Chinese satellite broadband launch rocket breaks up into space junk US Space Command last Friday warned that a Chinese Long March 6a rocket launched on August 6 broke up in orbit and created at least 300 pieces of debris. Chinese state media reported a launch on that day: a mission to launch 18 low Earth orbit broadband-beaming satellites. As The Register reported, Chinese authorities deemed …
Local maps cache + checksummed, cryptographically-signed map updates > must-be-continuously-online-to-work map services
(And what do these self-driving cars do when confronted with a physical reality -- road crash, HGV on fire, a ruptured tanker which has dumped water and live eels across the road, landslide, or cave-in* -- not included in the map data?)
Resilianse, you fools!
* I have encountered all of these listed situations.
This is why we need national regulation around autonomous vehicles before we let things go any further than they already are. They need to be 100% resilient to loss of connectivity as one of the requirements. Apple Maps (and I'm sure Google Maps) support downloading maps inside a given area to be used as an offline cache. It will still update the maps regularly so it captures changes. The data requirements are tiny - I set the largest box I could my iPhone which was in the ballpark of 100,000 km^2 and it required about 2GB. As the car moves then extend the box in the area of travel so there's always at least an hour in any direction. Or check, just download the entire US (or EU) I mean, you can buy phones with 1 TB and more so its hard to argue that's an unreasonable requirement.
When they encounter something that's not in the road data then they have to rely on their sensors and onboard "smarts" to handle it. Or if its a Tesla, just plow into the back of an emergency vehicle stopped on the side of the road at full speed.
Someday, sure, but that's not the case today.
Tesla fanboys will tout claims that Teslas using FSD have a lower accident rate than human driving, but there's a massive selection bias there. People aren't enabling FSD on windy country roads when it is raining, or interstates when there's snow blowing across the lanes, or in areas with a lot of construction. They are enabling it in the areas where it is easy to drive, a nice clear interstate or highway, etc.
Now maybe in bumper to bumper traffic it is better, but that's true for just about every modern car which have cruise control that can set a following distance, so if you get bored or distracted and miss someone slamming on their brakes ahead of you, the car will react. Those are rarely fatal crashes, but they'll ruin your day if you're in one!
They also fail in ways that humans don't, i.e. Teslas running into stopped emergency vehicles on the side of the road at full speed. People would be more accepting of having a car drive for them if it got into the same sort of accidents they do. But in reality it won't get into the accidents they do (because the software does not get drunk, distracted, bored, etc.) but it will get into accidents a human never will. So if someone's wife dies in a crash caused by an autonomous car that wouldn't have happened if she was driving herself, it makes it a lot easier for that person to speak up and say "these things are nowhere near ready". Because it is impossible to know if she might have been in a fatal accident had she been driving last month but the car avoided that circumstance so well no one could have ever known the potential other outcome.
"They won't be perfect, but they might be better at driving safely than the majority of humans."
For some things, yes. The problem is there have been prominent examples of some cars driving flat out into stopped emergency vehicles, not seeing large trailers across a roadway and doing bad things to the former owner as they pass beneath, gore points, K-rails, etc. Type "autonomous fail" into youtube and have a beverage and snacks to hand.
Many years ago I was reporting on the DARPA Grand Challenge, a self-navigated off-road competition and had a chance to talk with most of the teams. They were telling me about considerations that made perfect sense, but weren't the sort of things you'd think about until .... bang! Oops. Humans are far more flexible when it comes to off-nominal situations, but our attention can waiver.
Being in the grey zone of a system working really well, but not quite perfect is still "fail".
It has to be more than "just a map" to use with an autonomous car. GPS is good for around 3m of accuracy and a car will need 10-20cm accuracy. That can be done, but requires differential GPS with error correcting signals being broadcast from a fixed base station that is fairly close. Companies such as Waymo create highly detailed studies of the routes their cars will use and GPS is there for a very rough approximation to add into Lidar, vision and inertial data to make sure the car is where it's supposed to be and not bypass stopped traffic by using the pavements.
I never understood why those mountaineers leave so much junk up there... Have they no scruples at all.
I saw an emission on YT the other day and the top of Everest is covered in junk. It's simply sad to see and I'm not even one of the environmental kind of people.
