A sensor in an ear tag or collar is not a substitute for a physical health check. The sensor can not tell if the cow has a rock stuck in its hoof for example, or if the cow has gastrointestinal problems. I worked in this field ten years ago building a multi-hop wireless network on cattle to electronically herd and track them. There are long standing patents on this technology in the US (USDA) and Australia (CSIRO). We decided to stop research after the US prison industry expressed interest.
This reads like a typical management wish list. You could add "All programmers writing AI systems must be ethical and empathic" but how are you going to enforce that? We can't make ordinary computers secure, the IoT is being built without security, AI will be no different. The security flaws already demonstrated in AI systems (making a Tesla drive into oncoming traffic, racial bias in AI recommendations for prison release) do not look easy to solve, in part because the only existing source of training data is human behavior, which will always have biases, and because the basis of AI decision making is obscured by the nature of the algorithms. AI systems would have to constantly be retrained to adjust to changes in society, which is expensive and difficult and does not increase profit. Security makes systems more difficult to use. Privacy is difficult to define or implement because it's relative to context and computers do not understand human context. The people who are tasked with following this list will instantly water it down to "best practices" which for computer security so far has left system security a few steps behind attackers. However, it's worthwhile to create such a list, if just to start to understand the potential damage AI is going to cause. Since all security and privacy comes down to human behavior, it would be better to try to build a better society which encourages good behavior and to solve inequities and social problems. If most people are relatively comfortable and secure in their lives, fewer of them will try to attack or misuse systems and they won't feel inclined to take advantage of flaws when they appear. Trying to automate ethics in systems built by people using todays technology seems like a lost cause. Start turning psychology and sociology into real sciences used to build a better society rather than using them to get more work and money out of people.
so there may be some attempt to limit it. Try to get a grant to explore something that is NOT an extension of an idea from some previous grant. On the other hand, it might just be that the old scientists now control the grants, so what gets explored is limited to what they find acceptable. Very little basic science seems to be done these days. Try getting a grant to just 'explore' gravity with no specific plan, just to see what you might find playing with some equipment and thinking about the results, something that experimentalists used to do all the time.
Experimentalists used to be self funded, either rich from some other enterprise, or funded by discoveries they had already made and sold or turned into businesses. So they were much freer to explore whatever they wanted to.
but I think there are loopholes. The NSA and CIA are tasked by Congress and give the reports to Congress. So the NSA/CIA can say they do no economic spying, but the congresspeople can whisper what they know to their donors and corporate friends. Or some economic spying can be "in the interest of national security". Leaks have indicated that the US spied on other countries as part of the lead up to trade conferences so the US could cheat: https://medium.com/economic-po...
Is that not spying for economic gain?
Lower your fridge temperature (but not too close to freezing) and add an ozone generator to the inside of your fridge and most everything in it will last longer. The ozone generator only needs to send out a pulse of ozone occasionally so it can run on batteries. There are a lot of them for sale on eBay. Seems like a better solution than modifying the vegetables.
Self driving cars, especially electric, are simpler mechanically, so automotive companies may not have as much of an advantage. Cars are going to get smaller and be electric (requires less resources, global warming, costs less); the current car companies still want to sell us $35,000+ multi-ton behemoths. Buy two small autonomous electric vehicles and have the second one tag along when you need extra space for transporting items or people; otherwise the second vehicle stays home (or it could be hired in the first place, rent one on demand when you need more cargo hauling capacity.) China and other countries are developing small, inexpensive electric vehicles; add some autonomous driving tech to them, and relax some of the safety requirements for lower speed vehicles, and a lot of the big US cars may disappear. A lot of the expertise in car design is in the companies supplying components (steering systems, braking systems) which are not owned by the car companies, and who will hire out to anyone with the cash.
As a side note, autonomous vehicles may mean people don't need a drivers license any more. That is going to shift the entire enforcement industry centered around cars and driving, and will likely encourage a lot of changes in vehicle and identity related laws. That may make it easier to get safety requirements relaxed for low speed autonomous electrics. So the current political advantages of the car companies may be moot as well.
