I think that's pretty much a certainty on the business side. Already it's almost certain that a business has a subscription, not periodic license purchases.
I strongly suspect that if they tried to pull that in the consumer space, that will make the PC platform too expensive for people to bother. People will just shrug and move to a platform more oriented toward either Apple or Google's home turf. They won't be happy about it (at first), but if owning that laptop has a monthly fee, that laptop is going to get shelved. While certainly Microsoft wants to make inroads to be a viable vendor in those markets, they would cede the rather significant home court advantage.
No, for consumer space, they'll do whatever it takes to keep that market as alive as possible and windows-centric. They will just spend a lot less money doing so. They might offer some integrated cloud data backup options and such, but they'd screw themselves over if they made it a requirement.
The cost to design a pretty meh chip... well that can be low.
In most cases, you would think a meh chip would not be a good target. In this case, in terms of power efficiency, well in an automotive context, it's all a rounding error. So then the question is whether terrible perfromance would be acceptable, well, in these cases, it's mostly GPU doing the heavy lifting.
The challenge comes in as you try to continue to get open ended devices in a world where you have more and more people locked into the google ecosytem or similar. Sure, different tools for different purposes, but that can cause difficulty when your tool of choice becomes more and more rare in the face of tools you do not like.
The point was to reply to the person saying that this story about chromeos somehow relates to Linux security model. While it does avail itself of certain linux features (SELinux), it's mostly about implementing a very limited sandbox and they can/do pretty much implement that wherever their browser runs. You can pretty much also get the same security by never running anything outside a browser context.
In many cases, sure, you are dealing with a situation where the owner of the device is not the operator of the device, and it's nice to limit them. However for security researchers protecting themselves, they should be able to do it either way.
I don't mind chromebooks, but I am a bit put off by the security community in how they sometimes treat enduser empowerment and their endorsement of ChromeOS rather than a more empowering linux distro reminds me of some negative interactions is all.
As a market and a society, we have allowed them a monopoly, and let it fester. Now, they don't need to even vaguely care about customer sentiment, and instead focus on how to make that relationship as exploitative as possible.
Really, it's about how much it doesn't let you do.
If you are trying to be productive, chromebooks are exceedingly annoying because they are so limited.
This plays well with a lot of security researcher mindset, that would rather see useless computers than tolerate what they could imagine to be a security problem.
Sometimes they find legitimate problems (e.g. Heartbleed), but often the declare some severe CVE for "administrator can do administrator things" sorts of behaviors.
Then they wonder at why when they find a very severe issue and get a lot of credibility, why it goes away in a matter of weeks as they try to open/brand a wave of 'vulnerabilites' that are perfectly actually expected/intended behaviors by the developers and the users of that software.
It really just means they recognize that no matter how well they do or how poorly they do, the role of the Windows OS in the marketplace is pretty well set. They tried massive investment and dramatic changes, they didn't break into new markets like they hoped. They inadvertently pissed off the userbase several times, never made a dent.
So the business call is to just start coasting. don't waste money trying to grow, and don't think that continued innovation is really required to hold on to their market. For their future evaluation of potential products, the question is easy: can this product be rented instead of purchased? If they can't figure out a way to rent the product or use the product to drive people to rent something, they would rather direct their efforts elsewhere.
In the commercial software space, there is a great deal of pressure to migrate/establish userbase to cloud customers. Not because it is a channel to deliver capabilities otherwise impossible, but to transform your customers from transactional to recurring revenue.
In the transactional situation, you have to find some compelling motivation to drive users to conduct another transaction with you. Rpoblem is at some point you pretty much "finish" your vision (office 97 can achieve largely the same results that current office can do) and further tweaks to your product aren't enough to drive revenue. Making companies live in fear of being end of support and the treadmill of office format compatibility can carry MS far, but nothing beats making the software evaporate when the periodic payment doesn't happen.
Note that with Office 365, the use of the internet can be quite low, you install it locally and can ignore online mostly, except it checks subscrption status, but that's not much bandwidth.
