Grades are also measures of conformance and memory which are both actually crutches that can reduce the degree of developed intelligence.
It is interesting that the only intelligence required here is the intelligence to know that memorable passwords are a security risk in general and that proper password security requires the use of a well-protected password vault and automatic password generator so that no compromised site will ever reveal the password you've used on any other site.
NT OS kernel is not a name. It's a description. As I remember it, the scuttlebutt at the time was that it was just an avoidance of carrying through or limiting it to the Mach name.
Like I said, in that case, all businessmen, politicians, etc. who have others make calls for them because they don't have the time need to be stopped. Your argument applies just as well to having other people make calls on behalf of a busy person as having AIs make calls on behalf of a busy person.
Most tech is anti-privilege because it brings the costs of privilege down. This is the problem that people have with AIs making calls. Any attempt to slow the propagation of this privilege to the masses should be met with a demand that the few who already exercise it via more expensive means give it up.
A better comparison might be Windows/Mach derivative. As near as I can tell, Microsoft hasn't even bother naming their kernel other than calling it the kernel or the Windows NT kernel which is no different than the "kernel in Windows NT".
If we cast Linux into a similar vein, "Linux" users are actually running "GNU/Gnome", "GNU/KDE", etc. which happen to run on multiple kernels including Linux.
From a practical marketing point of view though, maybe if GNU had come up with a better name, we'd be using it. The fault probably lies with the GNU program for not coming up with a marketable name. In the vacuum, users grabbed for "Linux".
This is why we need to ban mass call campaigns, not the technologies being abused to implement them. But, as long as they are perceived to be of value in political campaigns, we can't seem to get there. There is no technical difficulty in detecting mass call campaigns at the carrier level. The difficulty is in getting politicians to put our desires ahead of theirs.
The difference between a single call made on behalf of a person trying to accomplish a task and a spam email sent out to millions in hopes of having one in a few thousand bite is so vast that the comparison is laughable.
If we want to be upset about providing this to the less privileged, then we should create a law that says that nobody can allow any other entity, human or otherwise, to make a call on their behalf, not one that discriminates against those that can't afford true human assistants by just banning the assistants that fall into the price range of the masses.
Few can afford PAs today. In 1980, the division of the corporation I first worked at had over 200 secretaries helping about 600 engineers and managers. By 1986 when I started, that had been reduced to 4. So, today, their effectiveness in serving the people they served then is about 2%, not 100%. Google Assistant has a potential to be an infinitely better PA in a few years than the PA most of those who had PAs in the old days no longer has today.
No, I'm not. If there are, then we can make other exemptions for useful services.
Frankly, I find the existing recording laws to be highly discriminatory. They give a big advantage to people who have the memory to repeat quotes from conversations months ago verbatim. I don't have this ability and tend to have to force everything to email and messages to compensate.
I consider recordings to be an assistive technology. If someone else doesn't have to ask permission to use their biological memory, why should I have to ask permission to use my assistive memory tools? Restrictions on capability extensions just serve to slow advancement of people not born with these gifts.
I believe it would be better to treat personal recordings as an extension of an individual's memory with all of the same protections. If we do, then other concerns are erased. When consent is not given for public disclosure, they would fall into the hearsay bucket in court - no more valuable in testimony than words from the mouth of the individual who made the recording.
Somehow, I don't think those making complaints about this work phones in retail businesses. The demo was vastly more articulate and had better manners than most human callers to businesses. For those users with heavy accents or ESL users, this would be a nice accessibility feature.
I can see this being useful anywhere an executive used to use a secretary in the old days. For example, the meeting organization functions in calendar programs are rarely used. This could move them to a useful level where the assistant calls the parties involved and negotiates what might work. It could even make the calls simultaneously though a worst schedule first approach might work better.
In some use cases such as verifying store hours, it is going to reduce the numbers of calls the store has to handle. My wife was working in a store on Easter and said that the already overworked staff was answering the phone almost continuously all day to say the store was open. On New Year's Day this year, I called a store to ask that same question. They picked up the phone, said "we're open" without waiting for me to talk, and hung up.