Seems like any optical system will stop working the first time a bug, leaf, rain drop or snow hits it's lens and sticks. You can coat the lens to make it easier for the wind to blow them off, but in general the optics could be instantly obscured at any time, which basically means the car would have to stop immediately.
Making sure the clock is set accurately is a huge problem in itself. Maybe satellite internet connections could solve that, but radio connections are often intermittent and clocks drift when left to run on their own and can be disrupted by electrical events.
Some types of deep fake problems are only a problem for people whose faces are known and are no problem for everyone else. A fake video showing you proclaiming you are Jesus would elicit a "who cares" from most of the world. If you're well known though, it could be a problem. Other types of deep fakes need to get the details right to be convincing. The sun has to be at the right angle, the location needs to be right, the buildings need to be the right buildings, the flesh tones have to match, etc. See Hany Farid's work on detecting Photoshopping; the same can be done with video to detect deep fakes that combine different elements.
My guess is it is mainly people who have a known face and are widely known (e.g., politicians) who are the ones concerned about deep fakes, because even if the fake is detectable, that doesn't stop people from thinking (or wishing) it was true. In other words, it will be mostly used to make fun of the famous. Using it to pull off crimes or convince large populations that someone important has done something bad would require a very high bar be met by the fake. One simple solution for the famous might be to continuously track their location using a tamper resistant device so they can prove where they were. If the location track shows them moving, and the fake video shows them standing still...
The wheels on a scooter are too small to make the ride stable. And some scooters go too fast. Being able to go 25mph on little wheels is a recipe for disaster. Even small obstacles can cause your path to alter. Compare the rate of accidents of bicycles to the rate for scooters. Are the ER's reporting large numbers of biking accidents also?
Maybe the gyroscopic characteristics of larger wheels is what makes bicycles more stable? Maybe bikers tend to follow the rules of the road more than scooter riders do, as well, which could be solved with laws.
Cheap, small, electric, and eventually self driving also. After they become self driving if you need more cargo or passenger space you just rent a second or third one and it follows yours around. The car you rent might be your neighbors.
The cost of crime in the US huge. If UBI meant people would not need to steal to live, that might lift a large burden from society. Perhaps that could be one way to decide the amount of UBI, raise it until crime stops decreasing. I'm not sure what percentage the cost of theft by the poor is though. Some of it may be driven by the high cost of drugs (which has other solutions).
http://www.shopliftingpreventi...
"There are approximately 27 million shoplifters (or 1 in 11 people) in our nation today. More than 10 million people have been caught shoplifting in the last five years."
https://www.iii.org/fact-stati...
"in 2017, there were 16.7 million victims of identity fraud, a record high that followed a previous record the year before. Criminals are engaging in complex identity fraud schemes that are leaving record numbers of victims in their wake. The amount stolen hit $16.8 billion last year as 30 percent of U.S. consumers were notified of a data breach last year, an increase of 12 percent from 2016."
Say that crimes caused by poverty cost $50B/year in the US. That's $152/year per person. Not enough for UBI, but it could eventually offset part of the cost of UBI.
It's relatively easy to build your own facial surveillance tools using open source software and wouldn't cost much to deploy across a town or part of a city. Same thing for automated license plate readers. And walking gait biometric trackers. And heart rate biometric trackers using the small variations in face color caused by heart beats. And to add activity recognition trackers. And to add radio signal trackers. And it's easy to get people to install software that tracks their activities on phones and computers. These kinds of surveillance are not limited to just governments and big companies, _everyone_ with a few skills can do it. As soon as the CEO's and politicians start realizing this by finding their faces, and cars, and biometrics being tracked and their activities being broadcast to the world, then you'll start to see new laws. Not sure they'll do much good though, there is no obviously effective way to police this. How do you tell what the intent is behind a tiny camera and computer sitting in some random location? How do you even tell if it's a camera/computer/radio? It doesn't have to look like one. You could build your own car/face/biometric surveillance system using a few $100 of cheap cellphones. People are likely already doing this. Soon you'll be able to buy complete systems on eBay.
