The US government is going to find someone to help them with AI object recognition and target assistance. If it's not a new tech titan, it will be an establish defense contractor. A better implementation is safer for everyone. Both accuracy and speed of response are important in weapons systems.
As the AI becomes less effective, the risk of bad outcomes increases: collateral damage, misidentified innocents, and missed opportunities on real targets.
While I firmly believing that automating kill authority is very dangerous, it is very beneficial---even morally necessary---to automate analysis if we can make faster and better decisions by doing so.
- American privacy rights are trampled, yet again - American cloud providers lose access to EU markets since we cannot provide the privacy protections they require
+ American law enforcement and surveillance agencies get their Christmas wish at last
My vote this November is going to whoever promises to repeal this, regardless of the D/R/I after their name.
Indication of what? Given the insane amount of money Microsoft currently makes from cloud services it really only makes sense.
I am interested in secure operating systems that respect user privacy and function properly offline. The integration of cloud services and telemetry are contrary to this desire, and, therefore, this decision is not a positive indicator of their future decisions for me.
Obviously the company is going to do whatever they believe will make money. That doesn't mean it's in the users' best interests.
Myerson was a solid dev leader. He was "acquired" when Microsoft bought his analytics startup. Now he is being replaced by an established app/cloud guy. I do not see this as a positive indication, but it is consistent with Nadella's overall direction for the company.
I have no idea what it was like to work with him personally, so maybe it's something like that. But if decision reflects the corporate direction, then the telemetry/data issues with Windows 10 will probably grow.
It is very easy to decide whether you should take a call from your boss outside of work hours. If the answer to those questions is no, then don't answer the phone:
1. Do you have a personal relationship with your boss outside of the workplace?
2. Are you on-call or part of a backup rotation in accordance with a written company policy?
3. Is there a policy applicable to your positions which provides compensation for work requested during off-duty hours?
You could have been watching a movie, swimming in a pool, or competing on the sports field. You could sleep with the phone off, and your battery could die at any time. All of these things are normal. The only reason to avoid these situations is when your job description and compensation impose restrictions.
You are wrong. More than merely plausible, it is widely known to be true.
This study on the reaction times of drivers by the Monash Accident Research Centre cites a commonly assumed value of 2.5 seconds and argues that a value of 3 seconds is more reasonable. This corroborates the original poster's assertion.
Note that the US has traditionally used a value of three seconds, so this research would encourage Canada and European countries to align with our standard.
You are focused on the mechanical reaction, apparently, which is the easiest to test. This is addressed by religionofpeas as the "muscle response delay" portion. If you were assessed by a test that had you watch for a signal and then press a button, this is exactly what they were measuring.
In an unexpected event like this, however, the attention shift and decision-making process take longer than the physical activation of muscles. In all likelihood, an attentive driver would not have been able to respond. (Assuming the 15-yard distance and one second of travel time are reasonably accurate.)
In either case, the autonomous driving system is seriously deficient. If the radar/lidar cannot detect a human being on an open road, then it is not ready for prime time.
You can't convince me that Apple, or any of the voice recognition players, are dedicating more processing power in their central servers on a per usage-basis than the mobile devices have natively inside them. Apple phones are a multi-gigahertz computing devices with more DSP power inside them than your PC has.
First of all, speech recognition doesn't run on standard DSPs at all. The platform probably uses the DSP's native noise reduction functionality to get a cleaner input, but modern speech recognition is based on deep learning, which is accelerated on GPUs or custom ASICs. Neither of those things was present in any phone when Apple and Google first shipped voice recognition. All those fancy DSPs for talking and media consumption do jack for speech recognition.
Second, they don't need an insane amount of compute power. The servers alleviate memory pressure and reduce storage demands on the device. Even a limited language model dataset is over a GB. Compare that to a small buffer, maybe a few KB, to tx/rx the voice stream and response.
Mainstream phones ship with 2-3 GB RAM and 16-32 GB storage, which is a serious issue for voice recognition. Budget devices have even less. There should be no issues running the Siri backend service on a decent workstation, but a phone is too cramped. I wouldn't be surprised if Siri's model data is much larger than a GB.
As DRAM and flash densities increase, the ability to run speech recognition software locally will improve. You could get away with much smaller datasets if you are willing to trade off for a smaller vocabulary or lower accuracy, but who would do that? People have complained for years about poor accuracy---and we have finally beaten that problem, for the most part.
what is the point of gig-e when all io is on 1 usb bus?
