Nokia Bell Labs where they'd trained a neural network to classify urban scenes by beauty - looking at their results, it turned out that they'd built a complex system to count the number of trees in a scene, everything other architectural feature was ignored.
Machine learning systems are good for generating correlations. I often say that it's the technique that you use when you don't really understand the problem that you're trying to solve, but you have a lot of cheap compute to throw at it. You'll end up with a system that has detected some unknown correlation and has unknown error rates and unknown failure modes. But it's probably better than nothing...
Sometimes those are worth listening to. I read a few reviews of a wireless access point that said it was difficult to configure. I assumed that the reviewers were idiots. I have a PhD in computer science, so I was pretty sure I could figure it out. It turns out that the manufacturers had decided that technical terms were confusing and so they'd make up their own terms for every single thing in the configuration interface. It took hours of a combination of trial and error, poking the thing remotely with diagnostic tools, and reading forum posts by other people who had figured out what some of them meant, to work out how to translate from their nonsense wibble into standard networking terminology.
Mod this up. Most reviews are useless when comparison shopping because the people writing them have a sample size of one. I recently bought a new electric shaver. Most of the reviews are from people who have owned maybe one other shaver in their life. For a product with 600 reviews, 500 of them are left within a few hours of the new one arriving, so all that you really know from them is that it came in a box and didn't break in the first use. The only useful reviews were the ones where someone actually compared it to others that are still available.
Market value for the software, or a copy of the software? There are two parts involved in software development: writing the software and copying it. Writing it costs money, copying it is essentially free. Open source makes it harder to charge money for copying software, but makes it easy to charge money for writing it. In contrast, proprietary off-the-shelf software involves writing software for free and then charging to make copies of it. Which do you think makes more economic sense?
The problem with that comparison is that people refer to tasers as 'non-lethal' weapons, rather than their official designation of 'less-lethal'. If you pull a gun and shoot someone and they die, you're not surprised. If you pull a taser and shoot someone and they die, then you are. This means that people are a lot more willing to shoot someone with a taser.
Data centres produce a lot of jobs for sysadmin / SRE people, software developers, and so on. The problem for the state is that most of these jobs have no reason to be geographically close to the data centre. You need a few people to pull dead machines and rack replacements, and to maintain the power / cooling systems, everything else is remote.
I will happily create as many jobs as you want for $1m/job. I guarantee that the jobs will be around for at least 10 years. Even counting overheads, that's enough that I can pocket half the money and pay people to do nothing.
If you look through the "integrations" list on this page you can choose your favorite
I had a look through this list the first time that you posted it, but none of them seem applicable.
If you don't have any favorites, one of the Backblaze IT people here uses "Duplicity Linux" to do EXACTLY what you describe
Duplicity doesn't seem to be able to do this. It maintains its own change logs, which ZFS gives me for free. I can do zfs send and get a stream that contains the changes since the last snapshot. I want to send this stream, encrypted, to B2, named such that if I need to recover I can easily stream the snapshots back, in order, so that zfs receive can reapply them to the filesyystem. I also want to configure the ACLs on the remote backups so that the credentials that the machine that's being backed up has can only create new files, it cannot modify or delete old ones, so that if that machine is compromised it cannot affect old snapshots.
Steve Jobs died October 5, 2011. Just before that, AAPL was at $57.83/share, which is the lowest that it's been since then, aside from a small dip in the weeks immediately after his death. It's now at a bit over $159/share. That's around a 16% annual return on investment. What did you buy when you sold your AAPL shares that gave you a better return?
A fixed-wing aircraft uses something between 2 and 3 *times* as much energy per km.
As much energy as what? A ground vehicle? For the same weight or per passenger? Light aircraft are typically lighter than cars because they are less likely to be involved in collisions and so don't need roll cages and so on, so I'll assume that it's per passenger. Aircraft can go in a straight line, whereas ground vehicles often have to travel two sides of a triangle or three sides of a square, so even a factor of 2 is not that bad, if you can take off and land near you origin and destination.
