There is a huge difference between being killed or injured by a human driver and being killed or injured by a self-driving mechanism. In the first case, the human driver is either to be blamed or not. In the second case, you or your next of kin have to deal with a large corporation that is guaranteed to have top lawyers, and they will be constantly shifting the blame.
True unfortunately. Also wonder what will be the position insurance company? How are these cars insured and does fault have to be assigned before they will payout?
It makes no sense for a large corporation of the sort that makes self-driving systems to buy insurance. They'll self-insure. And, no, they won't pay until fault is assigned.
There is a huge difference between being killed or injured by a human driver and being killed or injured by a self-driving mechanism. In the first case, the human driver is either to be blamed or not. In the second case, you or your next of kin have to deal with a large corporation that is guaranteed to have top lawyers, and they will be constantly shifting the blame.
In the first case you'll be dealing with the lawyers of the insurance company of the human driver, until they've paid all they're going to pay. Their limits are clearly defined in the policy. Then you'll be dealing with an individual who has no money to give you.
In the second case you'll be dealing with the lawyers of the makers of the self-driving mechanism. Who will have deep pockets and no contractual limitation on liability.
I think you'll be better off in the second case, because that company will be more sensitive to worries about bad PR, regulatory agencies, etc. And you'll find it easy enough to get a good lawyer who will be happy to work for a cut of the settlement.
Google has a sort of a tradition, going back four or five years, of employees volunteering their salary information via a Google Docs spreadsheet (actually a Google Docs Form, with results summarized on a read-only spreadsheet). Not a lot of them, but enough to be interesting. In 2017 3.64% of them (us; I participated) did it. Sharing your name is optional. About 15% of those who share their salary info provide their name. I did. The system also obviously knows exactly who participated so if management wanted to they could find identify the "anonymous" contributors.
The sheet has columns for base salary, bonus and equity grants, as well as job ladder, level and location. For employees at major sites (e.g. Mountain View), there are enough entries to give people a reasonable view of what the range is, so they can see how they compare. The data is self-reported and not verified against anything, so people could lie, but I expect few do. If any.
The company doesn't encourage openness about compensation, but it doesn't chastise those who share, either. I think it's useful.
Turning off location history will prevent storage of new history. In the settings dialog in Android where you turn off location history there's also a button that will delete existing history. Alternatively, you can go to myactivity.google.com and delete it (as well as a bunch of other stuff) from there.
Obviously, all of this depends on Google actually deleting what they say they're going to delete, and not storing what they say they're not going to store, and you have no direct way of verifying. There would be significant negative consequences to Google if it were found to be lying about those things, though, both legal and PR. Plus I'll tell you I know some of the people who work on the deletion infrastructure, and they work hard to make sure gone is gone. But I'm some random guy on the Internet who could be lying, so my assurance is worth less than the observation that the FTC and many others would take a dim view of Google deceiving users.
I believe Lauren Weinstein used to be paid as some kind of Google shill and now that the agreement has been terminated, he seems to have turned on his former employer. Most of his posts now seem to have an anti-Google bias.
In your world, does anyone ever say anything merely because they believe it, rather than because it supports their tribe or pays their bills?
It scares me that google, a greedy for-profit company, has all that personal data.
Then you should turn off location history, so they won't have it. With history off but location on, location data is only uploaded as needed to satisfy app requests and is not retained.
Personally, I find location history to be very useful. I like being able to see where I was on any given day and time. I like it enough that I periodically go in and correct any errors in Google's guesses as to where I was while I still remember (there are a fair number of errors because for power efficiency GPS is used as little as possible).
No need to trust. A cheap RF signal meter can tell you for sure.
It can - so long as they aren't turning on the transmitters in the phone in short bursts once every minute or so. In that case you need something that can log those short bursts of RF over a specified band or bands.
