This is definitely possible, since you can actually controll cars (at least some models) via a (non-public, but the capability is there) API. Two security researchers at defcon were able to find a way how to control a vehicle remotely, even including things like stopping or turning, and eventually made an exploit that could be used remotely to any car of the same model. So, if they wanted to, they were able to stop or turn the wheel of IIRC hundreds of thousands of cars around the world instantly, since the cars are connected to the network through GSM, so you don’t even need to be anywhere near them.

It’s been a few years since I saw the video, but IIRC the vehicle controls are on a separate board that should not be reachable from the other smart vehicle system. However, they were able to reverse engineer a way how to abuse framework update mechanism as a bridge, and use it to patch the framework to get it under their control. And then they discovered that they could actually trigger the update remotely.

I disagree with this. There is one glaring issue with AI-powered weapons, in comparison to other traditional ones - the skill ceiling required to make massive damages at scale.

Sure, you can probably level a whole town if you get your hands on some kind of advanced artillery. But it’s still vastly more complex machine, that probably requires extensive training just to operate. You need an army for that, and army is made of people who will hopefully tell you “No, we’re not doing that”, if your request is not reasonable. And if you somehow try to do it yourself, good luck getting more than a few shots out before someone notices and tries to stop you.

If you have an army of hundreds or thousands of AI powered suicide drones, where you just slap an explosive on them, set a target and the whole fleet will start running, you only need one person with a computer. And once you send the fleet, it’s vastly more difficult to stop it. Hell, you probably don’t event need to physically get to the drones, if you can hack into the system that controls them.

And that’s the biggest issue with any AI-powered weapon, and a reason why they shouldn’t exist.

I don’t think we ever want to find out that P=NP. If I’m not mistaken, any kind of cryptography we have right now is dependent on P != NP.

I’m not sure if it’s even possible to solve cryptography once you have P=NP. Can’t you simply bruteforce anything instantly, once you have non-deterministic turing machine running in polynomial time? Since it just basically guesses the correct output. And that would be a huge worldwide problem. Given how massively expensive in damages is any kind of new zero-day vulnerabilities, such as the recent Confluence one or the Log4J vulnerability, imagine if someone came up with an algorithm to prove P=NP, effectively breaking every kind of security measure and crypto the whole internet builds on. Blockchain would be absolutely wrecked, since you would be able to just almost instantly mine whatever you want, you would be able to fake anyones private certificates used for authentication, and in general the following few days would be a global catastrophe, and a banquet for cybercriminals.

I vaguely remember the term “quantum resistant cryptography”, but I’m not sure if quantum computers equal non-deterministic turing machine, and as I tried quickly googling it, it’s way more complex to understand that I have time or capacity for right now, so I’ll just leave it as an open question.

Oh, I kind of assumed that “P versus NP problem resolved” meant that they somehow proved and solved that that P = NP, or that some NP problems can be solved using P algorithms. Which would be kind of a really big deal, especially from the point of cybersecurity.

But you are right, it may just mean that we proved that P != NP for sure, which wouldn’t really change anything.

lol

2038 - P versus NP millennium problem resolved.

2044 - Blockchain is adopted as the global standard for financial transactions.

Yeah, sure.