As to it violating "recording" laws, we obviously need to rewrite them. It is ridiculous to require that a person be paid to make these calls. Would a transcription device for a deaf person be considered a "recording device" under these same laws? What happens when we start getting cyber implants to help us remember things?
I'd rather the law move towards treating assistants as personal extensions and giving them all the rights and protections of the person they are extending at the moment.
That's the BFR/BFS plan. 24 hour turnarounds would be groundbreaking enough for now. Getting there a step at a time is much more cost-effective and is working.
And despite releasing six new more natural sounding voices for the Assistant product available today, none approached the humanity of its Duplex example.
There is a big difference in what Google can do and what they can do en masse. You may have also noticed a statement today about efforts in optimizing trained networks. The more complex networks aren't economical to run millions of times a day. Parallel work is under way both in hardware and algorithms to change that.
Duplex is not being deployed today because the compute costs are high enough that it is not yet economical to deploy.
Similarly, the voice we hear from the Assistant differs greatly from their best in-house efforts on those same voices. It is from a more energy optimized model.
An everyday example of this can be self-demonstrated with Google Maps in the driving mode. Some of the voice commands are produced by the server and sound very nice. Others are produced directly on the phone and sound like text to speech engines from previous decades. I'm not sure what the criteria are for using a locally produced direction versus one from the server, perhaps it loses the cell tower for a moment. But, you'll know it when you hear it.
No, the driver is the responsible party. This is very much like a driver's education situation in which there are two sets of controls in the car. Ultimately, the teacher is the responsible party. Any accident is the teacher's fault.
How dark or light the area was is irrelevant to the technology's response. This vehicle was not using the released video for navigation. That was from a camera capturing video for review in incidences like this one.
The vehicle had Lidar which can work fine in complete darkness with the headlights turned off. The sensors spotted this individual and the software ignored what they saw.
However, because the driver is ultimately in control and responsible and this area was lit well enough for a human (though obviously not for a cheap dashcam), this accident was the driver's fault -- no ifs, ands, or buts. Moreover, it appears that the fault was due to gross negligence on his part. Perhaps Uber also has some fault for not adequately monitoring the drivers or hiring people professional enough to be trusted.
The purpose of this testing is to catch problems in the software like this one. If the driver had been performing his job, he would have taken control, the incident would have been noted, and the engineers would have been able to make the necessary adjustments without any loss of life or ever making the news.
Another system in current use for doing similar police work is to make public calls for information that might be helpful to a case or broadcast sketches or grainy videos of suspects and ask for the public to call in. What percentage of those calls are false positives? My bet is it is vastly higher than 92%.
If anything, providing a tool such as this is getting out of the debate. Not having the tool is weighing in on the side of the debate that says users should always be tied to their devices.
Products such as Android should not choose sides in debates. They should deliver the tools necessary to serve all of their customers or at least not lock things down to the point that others can't deliver nicely integrated tools. They are responsible only to serve the needs of, at the least, the vast majority, not to steer those needs.
And, why would there even be a debate on this subject? 50 years ago, phones routinely had switches to turn off the ringers. At work, we had secretaries to filter the calls and allow us to stay focused. The secretary would also remind us if we spent too much time on the phone. At home, parents limited the teenagers' calls by picking up the line and saying it was time to get off the phone. Meals involved people sitting at the table together and discussing their day - not arguing over who gets to use the one home phone during dinner.
We've had a media usage battle for many decades. We've simply lost some control of it because of the independence of today's devices. It wasn't so difficult to control TV usage when most families only had a shared one in the living room. Of course we should have tools to allow those who want to to continue that control.
I dream of the home-based assistant tech - not just simple word and phrase recognition, but true natural language processing across a wide range of skills and the ability to learn skills through verbal or other interactions. But, there are some very difficult to overcome barriers to it.
First, we really need advances in AI that allow it to train itself or be trainable with simple expressions of personal preferences, beliefs, etc. A home-based AI assistant must have home-based training to be truly private. We aren't there yet. Even if we were, learning requires far more AI processing power than simply running pre-trained networks.