The goal was to detect whether it was transmitting, not to log the transmissions. You can detect short bursts with a cheap RF signal meter, you just need to watch it. In fact, phones do transmit short bursts quite a bit, to avoid having to keep the transmitters powered up unnecessarily. They suck a lot of power, which drains your battery.
Even then, your phone might be receiving in stealth mode, in which case there might be provisions for making it 'phone home' on demand.
Perhaps, but now you're into stratospheric levels of paranoia. Actually, this sort of behavior would be pretty easy to detect just by watching battery levels while airplane mode is active, because in order to receive in stealth mode, the phone would have to power up the radios periodically.
If that capability doesn't exist now, it almost certainly will in the future.
Could not the device continue to collect location data without emitting ref signals?
Airplane mode disables transmitters, not receivers, so obviously it could. The comment I replied to was implying that airplane mode didn't do what it claims to, not that it didn't do things it doesn't claim to.
Trust the PRISM brands with their hardware to say off is off?
No need to trust. A cheap RF signal meter can tell you for sure. And what are the odds that no one would have noticed and blown the whistle if airplane mode didn't actually work?
Sigh. This site used to be populated by people with a clue. This is like all of those people who believe that smart speakers must be sending 24/7 audio to the cloud, but don't bother to simply measure the data the devices send/receive at their routers and do the math.
Paranoia is well and good, but being paranoid about a possibility that you can easily check yourself is stupid. Computers aren't magic.
Although this is obviously the way the world works, I would point out it both a) leaves us open to insider trading and b) distorts the free market via incomplete information among consumers.
Most companies that make this sort of information generally available to employees inform those employees that they're legally considered insiders by the SEC and that they are barred from trading the company's stock except during specific trading windows.
Executives at all publicly-traded companies fall under these restrictions.
Oh, it'll take a while, but this, Xi Jinping's newly-minted dictatorship and other crackdowns make it quite clear that China is beginning a fast slide into 1984-land. Their screw-tightening will provoke resistance from the large, recently-created middle class, which will in turn provoke more oppression, and the need for tighter central control of everything -- including the economy. Bye, bye capitalism and the growth that it brings. I'm guessing the new leaders are smart enough not to eviscerate their country with another Great Leap Forward, but they'll still have to gut the free market economy that has been driving such huge growth.
Assuming we can survive Trump, maintain the rule of law and avoid descdending into a new patrimonialization, this means the US will soon be the world's sole superpower again (no, Putin's kleptocracy is no competition. Unless we emulate it, which is exactly what Trump would like, but I don't think there's a chance of it happening).
The flip side is that we really need China to continue their aggressive response to climate change, which they'll forget all about if they end up spending all their time on suppressing dissidents.
You're nuts. I'd agree that perhaps it should be four, maybe, and that three is a little shorter than is ideal. But "ridiculously"? I think you need to buy a dictionary.
Oh, it's also worth pointing out that AFAICT no one else commits to even three years. That includes Apple. Apple generally provides support for four, maybe five years, but they don't make any promises.
You're nuts. I'd agree that perhaps it should be four, maybe, and that three is a little shorter than is ideal. But "ridiculously"? I think you need to buy a dictionary.
Go to "Settings", then scroll down. The bottom 80% of the main Settings menu is a list of all your apps. Click on any one of those to see what permissions it has asked for and/or been granted.
Android has this, too, and it's not what was requested.
Okay, but if you're going to make a new system that meets your requirements better, would it kill you to include at least some semblance of the advantages of previous systems in the new one?
For example, what3words codes are easy to remember - they are simply 3 words. A typical Open Location Code is 7FG49QCJ+2V. So... yeah.
Because what3words suck, for many reasons.
1. They're proprietary, and kind of inherently so. Their design approach requires a big table providing all of the mappings, rather than a simple, easily open-sourced algorithm. Google could have tried to reproduce their approach, but likely would have run afoul of their patents.
2. Their approach also doesn't scale up and down like OLC does; in what3words you can't specify a region larger than 3m^2, or a smaller one. With OLC each additional pair of digits allows specification of a sub-region 1/400th the size of the "parent" region.