They can also, frankly, take their eye off the ball with Windows. It's market situation is pretty well set. They've tried to expand it and failed utterly (Windows 8 strategy of screwing the desktop users to try to drag them into liking a phone-capable UI), they've tried to move the application publishing norm to an Apple/Google like one (which is also a failure) and tried to make Windows lock users into the store to make it more appealing (S edition, which did nothing). At the same time, even as they obviously pissed off customers (Vista, 8), their market share was steady, so it's also the case they can't lose more than they have.
So simplify their support burden as much as they can (you *will* be updated to the most recent 6-ish month release, because it's the same) and coast. Have to continue to do sustaining investment as it is their foundation for other services, but growth investment is a lost cause (can't get beyond 95% share, wouldn't be worth it if they could, and can't leverage the platform to break into more markets).
I would say the data is too noisy to be that confident the trend is precisely as described. For example August looked pretty good, then september was suddenly way worse. February looked as good as October, but April was horrific.
It is nice to review the data in more detail. It affirms that it's pretty solid on highways, less so on streets. It also shows that the most prominent intervention is user taking over due to an unwanted maneuver by the car. It shows that early on, almost all the events were human taking over without the system, and August/September had an interesting run of the software disengaging itself for some reason.
Of course, this could make for an unacceptable taxi experience. Your taxi gives up a noticable period of time isn't going to be a particularly happy experience. If it otherwise works as advertised, it might make for a solid mode for your personal car or a rental, but not really 'taxi'.
Conversely, throw out older cars. Of the accidents I've seen actually occur, one of them was because an axle snapped and the car completely lost it. If measuring by serious injury, only twice in my experience was an accident so severe it warranted an ambulance for anyone, and that's because that car was older than airbags (the other was not wearing a seatbelt). Also throw out bad weather conditions,country and mountain roads, etc.
Modern cars are increasingly having things like collision alert and lane departure alerts. It is possible that a system that requires the driver to be fully engaged to operate, but also always double checking the human driver is the safest of all the options. It sucks because I'd love to be carted around, but it is an eventuality I must be prepared for.
Would be interesting the percentage of urgent interventions versus non urgent.
I imagine an intervention can be scenarios like car not slowing down despite a pedestrian clearly about to get into the roadway, which would trigger something rather urgent and would indicate at least an increased likelihood of an accident.
I am also imagining scenarios like the coast is clear, but the car refuses to execute some maneuver. Then the human driver intervenes and does that left turn or whatever.
Either way, it speaks to not being able to be a taxi, at least not for people who can't drive. While they may be confident based on some histrocial imporvement in intervention rates, this is the sort of statistic that tends to come down quickly at first, and then plateauing short of your goal and eternally stuck at some unacceptable number.
I personally am worried about the "never allow your customers to own anything ever" philosophy of the tech industry as of late, so "autonomous driving as only for taxi" greatly bothers me anyway.
For Windows 10, the telemetry is unacceptable. Apart from that, it acts ok (I of course prefer other desktop environments).
The update situation is ugly, however. It's all 'Windows 10', but are you on 1507? 1803? LSTB? CBB? A device that was supported for Windows 10 (1507) might not work with Windows 10 (1803). You may still get updates to Windows 10 1507, but it won't act the same as windows 10 1803 and if you try to reinstall it you might fail because your windows 10 media isn't the same windows 10 as you need.
This is pretty much a branding failure more than a technical failure, but there are technical implications of pretending things are simpler than they are.
I think in that scenario, there may be differences to be had, but there's much less chance a consumer will look at far-away scenery and think a marginal increase in that is worth a couple hundred more dollars.
Contrasted with lighting/reflection/refraction improvements, which can be pretty dramatic.
That was simply my point, that going further may be possible without pulling in some raytracing, but the subjective impression isn't going to differ nearly as much.
Seems like an oversight of the VSC requirements. If the goal is to preserve the race state and pit stops are somehow exempted, then that seems like a loophole.
