Considering the power of AI, what if some evil genius secretly decided to make something called EvilAI that has the sole purpose of infecting networks and taking over all technology connected to the internet?
Would we all be fucked? Are there even any laws that prohibit him from making such an AI?
We should make NiceAI to fight EvilAI
Motherfucker we'd just invent a different internet or go back to a time before internet.
Seems too specific of a job for a neural network, that guy would have to wait quite a long time for it to bear fruit, if any. On top of that, the surface of attack doesn't get any bigger than it already is, and having efficient malicious attackers with no agenda other than "fuck shit up" means that code occlusion isn't a problem and thus countermeasures wouldn't be hard to deploy, or if extensive damage is done to a service it can be shut down to prevent further infections.
Still, no reason to believe finding vulnerabilities in a system is AI territory, since it'd require extensive knowledge of how process works, and simply brute-forcing it just won't cut it
>>9103120
Ez. You give it a database of known exploits and allow it to try them randomly on test systems until it learns to select a known exploit given the right environment. When it's good enough it can be given a new environment which none of its known exploits will handle, so it learns to combine the exploits until something works. Repeat long enough and you've got Ops described AI. Proof of concept for this was public last year. Look up cgc 2016.
>>9103111
the nice ai
>>9103130
Again, when it comes to system security, brute forcing isn't going to work. Even giving this "AI" full knowledge of all possible exploits of all possible systems that ever existed since the 60s to today, the average sysadmin will notice such intrusions, and good sysadmins already have automated responses for known and already-fixed attack vectors to prevent automated brute-force solutions to exploits in a system.
If you're talking about single entry on a single machine, you're still trying to work your way through a default firewall that may block your connections and blacklist your IP/range if enough reports come back to it.
I must reiterate how RIGHT NOW there are bots that use known exploits all over the world on random IPs, an AI that did that wouldn't be doing anything new or exciting that we haven't already tried or already put in place by ourselves.
>>9103111
Best trips
There is no such thing as AI, even machine learning and all actions performed are within programmer set parameter.
>>9103263
Incorrect
>>9103263
Your brain performs with God set parameters.
>>9103139
>It would only be a matter of time before that IP is blacklisted along with any and all proxies this hypothetical AI might be using. It's kind of like playing whack-a-mole with programming.
>>9103109
Networks don't run code, end devices do.
You'd need to infect the end device's OS...
>>9103109
Worms do this without the need of an AI..
>>9103370
>attention grubbing tripfag makes ignorant post
this checks out
>>9103399
>last year computer engineering
Yeah, I'm ignorant.
And yeah, I'm lonely as fuck, any attention even your attack keeps me from crossing the line.
Whether your "Ai" executes its ANNs on the network devices (i.e. Routers and switches) or on the end devices (pcs phones, etc) it needs permission by the OS to run, unless you manage to execute in ring zero, which requires bypassing the kernel security (*nix, nt). You can still run in user mode, but you won't be able to spread (like worms do).
>>9103386
But worms get stopped. This AI would be unstoppable and would learn to bypass any protections people might think of.
>>9103109
Google rokkos basilisk
Side note:
Dont google rokkos basilisk if you believe everything you read
>>9103523
>rokko's basilisk