For that, we really need a leap in AI processing power. These applications, when looked at from a per user perspective, require intense bursts of complex, real-time operations. Response time is critical for usability. The centralized nature of the processing at this point allows you to employ a few milliseconds a few times a day of processing power from a large machine that you couldn't afford to monopolize as a single user at home. That large machine is required to hit the real-time mark in natural language processing.
I keep hoping that Samsung, Google, NVidia, Intel, AMD, or IBM (I'm not listing Apple because they would require me to buy Apple products) will go seriously out of the box and produce an AI processor that targets a 3 or 4 orders of magnitude density and power improvement over Google's best TPU. But, they have no incentive to do so because the current data center-based model is relatively predictable and profitable.
By "out of the box", I mean we need to open our minds back up to analog vs digital, asynchronous vs synchronous, photons vs electrons, masks vs self-assembly, 2D physical structure vs true 3D (not stacked 2D) or 4D, and even the concept of every device coming off the line being perfect versus just a sea of "processors" (could be nothing other than alterable points in crystals) and tiny connections (which could be virtual determined by wavelengths, phases, timing, etc) awaiting the adaptation of a seed ANN to it followed by learning.
We'd also need device and service developers to adopt a consistent framework for discovery and interaction. They are currently incentivized to be different. Google Home, Amazon Alexa, Siri, Cortana, etc. should not have to have any special "skill" to interact with any device or service. Sadly, we are a long ways from any of this. I feel like the central devices are at a stage of development similar to Windows before the introduction of "plug and play" nearly a quarter century ago. In fact, I've seen indications that Alexa is retracing some steps very similar to the evolution of plug and play right now. Sad that we haven't learned these lessons at the systemic level.
I was once a computer engineer in a major corporation that had unionized electricians. If we got caught (by an electrician) unplugging a piece of equipment, moving it to another place in the lab, and plugging it back in, we could get a grievance. When that happened, an electrician got to sit down at a table on a Saturday and get overtime pay. Some electricians would actually keep records of where things were plugged in to increase their chances of catching someone.
What they did is ultimately good, but how they likely did it???
A citizen / corporation witnessing a crime and reporting it is one thing. Suspecting one and actively investigating it across international boundaries using powerful spying tools that the police might not have been able to use without subpoenas or warrants is very different. Is this not vigilantism?
If such capabilities were automated and deployed en masse to detect patterns instead of investigate specific reported instances, we'd start running into the problem of innocents being investigated. That happens in the justice system too with often devastating results to those whose lives are destroyed for nothing, but not as often as can happen when big data mining is employed.
We shouldn't go there. It fails the risk/reward test.
It depends. Setting up "merchants" can be automated just like sending millions of spams. You just create a script to set them up from a template that is varied by a table in a database. I see "websites" all the time that are obviously just rows in a database created to build 1000s of sites with hours of work just to match a search term now and then and achieve a view on a paid ad.
One thing that usually gives these away is that they aren't paying for good domain names. I won't shop at anything that is xyz.unrelatedname.com and the xyz is the store name or garbage. It would be nice to just filter those out.
True. I'd go a bit further though. When I was a kid and teenager, I rode my bike everywhere and was of course very good at bike riding. After college, I stopped. Twenty years later I got a nice bike for exercise and found that, though I could stay up with my hands on the handlebars, I could no longer ride a bike at all like I did as a teen. I didn't forget the totality of how to ride a bike, but I did forget a lot. How long that really took, I don't know.
I would bet there is a measurable decrease in responsiveness as a driver in a short time period of not driving - probably less than a year.
Yes, the cars need to be able to handle everything ASAP. I suspect this means the AI actually needs to be constrained some to force abstraction but inputs need to be increased. It can't learn every situation, so it must learn to recognize situations that are a danger to the vehicle or others according to the nature of the situation - using a deep understanding of physics and potentials. It's the potentials that are difficult. What is the potential of a deer at the edge of the trees 20 feet from the road? Very different from a dog or cow? What is the potential of a plane descending toward the road in the distance? A visible tornado a mile away?