It shouldn't be 'guessing' what the pit time is going to be to slow down for, it should be some mandatory amount.
Demos of real time raytracing have dated back to 2009 or earlier, albeit with various limitations. Raster based rendering is faster and going to raytracing means much better lighting, at the expense of some resolution/geometric detail/hardware requirements.
I think what is being seen is that we've been well beyond the point of diminishing returns as far as raster can reasonably get in terms of better quality. Sure, we can cram more and more polygons and sure we can raise the bar to 4k resolution, but the bang for the back is small. Given that situation, video card market has an issue, they need to do demand generation.
So pushing 4k for gaming helps, and 8k would be nice, but it's just really hard to tell the difference at this point.
If that's a hard sell, then VR certainly can knock things back if it gets traction. With wider FOV, stereoscopic rendering, and optimal experience being at least 90Hz, that would certainly deliver. However, as much as I am a fan of it, it's far from a given that VR is ever going to be large enough to drive adoption. at volumes that can sate the business needs of the GPU vendors.
So raytracing is a third option to make best of breed graphics card noticeably struggle with something that's very visually apparent. Suddenly the market content with status quo of ever refined raster graphics simply must make the leap to have some marketable advancement.
The people pessimistic about any such advancement just continuously have their expectations calibrated to how fast it can perform raster graphics. Raytracing does mean having to step back, but we have enough headroom to take the hit.
Privatized space companies seem to have been working out ok, at least better than I thought it would.
This may end up being a bust for now at least, but this specific article suggests that Uber isn't good at this compared to others, and this deficiency may have cost a human life and thrown a huge roadblock for the entire industry. Other companies may have well been doing well enough, though it of course may be the case things are far worse than they imagine. Certainly I feel the reporting is more enthusiastic than the state of the technology (reporting seems to suggest go anywhere is right around the corner, as yet the trials are running in very favorable conditions only).
When Uber started as a 'ride sharing app', ostensibly helping people coordinate carpooling where they were going to be going anyway, it was fine.
When it became "a taxi, but paying drivers less and trying to get out of the same regulations for no other reason than somehow being 'cooler' than taxi companies", a lot of deserved criticism came about.
Prior to the early 90s, you were probably stuck using a compute a few years old on average, because they simply cost way too much to refresh that often. Pace of advancement was actually relatively slow.
From about 95 to ~2010, they were releasing crazy better specs every year *and* relatively cheap, so you'd rarely find yourself with a computer more than 2 years old.
From 2010 onward, they were still cheap, but no longer was the delta that big a deal for desktop computing and/or focused more on reducing size and power than on performance increases.
They also found that Autopilot reduces the rate of accidents.
To be fair, they make that statement by comparing autopilot miles, which are self-selecting for ideal conditions on good roads, in cars that are only a couple years old at most, to general miles driven, on any road in any weather, by any beat up car with bald tires and no brake pads.
There has not been an effort to, say, compare full autonomous driving scenario to a human operated car, but limited to newer cars on nice highways and with auto-braking and lane-keeping systems (that only correct in exceptional cases, forcing the human to pay attention most of the time, only firing when it thinks the human is missing something or being too slow). It is highly probably that a mixed system where humans simply can't be completely checked out is the safest approach.
However it would be relatively crappier compared to a full-autonomous system, in terms of convenience.
Of the autonomous or near-autonomous systems, I'd much rather the likes of Tesla, Ford, GM, et al get it than Uber/Waymo. The latter would foist a rental model upon the world, and I'd like the option to own to continue to be a prevalent option.
The whole concept of 'this is good enough to do it all for you, except pay attention, just in case' is deeply flawed. If not normally engaged, a human's attention will of course drift.
Unless the horn honked based on LIDAR (which is reasonable), the person could have taken no action in time. Agreed about the car not slowing down at all, I would have expected it to react, though I do not think a human could have reacted fast enough to even think about slowing down.
Slashdot Mirror
I think that's pretty much a certainty on the business side. Already it's almost certain that a business has a subscription, not periodic license purchases.