From the article:
"While the USB 2.0 connection to the application processor limits the available bandwidth, we still see roughly a threefold increase in throughput compared to Raspberry Pi 3B."
You'd have to generate a lot of IO to drop below the original 3B throughput.
GPUs can be virtualized as well. Plus, all of the big players are customizing hardware to accelerate their cloud workloads. That includes Microsoft, Google, and Facebook for sure, and probably Amazon as well---though Amazon has never said anything publicly.
If Microsoft wants to do NN training, they could easily add GPUs, custom ASICs, or FPGAs to support those loads. They have done ASICs and FPGAs before, so I'd be shocked if they aren't already gearing up for AI workloads. Since they control the entire stack from the UI down to the hardware, they have almost unlimited flexibility with Visual Studio in Azure.
Moore's Law relied on technological advances in the semiconductor industry to fuel its projected growth. There is no practical or predictable limit on such growth until you run into a wall dealing with fundamental physics.
Lass's Law relies on adoption of a technology by commercial, state, and private entities for its growth---of which there are a limited supply. We are most likely looking at the beginning of an S-curve and mistaking it for an exponential or geometric curve. It is quite conceivable that the market for these devices will be saturated in time.
Per the article, they will let developers train their AI in Azure and then import directly into applications. Training a neural network is exactly the kind of limited-duration, CPU-heavy activity that the cloud is designed for. Borrow a thousand CPUs to knock it out in short order and get on with your work.
And imagine if you wanted to train an algorithm with different inputs to see which method yields the best results in your application. You can burn through the training process in parallel in the cloud quickly, and then start building packages for testing immediately. You can iterate faster to fine tune things once you've picked the best baseline training. Without paying for an expensive AI "render farm" up front. The idea is promising, although the devil is always in the details.
And, obviously, any decent hardware-level support for AI would be great. The article only refers to the Azure integration though, so it appears the Slashdot headline is misleading.
By handing your computer over to Geek Squad you are consenting to a search.
First of all, you are consenting to have your computer repaired, not to have its contents searched and reviewed. There are privacy issues with technicians accessing user files at all---unless specifically requested as part of the service, of course. E.g., when the customer says, "make sure Outlook has all my mails in it when you're done", then it's OK. Otherwise, no.
Second, if Geek Squad is getting paid for looking through the drives and reporting to the FBI, they may be acting as agents of the government. This is a legal distinction that the courts will have to decide, and it may get some people out of prison time. There is a difference between "informant" and "agent", including the applicability of the 4th Amendment.
This is the kind of news I hate. The kind where you read it, and you realize everybody involved is being an asshole.
Either that, or you're copying files off the hard drive before wiping the OS... and you see some questionable filenames flash across the file copy dialog.
A long time ago, I worked at a place where an idiot got fired for something similar after running a file copy while on a Remote Assistance session with the help desk.
The government demands that its loads be run on separate servers from public customers. (This is somewhat well-founded in light of Rowhammer and more recent attacks.)
The government also has rather comprehensive accreditation requirements for any information system that wants to connect to a government network.
From those two requirements, the need for a separate government product is born. This has nothing to do with backdoors and everything to do with ritualistic IT security practices. Most of the bureaucrats understand nothing that they accredit, but the rules are generally based on real risks. So here we are.
If memcached is running somewhere on your backend, that's fine. E.g., a user hits a web page, so your web frontend talks to database and application servers over your intranet to generate a page for that user. Those servers are perfectly fine with unauthenticated memcached on a private LAN. It's not ideal from a security standpoint, but it's enough to prevent this type of attack.
Something is terribly wrong if memcached is responding directly to requests from internet clients. People can use it for reflected DDoS attacks or exfiltrate/contaminate your data. Anything exposed to the internet should have layered security---firewall, encryption, and authentication.
If your memcache infrastructure must be shared with a third party, configure appropriate cypto between your servers and theirs. And setup firewall rules to ensure that your servers are only talking to theirs. Depending on how much of the circuit you control, you can secure outside communications with VPLS, VPN, or IPsec. Those are all widely supported and well-understood standards, and ignorance is no exuse for a system administrator.
Because they're already good at programming, they don't need to be trained further. If you have programmers who are extremely proficient in C/C++/Ruby/Python/etc but have minimal social skills and no experience in project management, then you train them on communication and collaboration to make them more effective.