3. The approach provides no indication of proximity. Given two locations you have no idea how far apart they are without first translating them to some other system.
4. They're language-dependent. A triplet of English words is no more sensible to a Tamil-speaker than an OLC code. A Tamil-language version could be constructed, but the entire mapping would need to be built for every language. And not every language has as many words as English. what3words has not included the oceans in either of the other two languages they support for this reason.
In any case, it would be trivial to layer a mechanical word mapping on top of OLC if you wanted. OLC values are fundamentally numeric (base 20), and it's trivial to pick an appropriately-sized dictionary and map onto a different base. To preserve the area/local distinction in OLC (note that typical OLC codes are not " 7FG49QCJ+2V" they're " ". In your example it would be "Tinzaouten 9QCJ+2V") I'd probably choose an 800-word dictionary so that one word corresponds to two "digits". A full 10-digit location would require five words, not three, but normally it would be city name plus three words.
Yes, and a few other location coding systems are similar as well. However, Google have their reasons for creating a new system. You can find their evaluation of the various systems explained here:
Only if you don't understand the comic, or don't understand the write-up, or both.
The point of the comic is that there are a whole bunch of standards and the idea is to invent a single new standard to replace them all. Which doesn't work, and just adds to the pile of standards.
The point of the plus codes writeup is to evaluate the existing standards to see if any of them meets the requirements of one particular set of use cases. Since it's determined that no existing standard does the job, a new one is created, not to replace the others but to address the requirements at hand.
e.g. what are the specific risks? Because all I ever hear is backhanded fearmongering.
If you're really interested, get a copy of Nick Bostrom's book "Superintelligence" from the library. You don't even need to read the whole thing, just the first few chapters will give you a good overview of the risks and how extraordinarily challenging solutions may be.
I'll try to give a summary here, but, really, it's worth investing a couple hours to read the book because I won't do it justice.
The core problem is that there is no plausible reason to expect that artificial intelligence will be limited to the level of human intelligence. On the contrary, it makes sense that once we figure out how to build an artificial general intelligence (AGI -- note, qualitatively completely different from the task-focused ML/AI we have now) that is even slightly smarter than we are, that AGI will be able to use the knowledge we had to discover in designing and building it to design its successor (or redesign itself) to be smarter yet, and so on. The rate at which AGI will increase will be limited only by the rate at which it can design and build (or cause to be built) improved versions.
This is known as the "AI Singularity", the point at which AGI takes over the development of AGI and it explodes out of control, becoming a superintelligence (SI) or group of SIs vastly, vastly smarter than any human has ever been, or could ever be given the limitations of our wetware.
This could be the greatest boon humanity has ever seen, if the superintelligence(s) choose helping humanity as their goal. Assuming, of course, that their definition of "helping" is something that we would agree with. Bostrom points out that even if we are in complete control of the goals of the SI, what goal would we set that could not backfire? Suppose we set the SI the goal of maximizing human happiness. The SI might decide that it can create maximal human happiness by (a) maximizing the number of humans and (b) making each of them maximally happy. To accomplish (b), it might implant electrodes in the pleasure centers of our brains and stimulate them constantly. To accomplish (a) it might choose any of many means of maximizing reproduction, and minimize the cost of maintaining each human life by extracting our brains and placing them in vats, fed by a trickle of nutrients and a trickly of happiness-inducing electricity. It could probably support hundreds of trillions of such brains on Earth by dismantling the entire ecosystem and turning it to this one purpose.
So "maximum happiness" is a bad goal. What's a good one? We don't know. It's a hard problem.
But what if we don't get to set the goal, or what if we inadvertently set it to something really stupid. What if the system in charge of ensuring office supply availability somehow made the jump to AGI, and in order to make itself more effective turned itself into an SI whose sole goal was making paperclips. It might ultimately disassemble the whole planet for paperclip-making resources.
