If you ever tune in to the Bigfoot/Sasquatch enthusiast groups you will find that there are tons of sightings, and lots and lots of physical evidence, hair, coprolites, stuff that would be loaded with Bigfoot DNA. It it was real.
Despite the fact that anyone can run a DNA test that identifies species for under $100 today, and that finding a single sample of DNA from an unknown primate would make the person finding it famous forever, not a single specimen claimed to be from a Bigfoot has ever been found to have anything but a known animal (usually human, sometimes bear, or other known mammal) as its source.
True story, a friend of mine told the pain clinic she was using a THC creme on her knee to help with the pain and that it was working great, the pain clinic said that by self treating they could no longer prescribe her pain meds. So now she can't get legal meds and is relegated to only the 'illegal' ones.
Not in Washington State, Oregon, Colorado, and a growing (heh, growing...) list of others...
You are profoundly mistaken. Opioid painkillers are regulated directly by the Federal government, and state laws are irrelevant. They piss-test opioid patients in California, the home of medical cannabis, and where recreational pot is also now legal, and if they detect cannabis residue they cut off access to the opioid medication.
I calculated once how much light was put out in London during the gas light era. It turns out the total number of lumens emitted to light the largest and most brightly lit city in the world at the time (four million people) is equal to the amount of light used to illuminate one football field for a night game.
Is the park actually safer or do you just think it is because it's well-lit?
This is a very important point. People assume that lights provide security, but the evidence for this is at best very weak. Most studies that have been done (research on this is limited) have found lighting increases the crime rate!
And so you see completely empty areas that are closed and fenced off at night brightly lit for "security", and remote service buildings, water towers, etc. with lights on them for no benefit whatsoever. It is similar to car alarms. If you hear a car alarm go off do you rush out to frustate the possible car thief? No one does this. Everyone just ignores it. Similarly unless there is someone actually watching the light deters no one, it just helps the criminals case the joint and spot potential victims to attack.
Everyone wants "security" so bright lights go up everywhere.
Simply redesigning them (and other outside illumination lighting) would be a huge gain. A bit less than 10% of light hitting the ground in an urban area gets scattered back up into the sky. So if the lights are designed to shine light on the ground only (no light being sent horizontally, where it just creates glare, or worse directly into the sky) there would be a large reduction in light pollution.
LEDs can help a lot here since they are inherently directional, it takes effort to make them spew light in other directions. But light makers are willing to provide them to the market since people want to buy lights that resemble lights they are used to.
Similarly regulating sign lighting so that you don't have bright lights at the bottom of a sign pointing straight-up, and regulating the use of light as a form of advertising and promotion. A lot of commercial light use is abusive, using brilliant light-emitting signage throwing light everywhere.
On highways with light traffic late at night "smart" lights can be implemented that turn off when there are no cars for them to aid.
The problem is that it isn't just (or primarily) with ad hoc and custom analyses.
It is that regular business functions are run with these sheets all the time. Business types do this because the are familiar with the tool, and they can implement the process themselves without calling in a dev team.
And all of this is perfectly understandable. Would you call in an outside dev team, explain requirements, and then have to wait for an acceptable product to be produced, when you could do it yourself quickly?
This is inevitable unless considerable effort is expended by the organization to identify and pull these business functions into formal, administered, monitored systems.
A decree not to use Excel at all (if this is what it is) is stupid.
The emphasis should be on educating the business on treating these "normal function" spreadsheets as prototypes of the function that must be implemented formally going forward, and the necessary resources must be provided to make this happen, and suitable reward structures must exist to encourage businessmen to identify and bring forward these functions for proper automation. Without all of this this decree will be useless.
The mention of 106Ru without any mention of the other more prominent Radionuclides leads me to the conclusion that whatever bad happened was very recent, and not six decades back. That would be some 59 Half-Lifes....
Mod this guy up, everything he says is exactly right (and he provides a very informative link).
This is a release from an active plutonium processing plant which is handling relatively fresh fuel (less than a decade old, perhaps very fresh).
The AID virus is not "finding a way" to defeat three drug therapy. Three different protease inhibitors target the critical protease molecule in different ways such that it cannot mutate to defeat all three and still work.