> When children play role-playing games, they aren't learning about real life.
So classify real life as a source of mental illness and start prescribing cures for it! Reduce competition and encourage learning until you get it right rather than awarding effort with failing grades, allow more flexible deadlines, no working late to get it done, teach politeness and charm, and much more. People avoid reality for a reason. Saying it is what is and you can't change it is stupid because reality is obviously infinitely malleable. Make reality fun!
For further training and without permission of the person called? To some extent the software has to know what the content of the call was also, which could also be stored, and sold and accessed by businesses and law enforcement.
The linked article is based on a NYTimes article which is much more informative:
https://www.nytimes.com/2018/0...
ALL US cellphone location data is being aggregated and resold.
In part edge computing is driven by advances in the speed and low power cost of MCU's, which just now are capable of running some types of machine learning. There are several advantages to local computing, see our paper:
NoCloud: Exploring Network Disconnection through On-Device Data Analysis
https://www.cs.dartmouth.edu/~...
So all I need to do to protect my privacy in the US is to VPN myself via the EU? Of course that means the NSA + GCHQ will definitely collect all my metadata. Do the NSA + GCHQ have to comply with the GDPR?
Given all the problems in the world that need solving, why does the world find little value in PhD's? Maybe one answer is to create more research organizations outside of universities which can receive government grants, similar to CSIRO in Australia? Universities are not good at problem solving research; researchers choose problems at will not based on value to society but on how easy it is to advance their career and can abandon a problem on a whim, grad students graduate and abandon their research, there is no engineering support for most projects (engineers cost too much and are difficult to add to grant proposals), Corporations hire PhD's but mostly not to solve the worlds problems, just to make money. So there are few long term research groups actually studying crucial problems in depth and the ones that do exist typically are understaffed and have few resources. The system seems designed to keep the world as it is. "This is 20th Century America and we're going to keep it that way." - The Man Who Fell To Earth
Good observation. Also if cancers have any differences caused by individual biology then the training this AI received was only on 238 samples, not 30,000 and most of its successful identification results were for multiple cancer samples from the same individual.
Computers that can deal with all the details that computers require instead of making people specific every last dot and slash. Small $6000 electric cars, every other country is building them but the US is only making big $25,000+ electric cars. Networks that are reliable, self healing, easy to configure. Our current networks are continually breaking, require constant human management, are difficult to connect to, and insecure.
As you add more nodes they get slower and slower as the packets have to jump more hops to get to their destination. To build a scalable ad-hoc mostly mesh network you need backbone nodes that do long haul transfer between small mesh clouds. This has been well researched by the US military and the wireless sensor network research community. Also battery power is an issue for mobile nodes, are you willing to dedicate half your battery to forwarding packets for other people while your phone is in your pocket? A mesh mode for phones that only forwards very short text messages in a disaster area may have some utility, however building a nation sized mesh-only network won't work. So figure out how communities could provide some fiber or microwave links between small mesh clouds (maybe use cars as more powerful mobile backbone nodes, and homes as fixed location backbone nodes?) and then you'll have a workable, scalable partially mobile partially ad-hoc network. Routing in such a network is another difficult problem though, if everyone is moving, the data that needs to be sent to inform every node how to get packets to every other node is massive. There are some possible solutions for that, though those solutions and the backbone links may result in giving up some decentralization and possibly be a privacy vulnerability. Creating a completely independent mobile internet/web that does not rely on the existing internet is not a simple problem. How to grow your network is a chicken and egg problem as well (until the full network exists who will use it?) Learn from how UUNET started using UUCP over telephone lines; you could use existing networks to support your new network until it can stand on its own:
https://en.wikipedia.org/wiki/...