I have actually encountered a small twin tornado crossing I-70 in St. Louis on 4/14/98. One funnel passed in front of my vehicle and the other behind it at nearly the same time - perhaps a couple hundred yards apart. I saw them when they were perhaps 100 yards from the road and moving fast. I braked a bit to thread my way between them. They went on to trash a furniture store a short distance away. What does LIDAR see with a tornado? A self-driving car might have just stopped in which case the one that passed behind me would have hit me (if not the vehicles behind me).
As far as inputs go, I have noticed that heavily shaded driver's side windows really bother me because I can't see the driver's face at intersections. This led to a realization that drivers routinely intentionally or unintentionally signal intentions or requests. This needs to be processed and everyone should know it is processed through some sort of feedback, otherwise they may not give the signals to self-driving cars. The same is true with pedestrians. There are many ways in which we sense or they signal their intentions.
Also, I believe sound input is vital. I think I first noticed the tornadoes mentioned above by sound. A good 3D microphone array should be one of the inputs. This is vital for detecting emergency vehicles in the distance and comparing to the vehicle's knowledge of the map to detect when it might be necessary to switch lanes in preparation for the possibility of needing to give way to an emergency vehicle.
There are times that they can't see the faded lines in the road or the stop lights? Well, yeh. I agree. Same here. And I don't consider that safe either. There are many different inherently unsafe intersections near me. The only difference is that the self-driving vehicle can avoid taking the chance by turning it over to me to take the chance. I honestly don't know if it would be better if it kept control and tried it than handing control to me. But, the buck is passed.
The biggest problem that autonomous vehicles have to overcome is that they expose the faults in a very bad system through extensive data collection. Exposing faults in something people depend on is never an easy road. The messenger often becomes the victim.
Like most who have lived more than half a century, I've been in a number of accidents including:
During a light drizzling rain, an oncoming vehicle veered into my lane on a sharp turn (cutting the corner). I tried to turn even sharper to run into a yard and avoid the vehicle. My car broke loose and slid semi-sideways into the oncoming vehicle's driver side door. Contact was in my lane. I would not expect an autonomous vehicle to avoid this accident though if the other vehicle with a teenage driver (first year of driving) headed to school had been one of the new self-driving ones, I am confident it would have been avoided.
I was stopped in a two-lane highway with blinker on awaiting the passage of oncoming traffic before making a left turn. A vehicle rear ended my vehicle at full highway speed. The driver was a railroad engineer driving five hours to his home after having driven his train route. The collision woke him up. I am pretty sure that if he had been in a self-driving vehicle, that accident would have been avoided.
While driving into the sun through a parking lot, I drove down a lane that was not a "thru" lane. The only reason it wasn't was because eight inch high concrete stops had been placed across the lane to prevent thru traffic. I suspect a self-driving vehicle would have detected the barriers across the road and that single car accident would have been avoided. If it didn't, it would do no worse than what I did.
While driving through a construction zone at night on a two-lane road with the lanes separated by barrels, a driver from the opposing lanes suddenly attempted an illegal U-turn between the barrels in front of me. I barely had time to even move my foot from the gas to the brake much less stop before swiping across the drivers front end at about 40 mph. Both vehicles were totaled and the mark on my arm from the airbag deployment appears permanent. It is possible that if I had been in a self-driving vehicle that it might have seen her insane turn before I did, but I doubt it. On the other hand, if the distraught grandmother on the way to see her grandchild in the hospital who suddenly realized she had taken the wrong turn off the interstate had been in a self-driving vehicle, she likely would not have even been on that road, much less making an illegal turn.
Driving is dangerous. Four out of four of the accidents above were because of driving while impaired in some fashion - a teenager unprepared to drive in rain, driving while tired, blinded by the sun while driving (should have stopped or greatly slowed), and driving while distraught. And there are numerous other incidents that didn't rise to the level that I would term an accident that would be reported by these self-driving vehicle regulations.
Road maintenance is also atrocious in this country and human drivers die because of it every day. Human drivers are also prone to complain that others shouldn't drive while impaired and then make exceptions for their own needs.
Maybe we should start this conversion by just shining a very bright light on reality - require every new car to be equipped with the sensors to record the reality and disclose every bit of it to the DMV - every solid line crossed, every rolling stop, every time the light is red and we're stil
We failed to give our regular users decent tools to review content and easily find what they're looking for.