This work is not the final stage, it is simply showing a two step mechanism that requires two gene segments be lost in a particular order for the kill switch to be deactivated. More complex schemes are easily imagined, two different two component kill switches; mutually lethal two component kill switches so that they both must be lost together in the same mutation process, etc.
Seriously, has no one read TFA? (I know, I know, this is/.)
This work addresses exactly that problem, they have installed two mechanisms in different parts of the genome both of which produce different toxins. The regular kill switch, which kills the organism with toxin A when an environmental trigger is detected, also continuously produces an anti-toxin for toxin B, which the other mechanism continuously produces. Losing the trigger part also edits out antitoxin B, killing the cell immediately. A two step process of losing the toxin B gene, followed by losing the trigger would be needed to eliminate the kill switch.
If that is not secure enough then you could make a three part system following a similar strategy.
This is reminiscent of how they developed an effective treatment for AIDs that renders it non-fatal. Any drug they came up with that targeted a vulnerable site in the virus would fail since that vulnerable site would be lost in the the very rapid mutation process of the virus. Even using two drugs that targeted different sites had limited success. But a three drug therapy has worked - targeting different regions of the protease so that it cannot simultaneously mutate enough to deactivate all of them, without also deactivating its function.
A good directed fragmentation payload could get a reliable kill out to 25 meters (just think of it as one quarter segment of a Claymore mine), which is much more than "a couple of meters", a drone flying straight in and detonating at 25 meters might tax Secret Service perimeter protection to get the President behind cover before the kill shot was made. And only he and the VP have this level of protection. Anybody else in the U.S. would be far more vulnerable.
And sure booby trapping a car and getting within physical attack range works (if get close enough to throw the grenade accurately, why not just shoot him?) but VIPs with body guards or security details make these things hard, and attacking the target yourself is generally one-time thing, you never get another chance as you will end up in custody (and possibly deceased). With drones you can attack from long range and not get caught. Big, big difference.
A human piloted drone would be a better single-target assassination weapon (as the U.S. drone program proves regularly). The Slaughterbots video emphasizes AI and facial recognition because using those technologies you can have autonomous attacks at large scale. Even in the case of attacking a U.S. President a swarm attack of drones that simply recognize heads (no need to find recognize faces) will permit a small swarm to indiscriminantly kill everyone in the vicinity of the President, including him or her.
Also the video is absolutely correct if you can get 3 grams of HE in close contact with a human head, by having the drone simply fly into the target, it will kill the target. I cited the use of fragmentation devices simply to show that even actual contact is not needed so that impenetrable perimeter defenses have to be deployed quite a ways away.
If your knowledge of explosives is limited to fireworks, you don't know anything about the subject. Don't you think it odd that no military employs fireworks as weapons?
A drone/EFP system designed for assassination (aiming and attacking a specific target) could use an EFP weighing only 10 grams to fire a 3.6 gram projectile (same weight as a typical M-16 round) at 2000 m/sec (twice the velocity and four times the kinetic energy of the M-16 round at the muzzle). This actually a much more powerful projectile than what you need to kill someone. EFPs are quite easy to make, there are "energetic materials" hobbyists who post their "experiments" online. Large EFPs made with simple tools were a huge problem to U.S. forces in Iraq.
Not happy with taking only a single shot? Well then multiple fragmentation weapons are just the thing. You can use an array of EFPs like that described above to send a cluster of them at your target. Or consider the venerable Claymore mine that weighs 1600 grams and fires 700 projectiles at 1300 m/sec. At 25 meters these are 100% lethal in a zone 25 meters wide. For a drone you would probably make a scaled down version, even a narrow cone "micro" version to send a cluster of high velocity projectiles into your target.
Simply replicating the M67 fragmentation grenade which weighs 400 grams and is 100% lethal at 5 meters in all directions would work fine.
With a suitably designed fragmentation device, if you can get a 100 gram payload within 5 meters of a target you can get a kill with 100% certainty without even requiring sophisticated aiming or munition design. With more effort you can do the same at greater range, or with even lighter payloads.
You are profoundly ignorant on the subject, and yet you feel obligated to make pronouncements about it. You must be a/. regular!