Russia's fighters can be flown by remote control. With all the sensors and computers on board a fighter it would seem possible to turn it into an autonomous weapon. The first use of small, cheap autonomous weapons that can track specific individuals will certainly change the landscape. Big autonomous weapons either already exist or just require a software upgrade. If software is the only difference then verification of any treaty becomes almost impossible (maybe spies could detect field tests? Maybe not, just put a pilot/operator in place and have them do nothing or pretend to operate the system.) Without verification is there any other way to control these kinds of munitions? Maybe capturing failed devices that are still intact, reverse engineering them, putting the results on public display, and applying sanctions based on that? Who built the device could be anonymized, but it at least raises the bar of difficulty. Maybe simple methods of blinding sensors+cameras/radios could make autonomous weapons less useful and higher cost, deterring their use? In some sense this isn't a new problem, a weapon with a pilot is essentially an autonomous weapon run by a meat computer (e.g., a person with a gun). The main differences on the horizon are cheapness, smallness, being able to fly, and targeting improvements. Disrupt one or several of those factors and they become less effective or more costly to build.
Strong air jets strategically placed inside buildings could perhaps prevent drones from flying indoors without having much effect on people.
Wikipedia says "Cuba is located in an area with several active fault systems which produce on average about 2000 seismic events each year." One explanation for why only the diplomats were affected could be that native Cuban's have adapted via evolution to the effects of tremors (maybe native Cuban's think "Oh, an earth tremor, I should walk a few feet to the side so it doesn't hurt my head", though you'd think there would be stories about something like that). A natural cause seems like a possible explanation since the size of the equipment needed to cause this kind of damage would make it very difficult to hide (e.g., a large diesel engine at each location driving a massive transducer, though I suppose it could be disguised as a large truck.) . Low frequency sounds are not directional (without something like a 100 foot long horn to direct the sound), so setting up an interference pattern that attacked one location would require quite a few huge oscillators and there would still likely be resonance peaks in other locations and other people affected (although the oscillators might be easier to hide since they would not need to be right next to the target). Tuning the oscillators so you know you'll affect a specific target area would be difficult to do in secret. It's a stretch but a natural phenomenon so far seems like the simplest explanation. Installing some measurement gear in the affected locations would provide some facts for narrowing down possible causes; I'm surprised we haven't heard anything about measurement results given how long this has been going on. The US wouldn't need Cuba's approval to install some test equipment in an embassy. It might be easy to correlate the timing of attacks with seismographic data as well to test the possibility of a natural cause.
Slashdot Mirror
A sensor in an ear tag or collar is not a substitute for a physical health check. The sensor can not tell if the cow has a rock stuck in its hoof for example, or if the cow has gastrointestinal problems. I worked in this field ten years ago building a multi-hop wireless network on cattle to electronically herd and track them. There are long standing patents on this technology in the US (USDA) and Australia (CSIRO). We decided to stop research after the US prison industry expressed interest.
This reads like a typical management wish list. You could add "All programmers writing AI systems must be ethical and empathic" but how are you going to enforce that? We can't make ordinary computers secure, the IoT is being built without security, AI will be no different. The security flaws already demonstrated in AI systems (making a Tesla drive into oncoming traffic, racial bias in AI recommendations for prison release) do not look easy to solve, in part because the only existing source of training data is human behavior, which will always have biases, and because the basis of AI decision making is obscured by the nature of the algorithms. AI systems would have to constantly be retrained to adjust to changes in society, which is expensive and difficult and does not increase profit. Security makes systems more difficult to use. Privacy is difficult to define or implement because it's relative to context and computers do not understand human context. The people who are tasked with following this list will instantly water it down to "best practices" which for computer security so far has left system security a few steps behind attackers. However, it's worthwhile to create such a list, if just to start to understand the potential damage AI is going to cause. Since all security and privacy comes down to human behavior, it would be better to try to build a better society which encourages good behavior and to solve inequities and social problems. If most people are relatively comfortable and secure in their lives, fewer of them will try to attack or misuse systems and they won't feel inclined to take advantage of flaws when they appear. Trying to automate ethics in systems built by people using todays technology seems like a lost cause. Start turning psychology and sociology into real sciences used to build a better society rather than using them to get more work and money out of people.