That is the real crux of the problem. By "review content" here, he means "search", and he's absolutely right.
Coders use many different words to talk about the same concepts and the search system sucks at taking this into account. Unless you just happen to hit on the right keywords, you can spend an hour searching for an answer that is there, finally post a question about it, and get slammed by some elitist for allegedly not searching. After a few times, you just avoid SO.
People who've learned the particular code of referring to things on SO don't have the problem as much and seem to close their mind to it.
Perhaps SO could have something where you could indicate that results of a search aren't helpful and, after a few times of doing so, it would offer to allow you to pose a request to be pointed to a result for a question or be given help in asking it.
Slashdot Mirror
Since the lowest birth rate is amongst the educated white establishment, you're 100% correct.
Grades are also measures of conformance and memory which are both actually crutches that can reduce the degree of developed intelligence.
It is interesting that the only intelligence required here is the intelligence to know that memorable passwords are a security risk in general and that proper password security requires the use of a well-protected password vault and automatic password generator so that no compromised site will ever reveal the password you've used on any other site.
NT OS kernel is not a name. It's a description. As I remember it, the scuttlebutt at the time was that it was just an avoidance of carrying through or limiting it to the Mach name.
Like I said, in that case, all businessmen, politicians, etc. who have others make calls for them because they don't have the time need to be stopped. Your argument applies just as well to having other people make calls on behalf of a busy person as having AIs make calls on behalf of a busy person.
Most tech is anti-privilege because it brings the costs of privilege down. This is the problem that people have with AIs making calls. Any attempt to slow the propagation of this privilege to the masses should be met with a demand that the few who already exercise it via more expensive means give it up.
A better comparison might be Windows/Mach derivative. As near as I can tell, Microsoft hasn't even bother naming their kernel other than calling it the kernel or the Windows NT kernel which is no different than the "kernel in Windows NT".
If we cast Linux into a similar vein, "Linux" users are actually running "GNU/Gnome", "GNU/KDE", etc. which happen to run on multiple kernels including Linux.
From a practical marketing point of view though, maybe if GNU had come up with a better name, we'd be using it. The fault probably lies with the GNU program for not coming up with a marketable name. In the vacuum, users grabbed for "Linux".
This is why we need to ban mass call campaigns, not the technologies being abused to implement them. But, as long as they are perceived to be of value in political campaigns, we can't seem to get there. There is no technical difficulty in detecting mass call campaigns at the carrier level. The difficulty is in getting politicians to put our desires ahead of theirs.
The difference between a single call made on behalf of a person trying to accomplish a task and a spam email sent out to millions in hopes of having one in a few thousand bite is so vast that the comparison is laughable.
If we want to be upset about providing this to the less privileged, then we should create a law that says that nobody can allow any other entity, human or otherwise, to make a call on their behalf, not one that discriminates against those that can't afford true human assistants by just banning the assistants that fall into the price range of the masses.
Few can afford PAs today. In 1980, the division of the corporation I first worked at had over 200 secretaries helping about 600 engineers and managers. By 1986 when I started, that had been reduced to 4. So, today, their effectiveness in serving the people they served then is about 2%, not 100%. Google Assistant has a potential to be an infinitely better PA in a few years than the PA most of those who had PAs in the old days no longer has today.
No, I'm not. If there are, then we can make other exemptions for useful services.
Frankly, I find the existing recording laws to be highly discriminatory. They give a big advantage to people who have the memory to repeat quotes from conversations months ago verbatim. I don't have this ability and tend to have to force everything to email and messages to compensate.
I consider recordings to be an assistive technology. If someone else doesn't have to ask permission to use their biological memory, why should I have to ask permission to use my assistive memory tools? Restrictions on capability extensions just serve to slow advancement of people not born with these gifts.
I believe it would be better to treat personal recordings as an extension of an individual's memory with all of the same protections. If we do, then other concerns are erased. When consent is not given for public disclosure, they would fall into the hearsay bucket in court - no more valuable in testimony than words from the mouth of the individual who made the recording.