Fine for many purposes. I have a laptop. Stuff that I'm actively working on is in a remote git repo. Everything is backed up to a NAS in my house. I'd like to do off-site backups of that. The only time that I'd need the recovery urgently is if my laptop and my NAS were both fried at the same time.
I back up computers in my house to a FreeBSD NAS and I'm looking for an off-site backup solution. I want to be able to send ZFS snapshots, encrypted, to a remote location. Do you know of any tools that work well for this kind of use, or do I need to roll my own.
Many modern smartphones have HDMI out and can be plugged into a projector or HD TV, where you could tell the difference. Of course, in most such situations, you'd probably prefer to use WiFi, but if mobile data is your only option then this could be annoying.
That definition of Network Neutrality is the one that's pushed by ISPs, not by NN advocates. Typical NN definitions allow differentiating based on traffic type, but with some tight constraints (e.g. you can put things into latency-sensitive, jitter-sensitive, and bandwidth-sensitive buckets, but you can't treat one latency-sensitive protocol differently from another). QoS explicitly is allowed by all except for the straw-man NN definition used by ISPs.
What about your sharp kitchen knives, do you have to enter an access code before using the blade or do you just go for it?
My sharp kitchen knives came in a knife block the prevents me from accidentally grabbing the blade. It has handles that protect my hand. If someone sold a knife that had the blade going the full length to where the user was expected to hold it, then they'd be unpopular. Most devices come with basic safety designs and are safe in their normal and intended mode of use.
I'm not sure I totally agree with your classification, but it's far more common for me to meet a scientist with a solid knowledge of history, art, and literature than an arts or humanities person with a solid understanding of science. More importantly, scientists are more likely to be embarrassed about their lack of knowledge, whereas humanities scholars display almost pride in their lack of understanding of physics or computer science.
I came here to say the same thing. You can now buy Oreos in the UK, but only one flavour, and few people eat them as they're inferior to (and more expensive than) custard creams.
Maybe it has to do with it being too difficult, or perhaps even impossible, to upgrade to a newer version.
I only just upgraded from 5.1.1. I bought a Moto G because, at the time, Motorola was owned by Google and I expected good long-term support. It shipped with 4.3, got a quick upgrade to 4.4, eventually got an update to 5, and hasn't had security updates for well over a year. I reflashed it with LineageOS to get a new version (7.1), but that's well beyond the abilities of most users. The hardware is still completely adequate for all of the apps that I want to run, so I have no desire to replace it. I suspect a lot of users are in a similar situation. My old cheap phone has vastly more CPU and GPU power and about as much RAM as the computer that I did all of my work on as an undergrad - it's perfectly adequate for now.
Unfortunately, there are a lot of problems regulating dual-use technologies at any scale. We're seeing this at the small scale now, as terrorists learn that they can simply fill cars full of propane cylinders and improvise an explosive missile.
The real problem with a ban on autonomous weapons is that they're basically only useful to wealthy industrialised nations: i.e. the ones that can easily violate this kind of ban without fearing too much threat from sanctions. There's little need for, for example, North Korea to develop this kind of weapon because their military has a large number of expendable soldiers.
Slashdot Mirror
CHMOD? Check your privilege, shitlord.
uid == 0?
My partner just replaced her Nokia Lumia 1020. I really liked the UI and it never crashed, but it did have an issue where the sound subsystem would fall over and then nothing would play sound, which meant that she had to test the alarm before she went to sleep each night or it have died during the day and be silent in the morning. She replaced it with an Android device because of the complete lack of third-party app support on Windows phone.
At 1g acceleration, it takes around 11 seconds to travel half the distance in this tube (assuming that the other half is spent decelerating). That's a peak speed of 110m/s, or around 245 miles per hour, so this train had less horizontal acceleration than humans experience vertically just by being on this planet. Give them a comfy chair and they'll happily manage that level of acceleration for 30 seconds to a minute. And you can always trade a little bit time for comfort. Half the acceleration and accelerate for two minutes instead of one and you'll add two minutes (one at each end) to the total travel time, which won't make much difference in a half-hour journey.