An experienced programmer can become proficient in another language on his own, for the most part, if he has already learned one or two. Training is not really necessary. Honestly, you're better off giving an experienced hire some time to become familiar with the application design and codebase anyway.
Entry-level hires will usually be chosen based on knowledge of the project language, so there is no need to train them.
So, realistically, I wouldn't expect to see someone trained on a new language. They will generally pick that up on their own as needed. On the the other hand, people rarely improve their communication skills without explicit direction---or a lot of hard life experience. I can easily it being worthwhile to pay for that training instead of waiting for that "life experience" to happen within your organization.
You think Nokia isn't doing anything but playing on nostalgia?
Nokia produces solid smartphones---with vanilla Android and minimum two-year updates. The phones are otherwise comparable to the competition, and this is exactly the distinction I care about. I plan on buying a Nokia when my current phone is outdated.
Their lineup has been out for roughly a year, and a new generation is in the works. Unfortunately, they launched in Europe first with the current lineup, so the next gen may not be available stateside immediately.
Imagine: You're a physicist... In a society where invention and discovery are written out of the script, progress and improvement would suffer.
Who the hell is going to understand this alien "gift"? Who is going to figure out how these new equations relate to the physical world? Who is going to design materials and tools to use this new knowledge? Who is going to develop the next set of theories once we find the limits of this new science?
We might get answers to a lot of our outstanding questions, but we will not end up with fewer questions.
Explain again why they needed power-wall batteries for each installation instead of just using the solar power directly when it was needed the most
Electricity is relatively expensive in Hawaii. If they are buying panels anyway to offset AC consumption, it could easily make economic sense to buy batteries. The batteries allow them to continue using those panels even when the kids aren't in class.
And I don't buy the argument that they could just feed the excess power into the grid. Where I live, the utility company pays jack compared to what they charge. I double the value of my generation if I store/use instead of sell/buy. This makes batteries a very good economic decision---in my area, at least.
In Hawaii, it looks like there is a feed-in tariff. So going with batteries, you eliminate that right off the bat. I couldn't find published rates, but if Hawaiian Electric works like my utility, those batteries will pay for themselves well before their expiration.
Cloning is not an issue if the signed data includes physical descriptors and photographs. Ultimately, all government ID systems rely on a human matching the person in front of them to the person on the paperwork.
Preventing forgery is the major concern. And they have zero chance of stopping it if they cannot verify a fucking digital signature. Pathetic.
Hell, ADOBE has integrated support for digital signatures and document validation---and it actually works. Unless there was a proposal to fix this that couldn't get Congressional funding, someone needs to be shitcanned. This is a serious lapse, and they should have asked for money to buy the hardware/software years ago.
Once again, we get to hear about risks to national security. Laughable ones, at that.
You have to assume that every endpoint on your network can be compromised. If your network security model cannot cope with widespread host infection, then your security is garbage. If they really cared about security, their networks should already have mitigations for Meltdown/Spectre-class malware in place.
Meltdown and Spectre aren't the first exploits either. They should have a plan for unexpected malware. There is no reason to assume that a given exploit will be discovered by a responsible actor. Quite frankly, the US government should be happy they were notified at all. Black hat, Chinese, and Russian hackers sure as hell aren't disclosing their exploits.
Instead of "raising concerns", these officials should double down on hardening their networks properly. This is useless showmanship.
except for amateur radio, but that's too easy to track and shut down also.
Every telecom system ends with either a wire or a transmitter. The authorities can always track down the address/customer associated with the wire or triangulate the transmitter, so shutdown is always possible.
The authorities must license broadcast bands and enforce transmission regulations---at least, in the US. And since wires generally require pole or street access, there isn't any realistic scenario where private citizens run their own lines. Network connectivity will remain centralized, trackable, and identifiable for the foreseeable future.
There are a growing number of office jobs where a total of 2-3 weeks is the norm, and that includes vacation, personal, and sick time. I don't have one of those jobs because I can get better offers easily, but there will always people who can only get the baseline.
Two weeks per year (so, ten paid days) is too little---especially for people with children. There should be at least 5-10 days for personal illness or caring for a close family member. Discipline or fire people for misusing that time if necessary, but it needs to be there.
Plus, if sick people stay home like they should for their own health, they're less likely to infect their colleagues. This goes double for people with the disease factories known as children.
Slashdot Mirror
The US government is going to find someone to help them with AI object recognition and target assistance. If it's not a new tech titan, it will be an establish defense contractor. A better implementation is safer for everyone. Both accuracy and speed of response are important in weapons systems.