Produces energy and gets rid of nasty byproducts.
so there may be some attempt to limit it. Try to get a grant to explore something that is NOT an extension of an idea from some previous grant. On the other hand, it might just be that the old scientists now control the grants, so what gets explored is limited to what they find acceptable. Very little basic science seems to be done these days. Try getting a grant to just 'explore' gravity with no specific plan, just to see what you might find playing with some equipment and thinking about the results, something that experimentalists used to do all the time. Experimentalists used to be self funded, either rich from some other enterprise, or funded by discoveries they had already made and sold or turned into businesses. So they were much freer to explore whatever they wanted to.
but I think there are loopholes. The NSA and CIA are tasked by Congress and give the reports to Congress. So the NSA/CIA can say they do no economic spying, but the congresspeople can whisper what they know to their donors and corporate friends. Or some economic spying can be "in the interest of national security". Leaks have indicated that the US spied on other countries as part of the lead up to trade conferences so the US could cheat:
https://medium.com/economic-po...
Is that not spying for economic gain?
Lower your fridge temperature (but not too close to freezing) and add an ozone generator to the inside of your fridge and most everything in it will last longer. The ozone generator only needs to send out a pulse of ozone occasionally so it can run on batteries. There are a lot of them for sale on eBay. Seems like a better solution than modifying the vegetables.
Self driving cars, especially electric, are simpler mechanically, so automotive companies may not have as much of an advantage. Cars are going to get smaller and be electric (requires less resources, global warming, costs less); the current car companies still want to sell us $35,000+ multi-ton behemoths. Buy two small autonomous electric vehicles and have the second one tag along when you need extra space for transporting items or people; otherwise the second vehicle stays home (or it could be hired in the first place, rent one on demand when you need more cargo hauling capacity.) China and other countries are developing small, inexpensive electric vehicles; add some autonomous driving tech to them, and relax some of the safety requirements for lower speed vehicles, and a lot of the big US cars may disappear. A lot of the expertise in car design is in the companies supplying components (steering systems, braking systems) which are not owned by the car companies, and who will hire out to anyone with the cash.
As a side note, autonomous vehicles may mean people don't need a drivers license any more. That is going to shift the entire enforcement industry centered around cars and driving, and will likely encourage a lot of changes in vehicle and identity related laws. That may make it easier to get safety requirements relaxed for low speed autonomous electrics. So the current political advantages of the car companies may be moot as well.
Seems like any optical system will stop working the first time a bug, leaf, rain drop or snow hits it's lens and sticks. You can coat the lens to make it easier for the wind to blow them off, but in general the optics could be instantly obscured at any time, which basically means the car would have to stop immediately.
Making sure the clock is set accurately is a huge problem in itself. Maybe satellite internet connections could solve that, but radio connections are often intermittent and clocks drift when left to run on their own and can be disrupted by electrical events.
Some types of deep fake problems are only a problem for people whose faces are known and are no problem for everyone else. A fake video showing you proclaiming you are Jesus would elicit a "who cares" from most of the world. If you're well known though, it could be a problem. Other types of deep fakes need to get the details right to be convincing. The sun has to be at the right angle, the location needs to be right, the buildings need to be the right buildings, the flesh tones have to match, etc. See Hany Farid's work on detecting Photoshopping; the same can be done with video to detect deep fakes that combine different elements.
My guess is it is mainly people who have a known face and are widely known (e.g., politicians) who are the ones concerned about deep fakes, because even if the fake is detectable, that doesn't stop people from thinking (or wishing) it was true. In other words, it will be mostly used to make fun of the famous. Using it to pull off crimes or convince large populations that someone important has done something bad would require a very high bar be met by the fake. One simple solution for the famous might be to continuously track their location using a tamper resistant device so they can prove where they were. If the location track shows them moving, and the fake video shows them standing still...
The wheels on a scooter are too small to make the ride stable. And some scooters go too fast. Being able to go 25mph on little wheels is a recipe for disaster. Even small obstacles can cause your path to alter. Compare the rate of accidents of bicycles to the rate for scooters. Are the ER's reporting large numbers of biking accidents also? Maybe the gyroscopic characteristics of larger wheels is what makes bicycles more stable? Maybe bikers tend to follow the rules of the road more than scooter riders do, as well, which could be solved with laws.