There's a story from the '70s about an artillery control system that used neural networks to classify enemy targets and civilians. At the first live-fire demonstration, it immediately targeted and destroyed the general's car. It turned out that they'd trained it to recognise things seen in daylight as civilian and things seen at night as enemy vehicles. The project was cancelled. Something similar happened when Google's face tagging software learned that any dark-coloured face was a gorilla, because someone had tagged a gorilla in their photo and that was the only dark face in their training set. I was recently at a presentation by Nokia Bell Labs where they'd trained a neural network to classify urban scenes by beauty - looking at their results, it turned out that they'd built a complex system to count the number of trees in a scene, everything other architectural feature was ignored.
Machine learning systems are good for generating correlations. I often say that it's the technique that you use when you don't really understand the problem that you're trying to solve, but you have a lot of cheap compute to throw at it. You'll end up with a system that has detected some unknown correlation and has unknown error rates and unknown failure modes. But it's probably better than nothing...
Sometimes those are worth listening to. I read a few reviews of a wireless access point that said it was difficult to configure. I assumed that the reviewers were idiots. I have a PhD in computer science, so I was pretty sure I could figure it out. It turns out that the manufacturers had decided that technical terms were confusing and so they'd make up their own terms for every single thing in the configuration interface. It took hours of a combination of trial and error, poking the thing remotely with diagnostic tools, and reading forum posts by other people who had figured out what some of them meant, to work out how to translate from their nonsense wibble into standard networking terminology.
Mod this up. Most reviews are useless when comparison shopping because the people writing them have a sample size of one. I recently bought a new electric shaver. Most of the reviews are from people who have owned maybe one other shaver in their life. For a product with 600 reviews, 500 of them are left within a few hours of the new one arriving, so all that you really know from them is that it came in a box and didn't break in the first use. The only useful reviews were the ones where someone actually compared it to others that are still available.
Market value for the software, or a copy of the software? There are two parts involved in software development: writing the software and copying it. Writing it costs money, copying it is essentially free. Open source makes it harder to charge money for copying software, but makes it easy to charge money for writing it. In contrast, proprietary off-the-shelf software involves writing software for free and then charging to make copies of it. Which do you think makes more economic sense?
The problem with that comparison is that people refer to tasers as 'non-lethal' weapons, rather than their official designation of 'less-lethal'. If you pull a gun and shoot someone and they die, you're not surprised. If you pull a taser and shoot someone and they die, then you are. This means that people are a lot more willing to shoot someone with a taser.
Data centres produce a lot of jobs for sysadmin / SRE people, software developers, and so on. The problem for the state is that most of these jobs have no reason to be geographically close to the data centre. You need a few people to pull dead machines and rack replacements, and to maintain the power / cooling systems, everything else is remote.
I will happily create as many jobs as you want for $1m/job. I guarantee that the jobs will be around for at least 10 years. Even counting overheads, that's enough that I can pocket half the money and pay people to do nothing.
If you look through the "integrations" list on this page you can choose your favorite
I had a look through this list the first time that you posted it, but none of them seem applicable.
If you don't have any favorites, one of the Backblaze IT people here uses "Duplicity Linux" to do EXACTLY what you describe
Duplicity doesn't seem to be able to do this. It maintains its own change logs, which ZFS gives me for free. I can do zfs send and get a stream that contains the changes since the last snapshot. I want to send this stream, encrypted, to B2, named such that if I need to recover I can easily stream the snapshots back, in order, so that zfs receive can reapply them to the filesyystem. I also want to configure the ACLs on the remote backups so that the credentials that the machine that's being backed up has can only create new files, it cannot modify or delete old ones, so that if that machine is compromised it cannot affect old snapshots.