I strongly suspect that if they tried to pull that in the consumer space, that will make the PC platform too expensive for people to bother. People will just shrug and move to a platform more oriented toward either Apple or Google's home turf. They won't be happy about it (at first), but if owning that laptop has a monthly fee, that laptop is going to get shelved. While certainly Microsoft wants to make inroads to be a viable vendor in those markets, they would cede the rather significant home court advantage.
No, for consumer space, they'll do whatever it takes to keep that market as alive as possible and windows-centric. They will just spend a lot less money doing so. They might offer some integrated cloud data backup options and such, but they'd screw themselves over if they made it a requirement.
The cost to engineer a very good chip is high.
The cost to design a pretty meh chip... well that can be low.
In most cases, you would think a meh chip would not be a good target. In this case, in terms of power efficiency, well in an automotive context, it's all a rounding error. So then the question is whether terrible perfromance would be acceptable, well, in these cases, it's mostly GPU doing the heavy lifting.
The challenge comes in as you try to continue to get open ended devices in a world where you have more and more people locked into the google ecosytem or similar. Sure, different tools for different purposes, but that can cause difficulty when your tool of choice becomes more and more rare in the face of tools you do not like.
The point was to reply to the person saying that this story about chromeos somehow relates to Linux security model. While it does avail itself of certain linux features (SELinux), it's mostly about implementing a very limited sandbox and they can/do pretty much implement that wherever their browser runs. You can pretty much also get the same security by never running anything outside a browser context.
In many cases, sure, you are dealing with a situation where the owner of the device is not the operator of the device, and it's nice to limit them. However for security researchers protecting themselves, they should be able to do it either way.
I don't mind chromebooks, but I am a bit put off by the security community in how they sometimes treat enduser empowerment and their endorsement of ChromeOS rather than a more empowering linux distro reminds me of some negative interactions is all.
To be hated, yet still take their money.
As a market and a society, we have allowed them a monopoly, and let it fester. Now, they don't need to even vaguely care about customer sentiment, and instead focus on how to make that relationship as exploitative as possible.
Really, it's about how much it doesn't let you do.
If you are trying to be productive, chromebooks are exceedingly annoying because they are so limited.
This plays well with a lot of security researcher mindset, that would rather see useless computers than tolerate what they could imagine to be a security problem.
Sometimes they find legitimate problems (e.g. Heartbleed), but often the declare some severe CVE for "administrator can do administrator things" sorts of behaviors.
Then they wonder at why when they find a very severe issue and get a lot of credibility, why it goes away in a matter of weeks as they try to open/brand a wave of 'vulnerabilites' that are perfectly actually expected/intended behaviors by the developers and the users of that software.
It really just means they recognize that no matter how well they do or how poorly they do, the role of the Windows OS in the marketplace is pretty well set. They tried massive investment and dramatic changes, they didn't break into new markets like they hoped. They inadvertently pissed off the userbase several times, never made a dent.
So the business call is to just start coasting. don't waste money trying to grow, and don't think that continued innovation is really required to hold on to their market. For their future evaluation of potential products, the question is easy: can this product be rented instead of purchased? If they can't figure out a way to rent the product or use the product to drive people to rent something, they would rather direct their efforts elsewhere.
Grudgingly accepted is good enough for their bottom line to be happy, sadly.
In the commercial software space, there is a great deal of pressure to migrate/establish userbase to cloud customers. Not because it is a channel to deliver capabilities otherwise impossible, but to transform your customers from transactional to recurring revenue.
In the transactional situation, you have to find some compelling motivation to drive users to conduct another transaction with you. Rpoblem is at some point you pretty much "finish" your vision (office 97 can achieve largely the same results that current office can do) and further tweaks to your product aren't enough to drive revenue. Making companies live in fear of being end of support and the treadmill of office format compatibility can carry MS far, but nothing beats making the software evaporate when the periodic payment doesn't happen.
Note that with Office 365, the use of the internet can be quite low, you install it locally and can ignore online mostly, except it checks subscrption status, but that's not much bandwidth.