Really, if you are unaware - as you clearly are - that it is the rotation of galaxies, like the nearly Andromeda, and our own, that were the first evidence of the existence of dark matter then you know nothing about the subject.
Try reading the Wikipedia page which lists ten different pieces of evidence showing its existence, with links to further material on each.
Because what are the options here now? Windows 10 and MacOS. Spy ware and a fortune for closed hardware. Two corporations who, in different ways, think you do not own your own computer.
That is why some of us use Linux on the desktop and wish that making that painless would interest someone in Linux-land.
I also wish the British Museum would return all those things to the countries the British stole them from.
First crash it's done forever, lawsuits will bankrupt the company, and it will set flying taxis back by 50 years. Blade Runner I will never happen. Fuck these morons for trying to move too quickly.
These don't crash so easily. They have multiple redundancy in the rotor array, the craft itself is self-flying - the "pilot" just guides it where to go, and it does have that aircraft safety parachute if everything fails to bring the craft to the ground in a not-catastrophic manner.
The technology is apparently here to support this vehicle.
Which costs $350,000 and can carry 160 kg a maximum distance of 27 km at 50 km/h (that is 350 lb, 17 miles, and 42 MPH to those using pre-1795 measuring systems) and takes 2 hours to fully charge.
With a charge time:flight time ratio of 2:1 you could get in 8 hours of flight a day, around the clock, or about 400 km total travel distance. If used as a short range shuttle over a congested city center you might get in up to forty 10 km trips a day, maybe ~10,000 in a year. Looks like this could be a profitable service with a not-astronomical fare price.
There is however an (optional) full aircraft parachute that brings the entire vehicle to the ground. This is called a "ballistic total separation system".
I am puzzled though about what gets "separated" here. Ejecting any part of the vehicle is going to be a problem for people on the ground.
In addition there is the problem of those who people who actually lost their jobs from "creative destruction". Even a short period of "no job" is a very serious problem for nearly everybody, and even if new jobs come along "soon", say in a few years, those few years can be devastating. And even if new jobs do come along, is the person who lost work able to get any of those replacements?
The degree to which society can, or should, provide support for people whose jobs are eliminated by automation is a matter of serious debate. Most advanced economies (e.g. Germany) takes this seriously has well-supported programs in place and yet does very well economically and competitively. But Kurzweil dismisses this as even being a problem at all that needs to be addressed in any fashion.
That doesn't mean they came from elsewhere, it would also sync up really well with them coming from the blackhole in the center of our galaxy and being curved back inward by the gravity of the whole galaxy - sort of a galactic-scale particle accelerator. (Like the field lines of a magnet.)
Or instead of "really well", not all in any way, shape, or form. You have no idea of the physics involved, typing words is not a physical analysis.
The galactic escape velocity for an iron atom (a typical heavy cosmic ray particle) is about 10^5 eV. All of the cosmic rays under discussion have energies greater than 10^19 eV, or 100 trillion times more energetic than the galactic escape velocity energy.
The galactic magnetic field is much better at holding on to cosmic rays, but cannot confine them above an energy of 10^18 eV or so. Which is why the researchers are studying extragalactic cosmic rays with energies above 10^19 eV. They know these cannot be confined to the Milky Way.
If hundreds of professional astrophysicists are devoting their careers to studying a problem, you can be sure that nothing you come up with off the top of your head, without knowing anything about the subject, is going to have any merit.
And then there is the issue of whether we really need to emulate how natural brains work to get strong AI.
There is a Russell and Norvig quote that I rather like because it does help reveal the important issues: “The quest for ‘artificial flight’ succeeded when the Wright brothers and others stopped imitating birds and started using wind tunnels and learning about aerodynamics.”
Most people I have discussed AI with, and know of this (well known) quote draw the conclusion from this analogy that we don't really need to imitate brains to get AI, so we don't really need to learn about them first either. We will get strong AI through other (unspecified) means.
It is true that we don't make airplanes by imitating birds. But we did have learn how birds fly before we could build an airplane (all that stuff about "wind tunnels" and "aerodynamics") and as it happens we could make models that flew like birds before we could make an airplane. And we weren't happy being able to make bird-models, we needed something far larger and faster than any bird to be useful.