As the AI becomes less effective, the risk of bad outcomes increases: collateral damage, misidentified innocents, and missed opportunities on real targets.
While I firmly believing that automating kill authority is very dangerous, it is very beneficial---even morally necessary---to automate analysis if we can make faster and better decisions by doing so.
So, we have two losers and one winner here:
- American privacy rights are trampled, yet again
- American cloud providers lose access to EU markets since we cannot provide the privacy protections they require
+ American law enforcement and surveillance agencies get their Christmas wish at last
My vote this November is going to whoever promises to repeal this, regardless of the D/R/I after their name.
Indication of what? Given the insane amount of money Microsoft currently makes from cloud services it really only makes sense.
I am interested in secure operating systems that respect user privacy and function properly offline. The integration of cloud services and telemetry are contrary to this desire, and, therefore, this decision is not a positive indicator of their future decisions for me.
Obviously the company is going to do whatever they believe will make money. That doesn't mean it's in the users' best interests.
Myerson was a solid dev leader. He was "acquired" when Microsoft bought his analytics startup. Now he is being replaced by an established app/cloud guy. I do not see this as a positive indication, but it is consistent with Nadella's overall direction for the company.
I have no idea what it was like to work with him personally, so maybe it's something like that. But if decision reflects the corporate direction, then the telemetry/data issues with Windows 10 will probably grow.
It is very easy to decide whether you should take a call from your boss outside of work hours. If the answer to those questions is no, then don't answer the phone:
1. Do you have a personal relationship with your boss outside of the workplace?
2. Are you on-call or part of a backup rotation in accordance with a written company policy?
3. Is there a policy applicable to your positions which provides compensation for work requested during off-duty hours?
You could have been watching a movie, swimming in a pool, or competing on the sports field. You could sleep with the phone off, and your battery could die at any time. All of these things are normal. The only reason to avoid these situations is when your job description and compensation impose restrictions.
Sorry,
that is not remotely plausible.
You are wrong. More than merely plausible, it is widely known to be true.
This study on the reaction times of drivers by the Monash Accident Research Centre cites a commonly assumed value of 2.5 seconds and argues that a value of 3 seconds is more reasonable. This corroborates the original poster's assertion.
Note that the US has traditionally used a value of three seconds, so this research would encourage Canada and European countries to align with our standard.
There are two measures of reaction time.
You are focused on the mechanical reaction, apparently, which is the easiest to test. This is addressed by religionofpeas as the "muscle response delay" portion. If you were assessed by a test that had you watch for a signal and then press a button, this is exactly what they were measuring.
In an unexpected event like this, however, the attention shift and decision-making process take longer than the physical activation of muscles. In all likelihood, an attentive driver would not have been able to respond. (Assuming the 15-yard distance and one second of travel time are reasonably accurate.)
In either case, the autonomous driving system is seriously deficient. If the radar/lidar cannot detect a human being on an open road, then it is not ready for prime time.
That is an idiotic rant.
Governments and universities aren't paid to fact check Youtube videos. You want to pay taxes for that? Didn't think so.
You can't convince me that Apple, or any of the voice recognition players, are dedicating more processing power in their central servers on a per usage-basis than the mobile devices have natively inside them. Apple phones are a multi-gigahertz computing devices with more DSP power inside them than your PC has.
First of all, speech recognition doesn't run on standard DSPs at all. The platform probably uses the DSP's native noise reduction functionality to get a cleaner input, but modern speech recognition is based on deep learning, which is accelerated on GPUs or custom ASICs. Neither of those things was present in any phone when Apple and Google first shipped voice recognition. All those fancy DSPs for talking and media consumption do jack for speech recognition.
Second, they don't need an insane amount of compute power. The servers alleviate memory pressure and reduce storage demands on the device. Even a limited language model dataset is over a GB. Compare that to a small buffer, maybe a few KB, to tx/rx the voice stream and response.
Mainstream phones ship with 2-3 GB RAM and 16-32 GB storage, which is a serious issue for voice recognition. Budget devices have even less. There should be no issues running the Siri backend service on a decent workstation, but a phone is too cramped. I wouldn't be surprised if Siri's model data is much larger than a GB.
As DRAM and flash densities increase, the ability to run speech recognition software locally will improve. You could get away with much smaller datasets if you are willing to trade off for a smaller vocabulary or lower accuracy, but who would do that? People have complained for years about poor accuracy---and we have finally beaten that problem, for the most part.
what is the point of gig-e when all io is on 1 usb bus?
