I agree that the proper metric is if the AI is better than the average human driver. But how do you measure that? If we tested new medicines the way we're currently testing self-driving vehicles, we'd just release new drugs immediately and conduct longitudinal studies to see how many people go blind, go mad, die, etc. after using them. I'm HAPPY about self-driving cars; I'm blind and I'm looking forward to eventually being able to rent/own one. But the way the technology is being deployed today is dangerous and immoral.

Autonomous driving will save thousands of lives and billions in insurance claims, EVENTUALLY. I want to see a transparent testing regime that all self-driving cars must pass before deployment on public roads. What we have now is companies testing their software on the highways and shifting blame to the person behind the wheel if something goes wrong. The companies should be able to test on public roads, but only with their own employees behind the wheel, and only if the companies are fully liable for anything damaged or anyone hurt or killed.

My Apple TV stopped talking to my iTunes library once. The reason was that the debit card associated with my Apple account had changed numbers when all the cards got chipped sometime back. Subsequently, I bought a song off iTunes but the charge was apparently declined. There was no message from iTunes or Apple except to pull the plug on Home Sharing until I figured out what was wrong and paid Apple the $1.09 that I owed.

This is a clever trick by the cops but it won't be effective forever. Neural-net driven filters are _amazingly_ good at removing noise from image and sound streams. If you tell the filter that a particular song is noise and give it a recording to work with, the filter will clean up the sound source well enough that the content ID detectors will no longer trigger. Either the person recording the video will be able to remove the background music in post or a platform like YouTube can do it automatically when the copyrighted material is detected.

Well, the hardware and the truck itself both have non-negligible value even broken down for parts, so an armed guard makes sense if for no other reason than taking the truck should be at least as hard as knocking over a liquor store.

The same answer applies to any other forum, not just an Internet platform. You provide factual information because it is more moral than providing counterfactual information. "Ich soll niemals anders verfahren als so, dass ich auch wollen kÃnne, meine Maxime solle ein allgemeines Gesetz werden." Or "I ought never to act except in such a way that I could also will that my maxim should become a universal law." Kantian universalism. The world would be all kinds of better if people embraced this simple philosophy.

But that's what bittorrent is for, Charlie Brown.

He need not use stock directly. Options let you gamble with much smaller sums, using leverage to make big profits if you guess a large price swing right. In this case he would buy a pile of out-of-the-money puts that would become much more valuable when Tesla stock went through the floor. Of course the SEC would notice the transaction and eventually find his ass, so best be in a non-extradition country by then.

Using it once is all he needs.

1. Take a big short position in Tesla stock.

2. Send out an "update" that bricks Tesla cars everywhere, which would end Tesla as a profitable enterprise.

3. Profit!

If the hacker only got $50K he was robbed.

I think Goldsberry was miscast as Quellcrist Falconer. Michelle Yeoh, yes. Stephany Jacobsen, yes. Olga Kurylenko, even. Goldsberry just doesn't look like she has suffered enough to be Quell.

Be careful about "no symptoms". Some people don't notice the symptoms they have. Remember a few months ago when you couldn't buy a pulse oximeter for any price? Medics discovered that some "asymptomatic" infected people were unconsciously breathing more rapidly and deeply due to diminished lung capacity, and people got the idea of using oximeters as early warning devices. Just because you don't notice symptoms doesn't mean the virus isn't doing bad things to you. The first infection might soften you up enough that the second infection finishes you. There's also the matter of the virus mutating thanks to being passed around in large population.

How are they doing this on modern iOS devices when whole drive encryption is supposedly in use?

You're going to need a lot of water for manned space missions anyway, so why not use it for shielding?

It's more likely that AI will discover a strategy that defeats all humans than the reverse. Machine learning systems have a monomania about their field of study that a normal human can never achieve. The closest human analogs we have are idiot savants. But extreme competence in one area tends to produce gross incompetence in many other areas. Some of these people can't even dress themselves.

... we were supposed to have painless and automated reputation services to steer us clear of the inevitable morass of malware that bazaar-type marketplaces produce. Where are these services?