But with AI we are still debating what "intelligence" even is and have no knowledge about its fundamental principles yet. So no building AI "planes" any time in the foreseeable future. But we don't need to have AI that is "larger and faster" than any brain to be useful. If we were able to get anywhere close to human level intelligence, our AI problems would be mostly solved.
Deep learning and other related machine learning techniques are proving very useful for a wide range of tasks. We don't need to "start over" to advance useful machine learning techniques.
Hinton seems to mean to get "strong AI". Yes, I read TFA, but the strength of Axios articles is that they are very short, but that is also their weakness. Very little is actually said in TFA.
We are a long, long way from anything that emulates a natural neural system at any level.
Consider Caenorhabditis elegans. Every cell in this simple worm has been mapped, also the development of every cell from a single cell has been mapped (male worms have 1031 cells). We know every cell in its nervous system (there are 302), and every cell that each cell is connected to, and we know the type of connections for all. What's more we have completely sequenced its genome. We know more about this little multi-cell organism than any other multi-cell animal on the planet.
Since we know every cell in its nervous system, and every connection between every cell, we must be able to emulate this worm's "brain"! Heck we must be able to "upload" the worm's brain to a computer! Right? Right?
No.
We are still working on understanding the functioning and capabilities of a single neuron in its brain. That has proven so complex as to defy characterization thus far. We are essentially nowhere in understanding how this 302 cell brain works despite decades of effort.
Meanwhile Kurzweil has changed his prediction of "when computers will have human-level intelligence" from 2020 to 2029. I guess believing it was going to happen in the next 26 and a half months was cutting it a little too close. I have been reading about his predictions about AI for a couple of decades now and have yet to see any explanation of how he imagines this is going to happen - other than his expectations about hardware capabilities, and that there is still an unspecified "software issue" that needs to be solved. Indeed.
Slashdot Mirror
If you ever tune in to the Bigfoot/Sasquatch enthusiast groups you will find that there are tons of sightings, and lots and lots of physical evidence, hair, coprolites, stuff that would be loaded with Bigfoot DNA. It it was real.
Despite the fact that anyone can run a DNA test that identifies species for under $100 today, and that finding a single sample of DNA from an unknown primate would make the person finding it famous forever, not a single specimen claimed to be from a Bigfoot has ever been found to have anything but a known animal (usually human, sometimes bear, or other known mammal) as its source.
True story, a friend of mine told the pain clinic she was using a THC creme on her knee to help with the pain and that it was working great, the pain clinic said that by self treating they could no longer prescribe her pain meds. So now she can't get legal meds and is relegated to only the 'illegal' ones.
Not in Washington State, Oregon, Colorado, and a growing (heh, growing...) list of others...
You are profoundly mistaken. Opioid painkillers are regulated directly by the Federal government, and state laws are irrelevant. They piss-test opioid patients in California, the home of medical cannabis, and where recreational pot is also now legal, and if they detect cannabis residue they cut off access to the opioid medication.
Nah. It is entirely due to electric lighting.
I calculated once how much light was put out in London during the gas light era. It turns out the total number of lumens emitted to light the largest and most brightly lit city in the world at the time (four million people) is equal to the amount of light used to illuminate one football field for a night game.
Is the park actually safer or do you just think it is because it's well-lit?
This is a very important point. People assume that lights provide security, but the evidence for this is at best very weak. Most studies that have been done (research on this is limited) have found lighting increases the crime rate!
And so you see completely empty areas that are closed and fenced off at night brightly lit for "security", and remote service buildings, water towers, etc. with lights on them for no benefit whatsoever. It is similar to car alarms. If you hear a car alarm go off do you rush out to frustate the possible car thief? No one does this. Everyone just ignores it. Similarly unless there is someone actually watching the light deters no one, it just helps the criminals case the joint and spot potential victims to attack.
Everyone wants "security" so bright lights go up everywhere.
At least have them shut off after midnight.
Simply redesigning them (and other outside illumination lighting) would be a huge gain. A bit less than 10% of light hitting the ground in an urban area gets scattered back up into the sky. So if the lights are designed to shine light on the ground only (no light being sent horizontally, where it just creates glare, or worse directly into the sky) there would be a large reduction in light pollution.