I'm sure he'd agree - even if he hadn't been elected.
Somehow, I don't think those making complaints about this work phones in retail businesses. The demo was vastly more articulate and had better manners than most human callers to businesses. For those users with heavy accents or ESL users, this would be a nice accessibility feature.
I can see this being useful anywhere an executive used to use a secretary in the old days. For example, the meeting organization functions in calendar programs are rarely used. This could move them to a useful level where the assistant calls the parties involved and negotiates what might work. It could even make the calls simultaneously though a worst schedule first approach might work better.
In some use cases such as verifying store hours, it is going to reduce the numbers of calls the store has to handle. My wife was working in a store on Easter and said that the already overworked staff was answering the phone almost continuously all day to say the store was open. On New Year's Day this year, I called a store to ask that same question. They picked up the phone, said "we're open" without waiting for me to talk, and hung up.
As to it violating "recording" laws, we obviously need to rewrite them. It is ridiculous to require that a person be paid to make these calls. Would a transcription device for a deaf person be considered a "recording device" under these same laws? What happens when we start getting cyber implants to help us remember things?
I'd rather the law move towards treating assistants as personal extensions and giving them all the rights and protections of the person they are extending at the moment.
Six more good flights for the manned-flight rating! Go SpaceX! This company exemplifies how to make America great again!
That's the BFR/BFS plan. 24 hour turnarounds would be groundbreaking enough for now. Getting there a step at a time is much more cost-effective and is working.
And despite releasing six new more natural sounding voices for the Assistant product available today, none approached the humanity of its Duplex example.
There is a big difference in what Google can do and what they can do en masse. You may have also noticed a statement today about efforts in optimizing trained networks. The more complex networks aren't economical to run millions of times a day. Parallel work is under way both in hardware and algorithms to change that.
Duplex is not being deployed today because the compute costs are high enough that it is not yet economical to deploy.
Similarly, the voice we hear from the Assistant differs greatly from their best in-house efforts on those same voices. It is from a more energy optimized model.
An everyday example of this can be self-demonstrated with Google Maps in the driving mode. Some of the voice commands are produced by the server and sound very nice. Others are produced directly on the phone and sound like text to speech engines from previous decades. I'm not sure what the criteria are for using a locally produced direction versus one from the server, perhaps it loses the cell tower for a moment. But, you'll know it when you hear it.
No, the driver is the responsible party. This is very much like a driver's education situation in which there are two sets of controls in the car. Ultimately, the teacher is the responsible party. Any accident is the teacher's fault.
How dark or light the area was is irrelevant to the technology's response. This vehicle was not using the released video for navigation. That was from a camera capturing video for review in incidences like this one.
The vehicle had Lidar which can work fine in complete darkness with the headlights turned off. The sensors spotted this individual and the software ignored what they saw.
However, because the driver is ultimately in control and responsible and this area was lit well enough for a human (though obviously not for a cheap dashcam), this accident was the driver's fault -- no ifs, ands, or buts. Moreover, it appears that the fault was due to gross negligence on his part. Perhaps Uber also has some fault for not adequately monitoring the drivers or hiring people professional enough to be trusted.
The purpose of this testing is to catch problems in the software like this one. If the driver had been performing his job, he would have taken control, the incident would have been noted, and the engineers would have been able to make the necessary adjustments without any loss of life or ever making the news.
Another system in current use for doing similar police work is to make public calls for information that might be helpful to a case or broadcast sketches or grainy videos of suspects and ask for the public to call in. What percentage of those calls are false positives? My bet is it is vastly higher than 92%.
If anything, providing a tool such as this is getting out of the debate. Not having the tool is weighing in on the side of the debate that says users should always be tied to their devices.
Products such as Android should not choose sides in debates. They should deliver the tools necessary to serve all of their customers or at least not lock things down to the point that others can't deliver nicely integrated tools. They are responsible only to serve the needs of, at the least, the vast majority, not to steer those needs.