They can also, frankly, take their eye off the ball with Windows. It's market situation is pretty well set. They've tried to expand it and failed utterly (Windows 8 strategy of screwing the desktop users to try to drag them into liking a phone-capable UI), they've tried to move the application publishing norm to an Apple/Google like one (which is also a failure) and tried to make Windows lock users into the store to make it more appealing (S edition, which did nothing). At the same time, even as they obviously pissed off customers (Vista, 8), their market share was steady, so it's also the case they can't lose more than they have.
So simplify their support burden as much as they can (you *will* be updated to the most recent 6-ish month release, because it's the same) and coast. Have to continue to do sustaining investment as it is their foundation for other services, but growth investment is a lost cause (can't get beyond 95% share, wouldn't be worth it if they could, and can't leverage the platform to break into more markets).
I would say the data is too noisy to be that confident the trend is precisely as described. For example August looked pretty good, then september was suddenly way worse. February looked as good as October, but April was horrific.
It is nice to review the data in more detail. It affirms that it's pretty solid on highways, less so on streets. It also shows that the most prominent intervention is user taking over due to an unwanted maneuver by the car. It shows that early on, almost all the events were human taking over without the system, and August/September had an interesting run of the software disengaging itself for some reason.
Of course, this could make for an unacceptable taxi experience. Your taxi gives up a noticable period of time isn't going to be a particularly happy experience. If it otherwise works as advertised, it might make for a solid mode for your personal car or a rental, but not really 'taxi'.
Conversely, throw out older cars. Of the accidents I've seen actually occur, one of them was because an axle snapped and the car completely lost it. If measuring by serious injury, only twice in my experience was an accident so severe it warranted an ambulance for anyone, and that's because that car was older than airbags (the other was not wearing a seatbelt). Also throw out bad weather conditions,country and mountain roads, etc.
Modern cars are increasingly having things like collision alert and lane departure alerts. It is possible that a system that requires the driver to be fully engaged to operate, but also always double checking the human driver is the safest of all the options. It sucks because I'd love to be carted around, but it is an eventuality I must be prepared for.
Would be interesting the percentage of urgent interventions versus non urgent.
I imagine an intervention can be scenarios like car not slowing down despite a pedestrian clearly about to get into the roadway, which would trigger something rather urgent and would indicate at least an increased likelihood of an accident.
I am also imagining scenarios like the coast is clear, but the car refuses to execute some maneuver. Then the human driver intervenes and does that left turn or whatever.
Either way, it speaks to not being able to be a taxi, at least not for people who can't drive. While they may be confident based on some histrocial imporvement in intervention rates, this is the sort of statistic that tends to come down quickly at first, and then plateauing short of your goal and eternally stuck at some unacceptable number.
I personally am worried about the "never allow your customers to own anything ever" philosophy of the tech industry as of late, so "autonomous driving as only for taxi" greatly bothers me anyway.
Well, also Windows 8/8.1.
For Windows 10, the telemetry is unacceptable. Apart from that, it acts ok (I of course prefer other desktop environments).
The update situation is ugly, however. It's all 'Windows 10', but are you on 1507? 1803? LSTB? CBB? A device that was supported for Windows 10 (1507) might not work with Windows 10 (1803). You may still get updates to Windows 10 1507, but it won't act the same as windows 10 1803 and if you try to reinstall it you might fail because your windows 10 media isn't the same windows 10 as you need.
This is pretty much a branding failure more than a technical failure, but there are technical implications of pretending things are simpler than they are.
I think in that scenario, there may be differences to be had, but there's much less chance a consumer will look at far-away scenery and think a marginal increase in that is worth a couple hundred more dollars.
Contrasted with lighting/reflection/refraction improvements, which can be pretty dramatic.
That was simply my point, that going further may be possible without pulling in some raytracing, but the subjective impression isn't going to differ nearly as much.
Well it is an efficient pit crew, but *also* the fact that you can pass if you go through pit.