LEDs can help a lot here since they are inherently directional, it takes effort to make them spew light in other directions. But light makers are willing to provide them to the market since people want to buy lights that resemble lights they are used to.
Similarly regulating sign lighting so that you don't have bright lights at the bottom of a sign pointing straight-up, and regulating the use of light as a form of advertising and promotion. A lot of commercial light use is abusive, using brilliant light-emitting signage throwing light everywhere.
On highways with light traffic late at night "smart" lights can be implemented that turn off when there are no cars for them to aid.
The problem is that it isn't just (or primarily) with ad hoc and custom analyses.
It is that regular business functions are run with these sheets all the time. Business types do this because the are familiar with the tool, and they can implement the process themselves without calling in a dev team.
And all of this is perfectly understandable. Would you call in an outside dev team, explain requirements, and then have to wait for an acceptable product to be produced, when you could do it yourself quickly?
This is inevitable unless considerable effort is expended by the organization to identify and pull these business functions into formal, administered, monitored systems.
A decree not to use Excel at all (if this is what it is) is stupid.
The emphasis should be on educating the business on treating these "normal function" spreadsheets as prototypes of the function that must be implemented formally going forward, and the necessary resources must be provided to make this happen, and suitable reward structures must exist to encourage businessmen to identify and bring forward these functions for proper automation. Without all of this this decree will be useless.
The mention of 106Ru without any mention of the other more prominent Radionuclides leads me to the conclusion that whatever bad happened was very recent, and not six decades back. That would be some 59 Half-Lifes....
Mod this guy up, everything he says is exactly right (and he provides a very informative link).
This is a release from an active plutonium processing plant which is handling relatively fresh fuel (less than a decade old, perhaps very fresh).
The AID virus is not "finding a way" to defeat three drug therapy. Three different protease inhibitors target the critical protease molecule in different ways such that it cannot mutate to defeat all three and still work.
This work is not the final stage, it is simply showing a two step mechanism that requires two gene segments be lost in a particular order for the kill switch to be deactivated. More complex schemes are easily imagined, two different two component kill switches; mutually lethal two component kill switches so that they both must be lost together in the same mutation process, etc.
Seriously, has no one read TFA? (I know, I know, this is /.)
This work addresses exactly that problem, they have installed two mechanisms in different parts of the genome both of which produce different toxins. The regular kill switch, which kills the organism with toxin A when an environmental trigger is detected, also continuously produces an anti-toxin for toxin B, which the other mechanism continuously produces. Losing the trigger part also edits out antitoxin B, killing the cell immediately. A two step process of losing the toxin B gene, followed by losing the trigger would be needed to eliminate the kill switch.
If that is not secure enough then you could make a three part system following a similar strategy.
This is reminiscent of how they developed an effective treatment for AIDs that renders it non-fatal. Any drug they came up with that targeted a vulnerable site in the virus would fail since that vulnerable site would be lost in the the very rapid mutation process of the virus. Even using two drugs that targeted different sites had limited success. But a three drug therapy has worked - targeting different regions of the protease so that it cannot simultaneously mutate enough to deactivate all of them, without also deactivating its function.
A good directed fragmentation payload could get a reliable kill out to 25 meters (just think of it as one quarter segment of a Claymore mine), which is much more than "a couple of meters", a drone flying straight in and detonating at 25 meters might tax Secret Service perimeter protection to get the President behind cover before the kill shot was made. And only he and the VP have this level of protection. Anybody else in the U.S. would be far more vulnerable.
And sure booby trapping a car and getting within physical attack range works (if get close enough to throw the grenade accurately, why not just shoot him?) but VIPs with body guards or security details make these things hard, and attacking the target yourself is generally one-time thing, you never get another chance as you will end up in custody (and possibly deceased). With drones you can attack from long range and not get caught. Big, big difference.
A human piloted drone would be a better single-target assassination weapon (as the U.S. drone program proves regularly). The Slaughterbots video emphasizes AI and facial recognition because using those technologies you can have autonomous attacks at large scale. Even in the case of attacking a U.S. President a swarm attack of drones that simply recognize heads (no need to find recognize faces) will permit a small swarm to indiscriminantly kill everyone in the vicinity of the President, including him or her.
