Sure there has been ARM64 windows for 5+ years, but it generally has poor reviews - compatibility issues (x64 emulation is only in userland so drivers won't work unless there's a native version), poor performance (mostly caused by emulation) etc.

Linux applications already natively support ARM64, and it runs the same kernel with the same drivers so most of these problems don't exist.

Apple are also in a much better place as they can mandate ARM64 as the only way forward. For Apple developers porting to ARM64 is a hard requirement unless you want to abandon your software, whereas for Windows developers porting to ARM64 is a cost with very little benefit. That's why virtually all mac software is now native ARM64, whereas for Windows it's a very small percentage.

But it's the same with any new technology, the incumbents hate it and either ignore it and hope it goes away, or try to push back against it, stifling progress. Usually this is a kneejerk response from people who are afraid of change and don't even understand the new tech.

Sensible people evaluate the new technology properly and determine how to make the best use of it, while understanding what weaknesses it might have and when its use might not always be the best option.

Anyone who dismisses a new technology without having properly tested and understood it is not worth listening to.

More people do office work than ever before.

This story is so old it was the reason for Unix. Once the platform controls the essential business process, it owns the business. You work for them now, and all your aspirations are moot.

It makes sense in some cases to have a cleaning fee at the end rather than building it in to the nightly cost.

A decent hotel will clean your room every day, an airbnb will not - they will just clean the place after you leave ready for the next guest. Thus the cleaning costs are the same wether you stay 1 night or 1 month.
If you're not getting daily cleanings it doesn't make sense to factor the cost into the nightly rate.

Sent through a physical cable you either receive the original data or not.
Once you are receiving the original data - either through clean transmission, or because error correction has fixed any transmission errors, it doesn't get any better.

At the relatively low bandwidth needed for audio data, there are many protocols and physical carriers capable of transmitting the data and ensuring a perfect copy is received even over low quality cabling.

It's not like you can't use another model to validate the results you got with one model. That's a step you should be taking anyway. It's actually built into the system for test/dev of the agent and inference model itself. Active supervision, sandboxing and all that.

Or just pass the whole machine on to that entity, and let them work out how to hack it... Assuming that such entities don't already have access to such machines.

But you don't use a federal agency where a significant proportion of the employees wouldn't agree with the intended use of the vulnerabilities.
It only took one edward snowden to leak exploits from the NSA, and the NSA as an agency targets foreign countries so most americans wouldn't have a problem with it. Once you had an agency using exploits to subvert the domestic vote you can guarantee that the majority of employees would be against it.

You'd get a voting machine to a friendly location - die hard domestic supporters, or a foreign power that aligns with your goals. And on the subject of foreign powers, it's almost certain that at least the russians and chinese already have access to these machines and have studied them, possibly other countries like iran and north korea too.

So the machines did give fraudulent counts and the claim was correct.

Patching the vulnerabilities in the future wouldn't change the fact that the system was vulnerable last time it was used, and someone else could have independently found and exploited the vulnerabilities at the time.
Likely if vulnerabilities were found they would be used to claim that they were exploited during previous elections, and calling the results into question.

A fork truck moves predictably and has weak points, so yes you probably could beat it. Equally it could easily kill you if it was able to spear you with its forks.
If you were able to get around the forks, or even in between them you could start ripping out the hydraulic lines.

A "contest" requires more than 1 human participant... When those laws were written they likely never considered someone fighting a robot.
Until recently any human fighting a "non human" would have meant fighting an animal, and there are separate animal cruelty laws to cover such scenarios. But there are no laws that forbid cruelty to machines.

A fight against a robot could be considered use of training equipment rather than a contest, although it might fall foul of other laws relating to dangerous equipment in the workplace etc.

Precisely this... In those videos he's fighting the robot as if it was a human opponent, trying to hit it in places like the head where a human would be vulnerable.
The robot looked to have some exposed wiring around the back of the head, in a real fight you'd go for the wiring or joints in order to incapacitate the robot, not repeatedly punch it in the face.

Also a robot designed to kill wouldn't be humanoid or try to fight with punches and kicks, it's more likely to come on wheels, tracks or rotors and be armed with guns or explosives. See the combat drones in Ukraine for example.

It seems so. Although Netcraft confirms it is (ahem) a rounding error.

This post may contain historical references to Microsoft press releases you are not aware of.