You don't get much flatter than Flanders - the Netherlands perhaps - but you should see the amount of litter left behind by walkers, bikers, car- and lorry drivers, briefly: people away and sometimes at home. Mountaineers are just as filthy as the rest of humanity. Then there's the filth you don't see ...
Every gram of weight at altitude is exhausting. People die up there. If you want to argue against the commercialism of Everest I’ll 100% agree with you, but if people are there and have a choice of dropping something or dropping dead, they’re going to take the former.
" but if people are there and have a choice of dropping something or dropping dead, they’re going to take the former."
Easily solved. Jack the permit fee way up and run a track that can accommodate electric quads, build more permanent camps with trash and human waste stations. People that want to hike past a certain very low altitude have to pay for somebody else to tidy up. They can also be weighed going up and when they come back and charged a fee or paid for the difference on top of the permit. Want to offset the cost? Hike out a load of trash.
I think a lot of it is historical (maybe up to 10 years ago) and it is only more recently with greater awareness and climate change reducing ice that the scale of the issue is becoming clear.
It was always there, and the philosophy was to just leave anything that had been consumed.
The increasing numbers of people also doing these types of things is creating a massive environmental problem with all the waste.
On Everest Rainbow Valley is not named for picturesque beauty but for the number of colorful winter jackets on the deceased who died on their trek up the mountain.
I agree that keep beautiful natural places like Everest should be kept clean and there are efforts to get climbers to bring their trash back down from the mountain with them.
If they are not reclaiming the dead what hope do they have of cleaning up trash?
Perhaps the drone props can work better in thin air but as it is helicopters cannot reliably work on Everest above Camp 2 / 6,400 meters.
The air is too thin after that for the blades to reliably generate lift.
The fact that Musk is not building a factory in Thailand signals either slowing down global demand, or demand in Thailand itself.
If I recall correctly, cars not built in Thailand itself suffer a massive import duty (I think it's in some cases 100%), so if local demand drops I can see that removing the argument of setting up shop there.
I wonder if there are other EVs manufactured locally. Bangkok could definitely do with them..
"The fact that Musk is not building a factory in Thailand signals either slowing down global demand, or demand in Thailand itself."
More likely Musk felt the Thai authorities hadn't stroked his magnificent ego enough, nor thrown enough money and support his way. Thailand has a population not dissimilar to the UK, and already produces about twice as many cars as we do, and Thailand has a manufacturing presence from many of the global makers.
They'll be better off without Tesla.
Slowing global demand, slowing demand in Thailand or just people deciding that can't put up with Musk any longer?
Tesla lost their shine long ago and Musk's Xitter antics have destroyed what remained of his and its credibility.
No-one wants an expensive, poorly made, crash prone Nazimobile.
Crash proneness: search "Tesla FSD crashes" - you will find that there are many to the point of (finally) prompting regulatory attention. Less so for those turning into unquenchable fires where Tesla is using the same excuse as Microsoft to explain the remarkable volume disparity with other brands: "we're more popular". That's waning rapidly, thanks to Musk, which brings me to ..
.. Nazism, that's regrettable shorthand for 'right wing', and you won't have to search far for that. X and Musk's approach to freedom of speech which translates as 'anything I don't like gets banned' ought to be easy enough to find.
AWS and Australia's Red Cross announced a project that used quantum computing to optimize nurse scheduling
Yeah, "announced a project". Here's what they actually did: used classical techniques for the actual implementation, and did what they're calling a "proof-of-concept" using the QuEra Aquila machine that AWS hosts (see the linked article).
Aquila has 256 physical qubits, except they aren't even that, really. They're 256 trapped rubidium atoms that can be used for physical quantum simulations. There aren't any quantum gates; QuEra say they're working on "digital qubits", which means qubits as the term is properly used. QuEra recently announced that they're aiming for 100 error-corrected qubits in a couple of years.
In other words, this "quantum PoC" is irrelevant to digital quantum computing. Maybe it demonstrates a limited future for analog QC (in this case Hamiltonian simulation), much as D-Wave demonstrated a limited future for adiabatic QC. But on the whole this is the sort of announcement that encourages skepticism about QC.
Sure, demonstrations of new approaches to analog computing are of intellectual interest. (Remember Adelman's DNA-computing work?) But they rarely gain much traction in the market (remember Adelman's ... you get the idea), because there are enormous economies of scale for digital computing, and there always will be, because analog systems are less general.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
