That said, if you did have super AI more advanced than this, it'd be ahead of the technological singularity... which is to say, that you can't extrapolate what we know now to guess what they're capable of doing. Let's wait until the Internet gains consciousness as a super AI, then figure out what happens.
A little nitpick: A super AI would be the trigger for the singularity. An intelligence capable of self-improvement. Not quite after the singularity, but pretty much the edge.
Trigger or not, what happens will be after the singularity. That is, if you can actually create a brilliant super AI, you won't know what it's capable of doing or how it'd act, and only the experts capable of building one would be able to make a decent guess.
Ah, I'm doing the kind of stuff that'll enable someone to do this. Well, simple solution is that that we're extremely far away from AI that could kill people. AI can identify lines and stuff, but nowhere near the same level as the human brain. And even if so, it's easy to confuse them.
I'm extremely curious as to what sort of stuff you're working on, though you don't have to tell if you can't obviously 
I'm working on parts of this on my own, however egotistical that is, just to give myself further insight - it's a demanding hobby, and I have no illusions that I'll beat professional researchers to the prize (especially because I believe that it requires such a broad application of knowledge and engineering, plus ridiculous hardware), but I feel like I've made some leaps of insight of my own that may be of use to others. I just don't know anyone in the field enough that I could put the ideas forward to.
Lol, I'm not building a killer AI on my own. It's not like in sci-fi where one mad scientist creates an army of killer robots from scratch.. there's a lot people don't know and a takes an army of very skilled, very experienced researchers to even get there.
I do things like signal processing, which is used for any kind of robot to recognize faces and objects. Also do things like control systems and electronics, so I can make a good guess of where robotics technology is these days, how fast they can point a weapon, etc.. and it's quite good as long as they can easily recognize the right target. Thesis was based on simulation of emotions. Mom's a psychologist (very good with the neural brain stuff), which helps me associate the current technology with actual biological intelligence. And I work on a killer robot game, so it's fun hobby
I did attend a lecture from the CTO of Raytheon a few days ago. It's a world leader in defense equipment, even had a pretty awesome exoskeleton thing, and they're still far from finding the solution to killer robots. It was very insightful on this sort of thing, though.
And one thing about having a super AI is that doesn't mean that you can even get a killer AI. At best, you'll get a giant brain in a jar. This giant brain might be able to design good robots, but it's still very difficult. If you get it in turrets and combat robots, at best, it's still like a very good marksman with very poor vision. The worst I could imagine it being is if you let it control a fighter plane or a ICBM, or some other weapon that doesn't have to be accurate to be destructive. Even then, it's questionable. People don't spend millions training fighter pilots and risking human lives for no reason.. the human mind is still far superior to the best military grade AI because they can make split-second decisions easily and react much better to unexpected changes.
If you want a short answer.. I'll give you this. With pre-singularity technology, I'd say that robots will, at best, be able to specialize in a type of attack. You might get super-snipers or something. But even then, once a human realizes how they work, they'd be able to create a specialized stealth technique that renders current killer robot models obsolete.
Author
Topic: Killing AI's (Read 19336 times)