Also the video is absolutely correct if you can get 3 grams of HE in close contact with a human head, by having the drone simply fly into the target, it will kill the target. I cited the use of fragmentation devices simply to show that even actual contact is not needed so that impenetrable perimeter defenses have to be deployed quite a ways away.
If your knowledge of explosives is limited to fireworks, you don't know anything about the subject. Don't you think it odd that no military employs fireworks as weapons?
Here are three types of explosive munitions that would be quite deadly even in "hobby drones": explosive formed projectiles (EFPs), directed fragmentation munitions like the Claymore mine, and then there are ordinary fragmentation grenades.
A drone/EFP system designed for assassination (aiming and attacking a specific target) could use an EFP weighing only 10 grams to fire a 3.6 gram projectile (same weight as a typical M-16 round) at 2000 m/sec (twice the velocity and four times the kinetic energy of the M-16 round at the muzzle). This actually a much more powerful projectile than what you need to kill someone. EFPs are quite easy to make, there are "energetic materials" hobbyists who post their "experiments" online. Large EFPs made with simple tools were a huge problem to U.S. forces in Iraq.
Not happy with taking only a single shot? Well then multiple fragmentation weapons are just the thing. You can use an array of EFPs like that described above to send a cluster of them at your target. Or consider the venerable Claymore mine that weighs 1600 grams and fires 700 projectiles at 1300 m/sec. At 25 meters these are 100% lethal in a zone 25 meters wide. For a drone you would probably make a scaled down version, even a narrow cone "micro" version to send a cluster of high velocity projectiles into your target.
Simply replicating the M67 fragmentation grenade which weighs 400 grams and is 100% lethal at 5 meters in all directions would work fine.
With a suitably designed fragmentation device, if you can get a 100 gram payload within 5 meters of a target you can get a kill with 100% certainty without even requiring sophisticated aiming or munition design. With more effort you can do the same at greater range, or with even lighter payloads.
As if company directors even actually faced jail terms for misconduct.
You are profoundly ignorant on the subject, and yet you feel obligated to make pronouncements about it. You must be a /. regular!
Really, if you are unaware - as you clearly are - that it is the rotation of galaxies, like the nearly Andromeda, and our own, that were the first evidence of the existence of dark matter then you know nothing about the subject.
Try reading the Wikipedia page which lists ten different pieces of evidence showing its existence, with links to further material on each.
Because what are the options here now? Windows 10 and MacOS. Spy ware and a fortune for closed hardware. Two corporations who, in different ways, think you do not own your own computer.
That is why some of us use Linux on the desktop and wish that making that painless would interest someone in Linux-land.
I also wish the British Museum would return all those things to the countries the British stole them from.
First crash it's done forever, lawsuits will bankrupt the company, and it will set flying taxis back by 50 years. Blade Runner I will never happen. Fuck these morons for trying to move too quickly.
These don't crash so easily. They have multiple redundancy in the rotor array, the craft itself is self-flying - the "pilot" just guides it where to go, and it does have that aircraft safety parachute if everything fails to bring the craft to the ground in a not-catastrophic manner.
The SkyWedgie!
The technology is apparently here to support this vehicle.
Which costs $350,000 and can carry 160 kg a maximum distance of 27 km at 50 km/h (that is 350 lb, 17 miles, and 42 MPH to those using pre-1795 measuring systems) and takes 2 hours to fully charge.
With a charge time:flight time ratio of 2:1 you could get in 8 hours of flight a day, around the clock, or about 400 km total travel distance. If used as a short range shuttle over a congested city center you might get in up to forty 10 km trips a day, maybe ~10,000 in a year. Looks like this could be a profitable service with a not-astronomical fare price.
There is however an (optional) full aircraft parachute that brings the entire vehicle to the ground. This is called a "ballistic total separation system". I am puzzled though about what gets "separated" here. Ejecting any part of the vehicle is going to be a problem for people on the ground.
All true.
In addition there is the problem of those who people who actually lost their jobs from "creative destruction". Even a short period of "no job" is a very serious problem for nearly everybody, and even if new jobs come along "soon", say in a few years, those few years can be devastating. And even if new jobs do come along, is the person who lost work able to get any of those replacements?