These might seem like ridiculous sci-fi fantasy problems... but the problem is that they're not only plausible, there's nothing at all improbable about them. They depend on only one assumption: That it's possible to build an AGI which is smarter -- even a little bit -- than the smartest humans, and there's no logical basis for refuting that assumption.
Further, it's impossible to know how far we are from knowing how to build such an AGI. We lack the necessary theory of cognition to let us understand what it is, exactly, that makes us intelligent in the way we are, able to think abstractly and problem-solve. Given that we don't have that theory, we can have no idea how far we are from acquiring that theory. It's possible someone had the necessary flash of insight in the shower yesterday. Or last month. It's possible that we're still centuries away (though this seems very unlikely).
Even if we did decide to regulate AI development, it's far fro
“If Jaycee is capable, he can make his own money. If he is not, then he will just be wasting my money.” – Jackie Chan on why he wont be passing down his fortune.
Yep. Bill Gates and many others have done the same. Give your kids all the opportunities you can while they're gowing up, a stable, loving home life, opportunities to learn and grow and do things, excellent healthcare and the best education you can. If after all that, they still can't make their own success, then they shouldn't have it. Not that you should leave them to die in a gutter, if that's what it comes to. It's fine for parents to act as a safety net. But if the kids want to live well, they need to do it themselves.
Slashdot Mirror
Homeopathy usually argues that the memory of the "disease" will trigger a cure
Basically the notion underlying vaccination, but bonkers.
Google also has no leadership and has devolved into a high tech Lord of The Flies society.
Surely you mean Lord of the Files.
I mean that about as much as you "usually delete AC replies without reading them."
I usually do.
Google also has no leadership and has devolved into a high tech Lord of The Flies society.
Surely you mean Lord of the Files.
There is a huge difference between being killed or injured by a human driver and being killed or injured by a self-driving mechanism. In the first case, the human driver is either to be blamed or not. In the second case, you or your next of kin have to deal with a large corporation that is guaranteed to have top lawyers, and they will be constantly shifting the blame.
True unfortunately. Also wonder what will be the position insurance company? How are these cars insured and does fault have to be assigned before they will payout?
It makes no sense for a large corporation of the sort that makes self-driving systems to buy insurance. They'll self-insure. And, no, they won't pay until fault is assigned.
There is a huge difference between being killed or injured by a human driver and being killed or injured by a self-driving mechanism. In the first case, the human driver is either to be blamed or not. In the second case, you or your next of kin have to deal with a large corporation that is guaranteed to have top lawyers, and they will be constantly shifting the blame.
In the first case you'll be dealing with the lawyers of the insurance company of the human driver, until they've paid all they're going to pay. Their limits are clearly defined in the policy. Then you'll be dealing with an individual who has no money to give you.
In the second case you'll be dealing with the lawyers of the makers of the self-driving mechanism. Who will have deep pockets and no contractual limitation on liability.
I think you'll be better off in the second case, because that company will be more sensitive to worries about bad PR, regulatory agencies, etc. And you'll find it easy enough to get a good lawyer who will be happy to work for a cut of the settlement.
Google has a sort of a tradition, going back four or five years, of employees volunteering their salary information via a Google Docs spreadsheet (actually a Google Docs Form, with results summarized on a read-only spreadsheet). Not a lot of them, but enough to be interesting. In 2017 3.64% of them (us; I participated) did it. Sharing your name is optional. About 15% of those who share their salary info provide their name. I did. The system also obviously knows exactly who participated so if management wanted to they could find identify the "anonymous" contributors.
The sheet has columns for base salary, bonus and equity grants, as well as job ladder, level and location. For employees at major sites (e.g. Mountain View), there are enough entries to give people a reasonable view of what the range is, so they can see how they compare. The data is self-reported and not verified against anything, so people could lie, but I expect few do. If any.
The company doesn't encourage openness about compensation, but it doesn't chastise those who share, either. I think it's useful.