Sell Apple. You were foolish not to when he died
Steve Jobs died October 5, 2011. Just before that, AAPL was at $57.83/share, which is the lowest that it's been since then, aside from a small dip in the weeks immediately after his death. It's now at a bit over $159/share. That's around a 16% annual return on investment. What did you buy when you sold your AAPL shares that gave you a better return?
A fixed-wing aircraft uses something between 2 and 3 *times* as much energy per km.
As much energy as what? A ground vehicle? For the same weight or per passenger? Light aircraft are typically lighter than cars because they are less likely to be involved in collisions and so don't need roll cages and so on, so I'll assume that it's per passenger. Aircraft can go in a straight line, whereas ground vehicles often have to travel two sides of a triangle or three sides of a square, so even a factor of 2 is not that bad, if you can take off and land near you origin and destination.
Fine for many purposes. I have a laptop. Stuff that I'm actively working on is in a remote git repo. Everything is backed up to a NAS in my house. I'd like to do off-site backups of that. The only time that I'd need the recovery urgently is if my laptop and my NAS were both fried at the same time.
I back up computers in my house to a FreeBSD NAS and I'm looking for an off-site backup solution. I want to be able to send ZFS snapshots, encrypted, to a remote location. Do you know of any tools that work well for this kind of use, or do I need to roll my own.
Many modern smartphones have HDMI out and can be plugged into a projector or HD TV, where you could tell the difference. Of course, in most such situations, you'd probably prefer to use WiFi, but if mobile data is your only option then this could be annoying.
That definition of Network Neutrality is the one that's pushed by ISPs, not by NN advocates. Typical NN definitions allow differentiating based on traffic type, but with some tight constraints (e.g. you can put things into latency-sensitive, jitter-sensitive, and bandwidth-sensitive buckets, but you can't treat one latency-sensitive protocol differently from another). QoS explicitly is allowed by all except for the straw-man NN definition used by ISPs.
What about your sharp kitchen knives, do you have to enter an access code before using the blade or do you just go for it?
My sharp kitchen knives came in a knife block the prevents me from accidentally grabbing the blade. It has handles that protect my hand. If someone sold a knife that had the blade going the full length to where the user was expected to hold it, then they'd be unpopular. Most devices come with basic safety designs and are safe in their normal and intended mode of use.
Finding idiots doesn't make a trademark invalid.
It does if you find enough of them. Common usage doesn't mean usage by domain experts.
You're not going to write "I searched it on the Internet" because it's too long
No, I'll probably just write 'I searched', because the context of 'on the Internet' is implicit from the context.
The name Sellotape was derived from cellophane, at that time a trademarked name, with the "C" changed to "S" so that the new name could be trademarked
I'm not sure I totally agree with your classification, but it's far more common for me to meet a scientist with a solid knowledge of history, art, and literature than an arts or humanities person with a solid understanding of science. More importantly, scientists are more likely to be embarrassed about their lack of knowledge, whereas humanities scholars display almost pride in their lack of understanding of physics or computer science.
I came here to say the same thing. You can now buy Oreos in the UK, but only one flavour, and few people eat them as they're inferior to (and more expensive than) custard creams.
Maybe it has to do with it being too difficult, or perhaps even impossible, to upgrade to a newer version.
I only just upgraded from 5.1.1. I bought a Moto G because, at the time, Motorola was owned by Google and I expected good long-term support. It shipped with 4.3, got a quick upgrade to 4.4, eventually got an update to 5, and hasn't had security updates for well over a year. I reflashed it with LineageOS to get a new version (7.1), but that's well beyond the abilities of most users. The hardware is still completely adequate for all of the apps that I want to run, so I have no desire to replace it. I suspect a lot of users are in a similar situation. My old cheap phone has vastly more CPU and GPU power and about as much RAM as the computer that I did all of my work on as an undergrad - it's perfectly adequate for now.
The real problem with a ban on autonomous weapons is that they're basically only useful to wealthy industrialised nations: i.e. the ones that can easily violate this kind of ban without fearing too much threat from sanctions. There's little need for, for example, North Korea to develop this kind of weapon because their military has a large number of expendable soldiers.