The degree to which society can, or should, provide support for people whose jobs are eliminated by automation is a matter of serious debate. Most advanced economies (e.g. Germany) takes this seriously has well-supported programs in place and yet does very well economically and competitively. But Kurzweil dismisses this as even being a problem at all that needs to be addressed in any fashion.
That doesn't mean they came from elsewhere, it would also sync up really well with them coming from the blackhole in the center of our galaxy and being curved back inward by the gravity of the whole galaxy - sort of a galactic-scale particle accelerator. (Like the field lines of a magnet.)
Or instead of "really well", not all in any way, shape, or form. You have no idea of the physics involved, typing words is not a physical analysis.
The galactic escape velocity for an iron atom (a typical heavy cosmic ray particle) is about 10^5 eV. All of the cosmic rays under discussion have energies greater than 10^19 eV, or 100 trillion times more energetic than the galactic escape velocity energy.
The galactic magnetic field is much better at holding on to cosmic rays, but cannot confine them above an energy of 10^18 eV or so. Which is why the researchers are studying extragalactic cosmic rays with energies above 10^19 eV. They know these cannot be confined to the Milky Way.
If hundreds of professional astrophysicists are devoting their careers to studying a problem, you can be sure that nothing you come up with off the top of your head, without knowing anything about the subject, is going to have any merit.
Thanks!
And then there is the issue of whether we really need to emulate how natural brains work to get strong AI.
There is a Russell and Norvig quote that I rather like because it does help reveal the important issues: “The quest for ‘artificial flight’ succeeded when the Wright brothers and others stopped imitating birds and started using wind tunnels and learning about aerodynamics.”
Most people I have discussed AI with, and know of this (well known) quote draw the conclusion from this analogy that we don't really need to imitate brains to get AI, so we don't really need to learn about them first either. We will get strong AI through other (unspecified) means.
It is true that we don't make airplanes by imitating birds. But we did have learn how birds fly before we could build an airplane (all that stuff about "wind tunnels" and "aerodynamics") and as it happens we could make models that flew like birds before we could make an airplane. And we weren't happy being able to make bird-models, we needed something far larger and faster than any bird to be useful.
But with AI we are still debating what "intelligence" even is and have no knowledge about its fundamental principles yet. So no building AI "planes" any time in the foreseeable future. But we don't need to have AI that is "larger and faster" than any brain to be useful. If we were able to get anywhere close to human level intelligence, our AI problems would be mostly solved.
Deep learning and other related machine learning techniques are proving very useful for a wide range of tasks. We don't need to "start over" to advance useful machine learning techniques.
Hinton seems to mean to get "strong AI". Yes, I read TFA, but the strength of Axios articles is that they are very short, but that is also their weakness. Very little is actually said in TFA.
We are a long, long way from anything that emulates a natural neural system at any level.
Consider Caenorhabditis elegans. Every cell in this simple worm has been mapped, also the development of every cell from a single cell has been mapped (male worms have 1031 cells). We know every cell in its nervous system (there are 302), and every cell that each cell is connected to, and we know the type of connections for all. What's more we have completely sequenced its genome. We know more about this little multi-cell organism than any other multi-cell animal on the planet.
Since we know every cell in its nervous system, and every connection between every cell, we must be able to emulate this worm's "brain"! Heck we must be able to "upload" the worm's brain to a computer! Right? Right?
No.
We are still working on understanding the functioning and capabilities of a single neuron in its brain. That has proven so complex as to defy characterization thus far. We are essentially nowhere in understanding how this 302 cell brain works despite decades of effort.
Meanwhile Kurzweil has changed his prediction of "when computers will have human-level intelligence" from 2020 to 2029. I guess believing it was going to happen in the next 26 and a half months was cutting it a little too close. I have been reading about his predictions about AI for a couple of decades now and have yet to see any explanation of how he imagines this is going to happen - other than his expectations about hardware capabilities, and that there is still an unspecified "software issue" that needs to be solved. Indeed.
They need a solution for municipal water systems so all the plastic from everybody gets trapped in one place.
Google Pre-Delete?
It was all a confusion due to homophones. It was supposed to be "Do Know Evil".