> a properly implemented hypervisor
Very hard to do. There so many instructions that can be used to expose a HV. For VM hypervisors, MSRs aren't usually properly populated and all the junk that is exposed through WMI
interesting how these consequences only materialised after cloudflare took the first step and then because a bunch of loud people on twitter started whining
Nobody's upset that the cancel mob on Twitter was allowed to whine for Cloudflare to drop them. We're upset that Cloudflare gave into the whining and obeyed the cancel mob.
I think you are right. I think the unknowns are, how tiny will the script be that commands all the cars into a lake and will it be a cloud hack or a local broadcast hack?
Yep, I've had Google Maps direct me to drive into a wall or an empty field more than a couple times over the years. It's not uncommon for people to get stranded or even killed by blindly following bad GPS directions. The maps are often quite bad in less traveled areas. And these are the non-malicious cases!
Sorry, this bugged me enough to try and find some data:
> It's not uncommon for people to get stranded or even killed by blindly following bad GPS directions.
Google took me to Wikipedia[0], which took me to a conference paper[1]:
In a corpus of about 400 news articles from 2010 onwards (via Lexus Nexus search), they found 52 deaths related to navigation technologies, which accounted for about 25% of the incidents they recorded.
57% of the incidents were collisions; someone running into something due to GPS giving bad directions.
20% total involved being stranded.
That's over ~6 years of US, UK, Canadian & Austrailian news reports.
It may not be uncommon for GPS to kinda suck, but it is _very_ rare for GPS to kill people.
I haven't done any deep research on the topic but know of several specific fatal cases off the top of my head. It's not like someone dies every day, but for every reported case there are probably many that don't get reported as such.
The much more common case is getting stuck and needing a rescue. Google maps is absolutely terrible at dirt roads. It will confidently give you directions that make absolutley no sense once you get away from pavement. It never got me stuck anywhere, but easily could have many times if I had been less cautious. Nowadays I know to ignore those directions in less developed areas.
I think the broader point is that driving navigation tech is getting fairly good at happy path cases but is woefully underdeveloped outside of that.
there were already such bugs before, and my analysis is that even the older ECU cars before the 2000s had such bugs, just nobody bothers to look for them (also ECUs have been causing deaths from bugs but they just assume its the driver's fault). self driving cars will be the next order of magnitude of problems. ECU 1x, smart 10x, self driving 100x.
> In July 2015, IT security researchers announced a severe security flaw assumed to affect every Chrysler vehicle with Uconnect produced from late 2013 to early 2015.[120] It allows hackers to gain access to the car over the Internet, and in the case of a Jeep Cherokee was demonstrated to enable an attacker to take control not just of the radio, A/C, and windshield wipers, but also of the car's steering, brakes and transmission.[120] Chrysler published a patch that car owners can download and install via a USB stick, or have a car dealer install for them.[120]
I’ve had this worry for years of a state level attack via network connected FSD cars. But I’m hardly alone, it was shown in a Fast and Furious movie, so people are thinking of it.
No, the future is to command all self driving cars to immediately accelerate to 100 mph and do not stop for whatever reason no matter what. Pure remote code execution.
mp3 density: 30sec per 1MB (some instrumental music with background). jpg density: 12M pixels per MB (trees and some landscaping). I'd argue music has a lot more information, if we can compare seconds with pixels. Imho, OpenAI didnt do a great job: a small dataset and a limited model.