Turning off location history will prevent storage of new history. In the settings dialog in Android where you turn off location history there's also a button that will delete existing history. Alternatively, you can go to myactivity.google.com and delete it (as well as a bunch of other stuff) from there.
Obviously, all of this depends on Google actually deleting what they say they're going to delete, and not storing what they say they're not going to store, and you have no direct way of verifying. There would be significant negative consequences to Google if it were found to be lying about those things, though, both legal and PR. Plus I'll tell you I know some of the people who work on the deletion infrastructure, and they work hard to make sure gone is gone. But I'm some random guy on the Internet who could be lying, so my assurance is worth less than the observation that the FTC and many others would take a dim view of Google deceiving users.
I believe Lauren Weinstein used to be paid as some kind of Google shill and now that the agreement has been terminated, he seems to have turned on his former employer. Most of his posts now seem to have an anti-Google bias.
In your world, does anyone ever say anything merely because they believe it, rather than because it supports their tribe or pays their bills?
It scares me that google, a greedy for-profit company, has all that personal data.
Then you should turn off location history, so they won't have it. With history off but location on, location data is only uploaded as needed to satisfy app requests and is not retained.
Personally, I find location history to be very useful. I like being able to see where I was on any given day and time. I like it enough that I periodically go in and correct any errors in Google's guesses as to where I was while I still remember (there are a fair number of errors because for power efficiency GPS is used as little as possible).
No need to trust. A cheap RF signal meter can tell you for sure.
It can - so long as they aren't turning on the transmitters in the phone in short bursts once every minute or so. In that case you need something that can log those short bursts of RF over a specified band or bands.
The goal was to detect whether it was transmitting, not to log the transmissions. You can detect short bursts with a cheap RF signal meter, you just need to watch it. In fact, phones do transmit short bursts quite a bit, to avoid having to keep the transmitters powered up unnecessarily. They suck a lot of power, which drains your battery.
Even then, your phone might be receiving in stealth mode, in which case there might be provisions for making it 'phone home' on demand.
Perhaps, but now you're into stratospheric levels of paranoia. Actually, this sort of behavior would be pretty easy to detect just by watching battery levels while airplane mode is active, because in order to receive in stealth mode, the phone would have to power up the radios periodically.
If that capability doesn't exist now, it almost certainly will in the future.
Sure, all part of the Illuminati's grand plan.
Could not the device continue to collect location data without emitting ref signals?
Airplane mode disables transmitters, not receivers, so obviously it could. The comment I replied to was implying that airplane mode didn't do what it claims to, not that it didn't do things it doesn't claim to.
I could constantly monitor and analyse my network traffic or I could just save myself time and money and not buy something I don't trust.
And which mobile phone can you trust?
Trust the PRISM brands with their hardware to say off is off?
No need to trust. A cheap RF signal meter can tell you for sure. And what are the odds that no one would have noticed and blown the whistle if airplane mode didn't actually work?
Sigh. This site used to be populated by people with a clue. This is like all of those people who believe that smart speakers must be sending 24/7 audio to the cloud, but don't bother to simply measure the data the devices send/receive at their routers and do the math.
Paranoia is well and good, but being paranoid about a possibility that you can easily check yourself is stupid. Computers aren't magic.
Apple has no official update policy.
Although this is obviously the way the world works, I would point out it both a) leaves us open to insider trading and b) distorts the free market via incomplete information among consumers.
Most companies that make this sort of information generally available to employees inform those employees that they're legally considered insiders by the SEC and that they are barred from trading the company's stock except during specific trading windows.
Executives at all publicly-traded companies fall under these restrictions.
There goes China.
Oh, it'll take a while, but this, Xi Jinping's newly-minted dictatorship and other crackdowns make it quite clear that China is beginning a fast slide into 1984-land. Their screw-tightening will provoke resistance from the large, recently-created middle class, which will in turn provoke more oppression, and the need for tighter central control of everything -- including the economy. Bye, bye capitalism and the growth that it brings. I'm guessing the new leaders are smart enough not to eviscerate their country with another Great Leap Forward, but they'll still have to gut the free market economy that has been driving such huge growth.