From the article:
"While the USB 2.0 connection to the application processor limits the available bandwidth, we still see roughly a threefold increase in throughput compared to Raspberry Pi 3B."
You'd have to generate a lot of IO to drop below the original 3B throughput.
GPUs can be virtualized as well. Plus, all of the big players are customizing hardware to accelerate their cloud workloads. That includes Microsoft, Google, and Facebook for sure, and probably Amazon as well---though Amazon has never said anything publicly.
If Microsoft wants to do NN training, they could easily add GPUs, custom ASICs, or FPGAs to support those loads. They have done ASICs and FPGAs before, so I'd be shocked if they aren't already gearing up for AI workloads. Since they control the entire stack from the UI down to the hardware, they have almost unlimited flexibility with Visual Studio in Azure.
Moore's Law relied on technological advances in the semiconductor industry to fuel its projected growth. There is no practical or predictable limit on such growth until you run into a wall dealing with fundamental physics.
Lass's Law relies on adoption of a technology by commercial, state, and private entities for its growth---of which there are a limited supply. We are most likely looking at the beginning of an S-curve and mistaking it for an exponential or geometric curve. It is quite conceivable that the market for these devices will be saturated in time.
Per the article, they will let developers train their AI in Azure and then import directly into applications. Training a neural network is exactly the kind of limited-duration, CPU-heavy activity that the cloud is designed for. Borrow a thousand CPUs to knock it out in short order and get on with your work.
And imagine if you wanted to train an algorithm with different inputs to see which method yields the best results in your application. You can burn through the training process in parallel in the cloud quickly, and then start building packages for testing immediately. You can iterate faster to fine tune things once you've picked the best baseline training. Without paying for an expensive AI "render farm" up front. The idea is promising, although the devil is always in the details.
And, obviously, any decent hardware-level support for AI would be great. The article only refers to the Azure integration though, so it appears the Slashdot headline is misleading.
By handing your computer over to Geek Squad you are consenting to a search.
First of all, you are consenting to have your computer repaired, not to have its contents searched and reviewed. There are privacy issues with technicians accessing user files at all---unless specifically requested as part of the service, of course. E.g., when the customer says, "make sure Outlook has all my mails in it when you're done", then it's OK. Otherwise, no.
Second, if Geek Squad is getting paid for looking through the drives and reporting to the FBI, they may be acting as agents of the government. This is a legal distinction that the courts will have to decide, and it may get some people out of prison time. There is a difference between "informant" and "agent", including the applicability of the 4th Amendment.
This is the kind of news I hate. The kind where you read it, and you realize everybody involved is being an asshole.
Either that, or you're copying files off the hard drive before wiping the OS... and you see some questionable filenames flash across the file copy dialog.
A long time ago, I worked at a place where an idiot got fired for something similar after running a file copy while on a Remote Assistance session with the help desk.
The government demands that its loads be run on separate servers from public customers. (This is somewhat well-founded in light of Rowhammer and more recent attacks.)
The government also has rather comprehensive accreditation requirements for any information system that wants to connect to a government network.
From those two requirements, the need for a separate government product is born. This has nothing to do with backdoors and everything to do with ritualistic IT security practices. Most of the bureaucrats understand nothing that they accredit, but the rules are generally based on real risks. So here we are.
It's really that simple.
It depends on where it's exposed.
If memcached is running somewhere on your backend, that's fine. E.g., a user hits a web page, so your web frontend talks to database and application servers over your intranet to generate a page for that user. Those servers are perfectly fine with unauthenticated memcached on a private LAN. It's not ideal from a security standpoint, but it's enough to prevent this type of attack.
Something is terribly wrong if memcached is responding directly to requests from internet clients. People can use it for reflected DDoS attacks or exfiltrate/contaminate your data. Anything exposed to the internet should have layered security---firewall, encryption, and authentication.
If your memcache infrastructure must be shared with a third party, configure appropriate cypto between your servers and theirs. And setup firewall rules to ensure that your servers are only talking to theirs. Depending on how much of the circuit you control, you can secure outside communications with VPLS, VPN, or IPsec. Those are all widely supported and well-understood standards, and ignorance is no exuse for a system administrator.
Because they're already good at programming, they don't need to be trained further. If you have programmers who are extremely proficient in C/C++/Ruby/Python/etc but have minimal social skills and no experience in project management, then you train them on communication and collaboration to make them more effective.