And, why would there even be a debate on this subject? 50 years ago, phones routinely had switches to turn off the ringers. At work, we had secretaries to filter the calls and allow us to stay focused. The secretary would also remind us if we spent too much time on the phone. At home, parents limited the teenagers' calls by picking up the line and saying it was time to get off the phone. Meals involved people sitting at the table together and discussing their day - not arguing over who gets to use the one home phone during dinner.
We've had a media usage battle for many decades. We've simply lost some control of it because of the independence of today's devices. It wasn't so difficult to control TV usage when most families only had a shared one in the living room. Of course we should have tools to allow those who want to to continue that control.
I dream of the home-based assistant tech - not just simple word and phrase recognition, but true natural language processing across a wide range of skills and the ability to learn skills through verbal or other interactions. But, there are some very difficult to overcome barriers to it.
First, we really need advances in AI that allow it to train itself or be trainable with simple expressions of personal preferences, beliefs, etc. A home-based AI assistant must have home-based training to be truly private. We aren't there yet. Even if we were, learning requires far more AI processing power than simply running pre-trained networks.
For that, we really need a leap in AI processing power. These applications, when looked at from a per user perspective, require intense bursts of complex, real-time operations. Response time is critical for usability. The centralized nature of the processing at this point allows you to employ a few milliseconds a few times a day of processing power from a large machine that you couldn't afford to monopolize as a single user at home. That large machine is required to hit the real-time mark in natural language processing.
I keep hoping that Samsung, Google, NVidia, Intel, AMD, or IBM (I'm not listing Apple because they would require me to buy Apple products) will go seriously out of the box and produce an AI processor that targets a 3 or 4 orders of magnitude density and power improvement over Google's best TPU. But, they have no incentive to do so because the current data center-based model is relatively predictable and profitable.
By "out of the box", I mean we need to open our minds back up to analog vs digital, asynchronous vs synchronous, photons vs electrons, masks vs self-assembly, 2D physical structure vs true 3D (not stacked 2D) or 4D, and even the concept of every device coming off the line being perfect versus just a sea of "processors" (could be nothing other than alterable points in crystals) and tiny connections (which could be virtual determined by wavelengths, phases, timing, etc) awaiting the adaptation of a seed ANN to it followed by learning.
We'd also need device and service developers to adopt a consistent framework for discovery and interaction. They are currently incentivized to be different. Google Home, Amazon Alexa, Siri, Cortana, etc. should not have to have any special "skill" to interact with any device or service. Sadly, we are a long ways from any of this. I feel like the central devices are at a stage of development similar to Windows before the introduction of "plug and play" nearly a quarter century ago. In fact, I've seen indications that Alexa is retracing some steps very similar to the evolution of plug and play right now. Sad that we haven't learned these lessons at the systemic level.
I was once a computer engineer in a major corporation that had unionized electricians. If we got caught (by an electrician) unplugging a piece of equipment, moving it to another place in the lab, and plugging it back in, we could get a grievance. When that happened, an electrician got to sit down at a table on a Saturday and get overtime pay. Some electricians would actually keep records of where things were plugged in to increase their chances of catching someone.
What they did is ultimately good, but how they likely did it???
A citizen / corporation witnessing a crime and reporting it is one thing. Suspecting one and actively investigating it across international boundaries using powerful spying tools that the police might not have been able to use without subpoenas or warrants is very different. Is this not vigilantism?
If such capabilities were automated and deployed en masse to detect patterns instead of investigate specific reported instances, we'd start running into the problem of innocents being investigated. That happens in the justice system too with often devastating results to those whose lives are destroyed for nothing, but not as often as can happen when big data mining is employed.
We shouldn't go there. It fails the risk/reward test.
It depends. Setting up "merchants" can be automated just like sending millions of spams. You just create a script to set them up from a template that is varied by a table in a database. I see "websites" all the time that are obviously just rows in a database created to build 1000s of sites with hours of work just to match a search term now and then and achieve a view on a paid ad.
One thing that usually gives these away is that they aren't paying for good domain names. I won't shop at anything that is xyz.unrelatedname.com and the xyz is the store name or garbage. It would be nice to just filter those out.