Assuming we can survive Trump, maintain the rule of law and avoid descdending into a new patrimonialization, this means the US will soon be the world's sole superpower again (no, Putin's kleptocracy is no competition. Unless we emulate it, which is exactly what Trump would like, but I don't think there's a chance of it happening).
The flip side is that we really need China to continue their aggressive response to climate change, which they'll forget all about if they end up spending all their time on suppressing dissidents.
Three years is ridiculously short?
YES!
You're nuts. I'd agree that perhaps it should be four, maybe, and that three is a little shorter than is ideal. But "ridiculously"? I think you need to buy a dictionary.
Oh, it's also worth pointing out that AFAICT no one else commits to even three years. That includes Apple. Apple generally provides support for four, maybe five years, but they don't make any promises.
Three years is ridiculously short?
YES!
You're nuts. I'd agree that perhaps it should be four, maybe, and that three is a little shorter than is ideal. But "ridiculously"? I think you need to buy a dictionary.
Honest question: Where can I find this in iOS?
Go to "Settings", then scroll down. The bottom 80% of the main Settings menu is a list of all your apps. Click on any one of those to see what permissions it has asked for and/or been granted.
Android has this, too, and it's not what was requested.
Given the ridiculously short amount of time Android devices get updates -- including devices from Google itself
Three years is ridiculously short?
Okay, but if you're going to make a new system that meets your requirements better, would it kill you to include at least some semblance of the advantages of previous systems in the new one? For example, what3words codes are easy to remember - they are simply 3 words. A typical Open Location Code is 7FG49QCJ+2V. So... yeah.
Because what3words suck, for many reasons.
1. They're proprietary, and kind of inherently so. Their design approach requires a big table providing all of the mappings, rather than a simple, easily open-sourced algorithm. Google could have tried to reproduce their approach, but likely would have run afoul of their patents.
2. Their approach also doesn't scale up and down like OLC does; in what3words you can't specify a region larger than 3m^2, or a smaller one. With OLC each additional pair of digits allows specification of a sub-region 1/400th the size of the "parent" region.
3. The approach provides no indication of proximity. Given two locations you have no idea how far apart they are without first translating them to some other system.
4. They're language-dependent. A triplet of English words is no more sensible to a Tamil-speaker than an OLC code. A Tamil-language version could be constructed, but the entire mapping would need to be built for every language. And not every language has as many words as English. what3words has not included the oceans in either of the other two languages they support for this reason.
In any case, it would be trivial to layer a mechanical word mapping on top of OLC if you wanted. OLC values are fundamentally numeric (base 20), and it's trivial to pick an appropriately-sized dictionary and map onto a different base. To preserve the area/local distinction in OLC (note that typical OLC codes are not " 7FG49QCJ+2V" they're " ". In your example it would be "Tinzaouten 9QCJ+2V") I'd probably choose an 800-word dictionary so that one word corresponds to two "digits". A full 10-digit location would require five words, not three, but normally it would be city name plus three words.
If you'd like to fiddle with this, I have some code I put together to do something very similar. https://github.com/divegeek/me...
Yes, and a few other location coding systems are similar as well. However, Google have their reasons for creating a new system. You can find their evaluation of the various systems explained here:
https://github.com/google/open...
That write-up is pretty much a perfect case study of the classic xkcd comic "There are 14 competing standards".
Only if you don't understand the comic, or don't understand the write-up, or both.
The point of the comic is that there are a whole bunch of standards and the idea is to invent a single new standard to replace them all. Which doesn't work, and just adds to the pile of standards.
The point of the plus codes writeup is to evaluate the existing standards to see if any of them meets the requirements of one particular set of use cases. Since it's determined that no existing standard does the job, a new one is created, not to replace the others but to address the requirements at hand.
it's a standard feature on the phone's cell voice calling on nearly any smartphone I've had.