An experienced programmer can become proficient in another language on his own, for the most part, if he has already learned one or two. Training is not really necessary. Honestly, you're better off giving an experienced hire some time to become familiar with the application design and codebase anyway.
Entry-level hires will usually be chosen based on knowledge of the project language, so there is no need to train them.
So, realistically, I wouldn't expect to see someone trained on a new language. They will generally pick that up on their own as needed. On the the other hand, people rarely improve their communication skills without explicit direction---or a lot of hard life experience. I can easily it being worthwhile to pay for that training instead of waiting for that "life experience" to happen within your organization.
You think Nokia isn't doing anything but playing on nostalgia?
Nokia produces solid smartphones---with vanilla Android and minimum two-year updates. The phones are otherwise comparable to the competition, and this is exactly the distinction I care about. I plan on buying a Nokia when my current phone is outdated.
Their lineup has been out for roughly a year, and a new generation is in the works. Unfortunately, they launched in Europe first with the current lineup, so the next gen may not be available stateside immediately.
Imagine: You're a physicist... In a society where invention and discovery are written out of the script, progress and improvement would suffer.
Who the hell is going to understand this alien "gift"? Who is going to figure out how these new equations relate to the physical world? Who is going to design materials and tools to use this new knowledge? Who is going to develop the next set of theories once we find the limits of this new science?
We might get answers to a lot of our outstanding questions, but we will not end up with fewer questions.
Explain again why they needed power-wall batteries for each installation instead of just using the solar power directly when it was needed the most
Electricity is relatively expensive in Hawaii. If they are buying panels anyway to offset AC consumption, it could easily make economic sense to buy batteries. The batteries allow them to continue using those panels even when the kids aren't in class.
And I don't buy the argument that they could just feed the excess power into the grid. Where I live, the utility company pays jack compared to what they charge. I double the value of my generation if I store/use instead of sell/buy. This makes batteries a very good economic decision---in my area, at least.
In Hawaii, it looks like there is a feed-in tariff. So going with batteries, you eliminate that right off the bat. I couldn't find published rates, but if Hawaiian Electric works like my utility, those batteries will pay for themselves well before their expiration.
Cloning is not an issue if the signed data includes physical descriptors and photographs. Ultimately, all government ID systems rely on a human matching the person in front of them to the person on the paperwork.
Preventing forgery is the major concern. And they have zero chance of stopping it if they cannot verify a fucking digital signature. Pathetic.
Hell, ADOBE has integrated support for digital signatures and document validation---and it actually works. Unless there was a proposal to fix this that couldn't get Congressional funding, someone needs to be shitcanned. This is a serious lapse, and they should have asked for money to buy the hardware/software years ago.
Once again, we get to hear about risks to national security. Laughable ones, at that.
You have to assume that every endpoint on your network can be compromised. If your network security model cannot cope with widespread host infection, then your security is garbage. If they really cared about security, their networks should already have mitigations for Meltdown/Spectre-class malware in place.
Meltdown and Spectre aren't the first exploits either. They should have a plan for unexpected malware. There is no reason to assume that a given exploit will be discovered by a responsible actor. Quite frankly, the US government should be happy they were notified at all. Black hat, Chinese, and Russian hackers sure as hell aren't disclosing their exploits.
Instead of "raising concerns", these officials should double down on hardening their networks properly. This is useless showmanship.
except for amateur radio, but that's too easy to track and shut down also.
Every telecom system ends with either a wire or a transmitter. The authorities can always track down the address/customer associated with the wire or triangulate the transmitter, so shutdown is always possible.
The authorities must license broadcast bands and enforce transmission regulations---at least, in the US. And since wires generally require pole or street access, there isn't any realistic scenario where private citizens run their own lines. Network connectivity will remain centralized, trackable, and identifiable for the foreseeable future.
There are a growing number of office jobs where a total of 2-3 weeks is the norm, and that includes vacation, personal, and sick time. I don't have one of those jobs because I can get better offers easily, but there will always people who can only get the baseline.
Two weeks per year (so, ten paid days) is too little---especially for people with children. There should be at least 5-10 days for personal illness or caring for a close family member. Discipline or fire people for misusing that time if necessary, but it needs to be there.
Plus, if sick people stay home like they should for their own health, they're less likely to infect their colleagues. This goes double for people with the disease factories known as children.