Seems like an oversight of the VSC requirements. If the goal is to preserve the race state and pit stops are somehow exempted, then that seems like a loophole.
It shouldn't be 'guessing' what the pit time is going to be to slow down for, it should be some mandatory amount.
Demos of real time raytracing have dated back to 2009 or earlier, albeit with various limitations. Raster based rendering is faster and going to raytracing means much better lighting, at the expense of some resolution/geometric detail/hardware requirements.
I think what is being seen is that we've been well beyond the point of diminishing returns as far as raster can reasonably get in terms of better quality. Sure, we can cram more and more polygons and sure we can raise the bar to 4k resolution, but the bang for the back is small. Given that situation, video card market has an issue, they need to do demand generation.
So pushing 4k for gaming helps, and 8k would be nice, but it's just really hard to tell the difference at this point.
If that's a hard sell, then VR certainly can knock things back if it gets traction. With wider FOV, stereoscopic rendering, and optimal experience being at least 90Hz, that would certainly deliver. However, as much as I am a fan of it, it's far from a given that VR is ever going to be large enough to drive adoption. at volumes that can sate the business needs of the GPU vendors.
So raytracing is a third option to make best of breed graphics card noticeably struggle with something that's very visually apparent. Suddenly the market content with status quo of ever refined raster graphics simply must make the leap to have some marketable advancement.
The people pessimistic about any such advancement just continuously have their expectations calibrated to how fast it can perform raster graphics. Raytracing does mean having to step back, but we have enough headroom to take the hit.
Relatively less success
Privatized space companies seem to have been working out ok, at least better than I thought it would.
This may end up being a bust for now at least, but this specific article suggests that Uber isn't good at this compared to others, and this deficiency may have cost a human life and thrown a huge roadblock for the entire industry. Other companies may have well been doing well enough, though it of course may be the case things are far worse than they imagine. Certainly I feel the reporting is more enthusiastic than the state of the technology (reporting seems to suggest go anywhere is right around the corner, as yet the trials are running in very favorable conditions only).
When Uber started as a 'ride sharing app', ostensibly helping people coordinate carpooling where they were going to be going anyway, it was fine.
When it became "a taxi, but paying drivers less and trying to get out of the same regulations for no other reason than somehow being 'cooler' than taxi companies", a lot of deserved criticism came about.
There was a curve...
Prior to the early 90s, you were probably stuck using a compute a few years old on average, because they simply cost way too much to refresh that often. Pace of advancement was actually relatively slow.
From about 95 to ~2010, they were releasing crazy better specs every year *and* relatively cheap, so you'd rarely find yourself with a computer more than 2 years old.
From 2010 onward, they were still cheap, but no longer was the delta that big a deal for desktop computing and/or focused more on reducing size and power than on performance increases.
They also found that Autopilot reduces the rate of accidents.
To be fair, they make that statement by comparing autopilot miles, which are self-selecting for ideal conditions on good roads, in cars that are only a couple years old at most, to general miles driven, on any road in any weather, by any beat up car with bald tires and no brake pads.
There has not been an effort to, say, compare full autonomous driving scenario to a human operated car, but limited to newer cars on nice highways and with auto-braking and lane-keeping systems (that only correct in exceptional cases, forcing the human to pay attention most of the time, only firing when it thinks the human is missing something or being too slow). It is highly probably that a mixed system where humans simply can't be completely checked out is the safest approach.
However it would be relatively crappier compared to a full-autonomous system, in terms of convenience.
Of the autonomous or near-autonomous systems, I'd much rather the likes of Tesla, Ford, GM, et al get it than Uber/Waymo. The latter would foist a rental model upon the world, and I'd like the option to own to continue to be a prevalent option.
The whole concept of 'this is good enough to do it all for you, except pay attention, just in case' is deeply flawed. If not normally engaged, a human's attention will of course drift.
Unless the horn honked based on LIDAR (which is reasonable), the person could have taken no action in time. Agreed about the car not slowing down at all, I would have expected it to react, though I do not think a human could have reacted fast enough to even think about slowing down.