Except on TV and movies. It annoys me (very mildly, but still) that when actors "talk" on the phone the screen stays lit.
e.g. what are the specific risks? Because all I ever hear is backhanded fearmongering.
If you're really interested, get a copy of Nick Bostrom's book "Superintelligence" from the library. You don't even need to read the whole thing, just the first few chapters will give you a good overview of the risks and how extraordinarily challenging solutions may be.
I'll try to give a summary here, but, really, it's worth investing a couple hours to read the book because I won't do it justice.
The core problem is that there is no plausible reason to expect that artificial intelligence will be limited to the level of human intelligence. On the contrary, it makes sense that once we figure out how to build an artificial general intelligence (AGI -- note, qualitatively completely different from the task-focused ML/AI we have now) that is even slightly smarter than we are, that AGI will be able to use the knowledge we had to discover in designing and building it to design its successor (or redesign itself) to be smarter yet, and so on. The rate at which AGI will increase will be limited only by the rate at which it can design and build (or cause to be built) improved versions.
This is known as the "AI Singularity", the point at which AGI takes over the development of AGI and it explodes out of control, becoming a superintelligence (SI) or group of SIs vastly, vastly smarter than any human has ever been, or could ever be given the limitations of our wetware.
This could be the greatest boon humanity has ever seen, if the superintelligence(s) choose helping humanity as their goal. Assuming, of course, that their definition of "helping" is something that we would agree with. Bostrom points out that even if we are in complete control of the goals of the SI, what goal would we set that could not backfire? Suppose we set the SI the goal of maximizing human happiness. The SI might decide that it can create maximal human happiness by (a) maximizing the number of humans and (b) making each of them maximally happy. To accomplish (b), it might implant electrodes in the pleasure centers of our brains and stimulate them constantly. To accomplish (a) it might choose any of many means of maximizing reproduction, and minimize the cost of maintaining each human life by extracting our brains and placing them in vats, fed by a trickle of nutrients and a trickly of happiness-inducing electricity. It could probably support hundreds of trillions of such brains on Earth by dismantling the entire ecosystem and turning it to this one purpose.
So "maximum happiness" is a bad goal. What's a good one? We don't know. It's a hard problem.
But what if we don't get to set the goal, or what if we inadvertently set it to something really stupid. What if the system in charge of ensuring office supply availability somehow made the jump to AGI, and in order to make itself more effective turned itself into an SI whose sole goal was making paperclips. It might ultimately disassemble the whole planet for paperclip-making resources.
These might seem like ridiculous sci-fi fantasy problems... but the problem is that they're not only plausible, there's nothing at all improbable about them. They depend on only one assumption: That it's possible to build an AGI which is smarter -- even a little bit -- than the smartest humans, and there's no logical basis for refuting that assumption.
Further, it's impossible to know how far we are from knowing how to build such an AGI. We lack the necessary theory of cognition to let us understand what it is, exactly, that makes us intelligent in the way we are, able to think abstractly and problem-solve. Given that we don't have that theory, we can have no idea how far we are from acquiring that theory. It's possible someone had the necessary flash of insight in the shower yesterday. Or last month. It's possible that we're still centuries away (though this seems very unlikely).
Even if we did decide to regulate AI development, it's far fro
“If Jaycee is capable, he can make his own money. If he is not, then he will just be wasting my money.” – Jackie Chan on why he wont be passing down his fortune.
Yep. Bill Gates and many others have done the same. Give your kids all the opportunities you can while they're gowing up, a stable, loving home life, opportunities to learn and grow and do things, excellent healthcare and the best education you can. If after all that, they still can't make their own success, then they shouldn't have it. Not that you should leave them to die in a gutter, if that's what it comes to. It's fine for parents to act as a safety net. But if the kids want to live well, they need to do it themselves.