Cheap, small, electric, and eventually self driving also. After they become self driving if you need more cargo or passenger space you just rent a second or third one and it follows yours around. The car you rent might be your neighbors.
http://www.shopliftingpreventi...
"There are approximately 27 million shoplifters (or 1 in 11 people) in our nation today. More than 10 million people have been caught shoplifting in the last five years."
https://www.iii.org/fact-stati...
"in 2017, there were 16.7 million victims of identity fraud, a record high that followed a previous record the year before. Criminals are engaging in complex identity fraud schemes that are leaving record numbers of victims in their wake. The amount stolen hit $16.8 billion last year as 30 percent of U.S. consumers were notified of a data breach last year, an increase of 12 percent from 2016."
https://www.statista.com/stati...
https://www.census.gov/popcloc...
USA population is 328 million
Say that crimes caused by poverty cost $50B/year in the US. That's $152/year per person. Not enough for UBI, but it could eventually offset part of the cost of UBI.
It's relatively easy to build your own facial surveillance tools using open source software and wouldn't cost much to deploy across a town or part of a city. Same thing for automated license plate readers. And walking gait biometric trackers. And heart rate biometric trackers using the small variations in face color caused by heart beats. And to add activity recognition trackers. And to add radio signal trackers. And it's easy to get people to install software that tracks their activities on phones and computers. These kinds of surveillance are not limited to just governments and big companies, _everyone_ with a few skills can do it. As soon as the CEO's and politicians start realizing this by finding their faces, and cars, and biometrics being tracked and their activities being broadcast to the world, then you'll start to see new laws. Not sure they'll do much good though, there is no obviously effective way to police this. How do you tell what the intent is behind a tiny camera and computer sitting in some random location? How do you even tell if it's a camera/computer/radio? It doesn't have to look like one. You could build your own car/face/biometric surveillance system using a few $100 of cheap cellphones. People are likely already doing this. Soon you'll be able to buy complete systems on eBay.
So classify real life as a source of mental illness and start prescribing cures for it! Reduce competition and encourage learning until you get it right rather than awarding effort with failing grades, allow more flexible deadlines, no working late to get it done, teach politeness and charm, and much more. People avoid reality for a reason. Saying it is what is and you can't change it is stupid because reality is obviously infinitely malleable. Make reality fun!
For further training and without permission of the person called? To some extent the software has to know what the content of the call was also, which could also be stored, and sold and accessed by businesses and law enforcement.
The linked article is based on a NYTimes article which is much more informative: https://www.nytimes.com/2018/0... ALL US cellphone location data is being aggregated and resold.
In part edge computing is driven by advances in the speed and low power cost of MCU's, which just now are capable of running some types of machine learning. There are several advantages to local computing, see our paper: NoCloud: Exploring Network Disconnection through On-Device Data Analysis https://www.cs.dartmouth.edu/~...
So all I need to do to protect my privacy in the US is to VPN myself via the EU? Of course that means the NSA + GCHQ will definitely collect all my metadata. Do the NSA + GCHQ have to comply with the GDPR?
Given all the problems in the world that need solving, why does the world find little value in PhD's? Maybe one answer is to create more research organizations outside of universities which can receive government grants, similar to CSIRO in Australia? Universities are not good at problem solving research; researchers choose problems at will not based on value to society but on how easy it is to advance their career and can abandon a problem on a whim, grad students graduate and abandon their research, there is no engineering support for most projects (engineers cost too much and are difficult to add to grant proposals), Corporations hire PhD's but mostly not to solve the worlds problems, just to make money. So there are few long term research groups actually studying crucial problems in depth and the ones that do exist typically are understaffed and have few resources. The system seems designed to keep the world as it is. "This is 20th Century America and we're going to keep it that way." - The Man Who Fell To Earth
Good observation. Also if cancers have any differences caused by individual biology then the training this AI received was only on 238 samples, not 30,000 and most of its successful identification results were for multiple cancer samples from the same individual.