True. I'd go a bit further though. When I was a kid and teenager, I rode my bike everywhere and was of course very good at bike riding. After college, I stopped. Twenty years later I got a nice bike for exercise and found that, though I could stay up with my hands on the handlebars, I could no longer ride a bike at all like I did as a teen. I didn't forget the totality of how to ride a bike, but I did forget a lot. How long that really took, I don't know.
I would bet there is a measurable decrease in responsiveness as a driver in a short time period of not driving - probably less than a year.
Yes, the cars need to be able to handle everything ASAP. I suspect this means the AI actually needs to be constrained some to force abstraction but inputs need to be increased. It can't learn every situation, so it must learn to recognize situations that are a danger to the vehicle or others according to the nature of the situation - using a deep understanding of physics and potentials. It's the potentials that are difficult. What is the potential of a deer at the edge of the trees 20 feet from the road? Very different from a dog or cow? What is the potential of a plane descending toward the road in the distance? A visible tornado a mile away?
I have actually encountered a small twin tornado crossing I-70 in St. Louis on 4/14/98. One funnel passed in front of my vehicle and the other behind it at nearly the same time - perhaps a couple hundred yards apart. I saw them when they were perhaps 100 yards from the road and moving fast. I braked a bit to thread my way between them. They went on to trash a furniture store a short distance away. What does LIDAR see with a tornado? A self-driving car might have just stopped in which case the one that passed behind me would have hit me (if not the vehicles behind me).
As far as inputs go, I have noticed that heavily shaded driver's side windows really bother me because I can't see the driver's face at intersections. This led to a realization that drivers routinely intentionally or unintentionally signal intentions or requests. This needs to be processed and everyone should know it is processed through some sort of feedback, otherwise they may not give the signals to self-driving cars. The same is true with pedestrians. There are many ways in which we sense or they signal their intentions.
Also, I believe sound input is vital. I think I first noticed the tornadoes mentioned above by sound. A good 3D microphone array should be one of the inputs. This is vital for detecting emergency vehicles in the distance and comparing to the vehicle's knowledge of the map to detect when it might be necessary to switch lanes in preparation for the possibility of needing to give way to an emergency vehicle.
There are times that they can't see the faded lines in the road or the stop lights? Well, yeh. I agree. Same here. And I don't consider that safe either. There are many different inherently unsafe intersections near me. The only difference is that the self-driving vehicle can avoid taking the chance by turning it over to me to take the chance. I honestly don't know if it would be better if it kept control and tried it than handing control to me. But, the buck is passed.
The biggest problem that autonomous vehicles have to overcome is that they expose the faults in a very bad system through extensive data collection. Exposing faults in something people depend on is never an easy road. The messenger often becomes the victim.
Like most who have lived more than half a century, I've been in a number of accidents including:
Driving is dangerous. Four out of four of the accidents above were because of driving while impaired in some fashion - a teenager unprepared to drive in rain, driving while tired, blinded by the sun while driving (should have stopped or greatly slowed), and driving while distraught. And there are numerous other incidents that didn't rise to the level that I would term an accident that would be reported by these self-driving vehicle regulations.
Road maintenance is also atrocious in this country and human drivers die because of it every day. Human drivers are also prone to complain that others shouldn't drive while impaired and then make exceptions for their own needs.
Maybe we should start this conversion by just shining a very bright light on reality - require every new car to be equipped with the sensors to record the reality and disclose every bit of it to the DMV - every solid line crossed, every rolling stop, every time the light is red and we're stil
We failed to give our regular users decent tools to review content and easily find what they're looking for.
That is the real crux of the problem. By "review content" here, he means "search", and he's absolutely right.
Coders use many different words to talk about the same concepts and the search system sucks at taking this into account. Unless you just happen to hit on the right keywords, you can spend an hour searching for an answer that is there, finally post a question about it, and get slammed by some elitist for allegedly not searching. After a few times, you just avoid SO.
People who've learned the particular code of referring to things on SO don't have the problem as much and seem to close their mind to it.
Perhaps SO could have something where you could indicate that results of a search aren't helpful and, after a few times of doing so, it would offer to allow you to pose a request to be pointed to a result for a question or be given help in asking it.