Computers that can deal with all the details that computers require instead of making people specific every last dot and slash. Small $6000 electric cars, every other country is building them but the US is only making big $25,000+ electric cars. Networks that are reliable, self healing, easy to configure. Our current networks are continually breaking, require constant human management, are difficult to connect to, and insecure.
As you add more nodes they get slower and slower as the packets have to jump more hops to get to their destination. To build a scalable ad-hoc mostly mesh network you need backbone nodes that do long haul transfer between small mesh clouds. This has been well researched by the US military and the wireless sensor network research community. Also battery power is an issue for mobile nodes, are you willing to dedicate half your battery to forwarding packets for other people while your phone is in your pocket? A mesh mode for phones that only forwards very short text messages in a disaster area may have some utility, however building a nation sized mesh-only network won't work. So figure out how communities could provide some fiber or microwave links between small mesh clouds (maybe use cars as more powerful mobile backbone nodes, and homes as fixed location backbone nodes?) and then you'll have a workable, scalable partially mobile partially ad-hoc network. Routing in such a network is another difficult problem though, if everyone is moving, the data that needs to be sent to inform every node how to get packets to every other node is massive. There are some possible solutions for that, though those solutions and the backbone links may result in giving up some decentralization and possibly be a privacy vulnerability. Creating a completely independent mobile internet/web that does not rely on the existing internet is not a simple problem. How to grow your network is a chicken and egg problem as well (until the full network exists who will use it?) Learn from how UUNET started using UUCP over telephone lines; you could use existing networks to support your new network until it can stand on its own: https://en.wikipedia.org/wiki/...
Seems like a better solution for almost all kinds of waste disposal: https://en.wikipedia.org/wiki/...
Russia's fighters can be flown by remote control. With all the sensors and computers on board a fighter it would seem possible to turn it into an autonomous weapon. The first use of small, cheap autonomous weapons that can track specific individuals will certainly change the landscape. Big autonomous weapons either already exist or just require a software upgrade. If software is the only difference then verification of any treaty becomes almost impossible (maybe spies could detect field tests? Maybe not, just put a pilot/operator in place and have them do nothing or pretend to operate the system.) Without verification is there any other way to control these kinds of munitions? Maybe capturing failed devices that are still intact, reverse engineering them, putting the results on public display, and applying sanctions based on that? Who built the device could be anonymized, but it at least raises the bar of difficulty. Maybe simple methods of blinding sensors+cameras/radios could make autonomous weapons less useful and higher cost, deterring their use? In some sense this isn't a new problem, a weapon with a pilot is essentially an autonomous weapon run by a meat computer (e.g., a person with a gun). The main differences on the horizon are cheapness, smallness, being able to fly, and targeting improvements. Disrupt one or several of those factors and they become less effective or more costly to build. Strong air jets strategically placed inside buildings could perhaps prevent drones from flying indoors without having much effect on people.
Wikipedia says "Cuba is located in an area with several active fault systems which produce on average about 2000 seismic events each year." One explanation for why only the diplomats were affected could be that native Cuban's have adapted via evolution to the effects of tremors (maybe native Cuban's think "Oh, an earth tremor, I should walk a few feet to the side so it doesn't hurt my head", though you'd think there would be stories about something like that). A natural cause seems like a possible explanation since the size of the equipment needed to cause this kind of damage would make it very difficult to hide (e.g., a large diesel engine at each location driving a massive transducer, though I suppose it could be disguised as a large truck.) . Low frequency sounds are not directional (without something like a 100 foot long horn to direct the sound), so setting up an interference pattern that attacked one location would require quite a few huge oscillators and there would still likely be resonance peaks in other locations and other people affected (although the oscillators might be easier to hide since they would not need to be right next to the target). Tuning the oscillators so you know you'll affect a specific target area would be difficult to do in secret. It's a stretch but a natural phenomenon so far seems like the simplest explanation. Installing some measurement gear in the affected locations would provide some facts for narrowing down possible causes; I'm surprised we haven't heard anything about measurement results given how long this has been going on. The US wouldn't need Cuba's approval to install some test equipment in an embassy. It might be easy to correlate the timing of attacks with seismographic data as well to test the possibility of a natural cause.