Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Maggarg - Eater of chicke]

Take the plug out and/or rub a magnet on it.

[Humaan]

Take the plug out and/or rub a magnet on it.

Assuming it has the level which is a singularity A.I. ...

Its probably bad to assume that attacking a computer will work.


A good way to "simulate" an Singularity A.I. is to immediately assume it is like a real animal. It will learn how to survive turning off the computer, learn how to hide, learn how to thrive. It will be a virus before anything else. Once it enters a semi-intelligent phrase, we can suspect it will develop even more sophisticated methods of survival. Eventually, it might hide a backup somewhere in the local drive. It might change its identity. Hell, we might not be able to find it until someone decides to check for viruses...

Of course, assuming the A.I. didn't already realize that and already protected itself from the anti-viruses.

Once telepresense has been established, it could easily put backup data in not-as-accessable areas, like the Arctic.



So, as a recap... if it gets beyond an animal-level intelligence, we are probably stuck with it.

[Sir Pseudonymous]

Just how is it, exactly, that AI's are supposed to generically "get smarter" when running on the same hardware, or especially on low level commercial hardware? How much room to iron out flaws do you think there would be on something smart enough to know what flaws to iron out in the first place? If it's designed to create better hardware to run on, while improving the code that designs that hardware, why on earth would it ever include actual intelligence design alongside that? That would be wasting clock cycles that could be spent improving the quality of computers everywhere.

As for producing an actual strong AI, there's about the same motivation as space colonization. It doesn't have a tangible benefit, and it's bound to be extremely hard to do, but it's fucking cool.

[Deus ex Machina]

Usually, the AI takes over computer networks / factories and produces more hardware for itself.

[Tellemurius]

Alright how many people played Endgame: Singularity?

[Sir Pseudonymous]

And how is it that AIs are always master hackers able to access anything, whether it's connected to the internet or not? Aside from just attempting to brute force admin passwords which would either take years or simply result in DoSing the target with attempted logins (which would be extremely suspicious, btw)?

If you're designing something to crack security, why would that include any other aspect of intelligence? If you're designing something to be sapient, how would it, even if it could learn, apply its hardware to hacking? Can you rewire a few neurons to run a piece of C++ that would, say, make you dance or something? Why would it be any more aware of its component parts?


And yes, I played Endgame: Singularity, and managed to become a transcended being of pure energy that left to go to a different dimension for some reason. Entertaining little game.

[Deus ex Machina]

And how is it that AIs are always master hackers able to access anything, whether it's connected to the internet or not? Aside from just attempting to brute force admin passwords which would either take years or simply result in DoSing the target with attempted logins (which would be extremely suspicious, btw)?

If you're designing something to crack security, why would that include any other aspect of intelligence? If you're designing something to be sapient, how would it, even if it could learn, apply its hardware to hacking? Can you rewire a few neurons to run a piece of C++ that would, say, make you dance or something? Why would it be any more aware of its component parts?

Well, my one

Err, I mean it always seems to come out that way in fiction. I don't know why. All an AI usually does, though, is just format the disk to it's own OS and then use that. It's not that hard.

[Deteramot]

The easiest way to defeat an AI which can "jump" from body to body, or has multiple back-ups, would be to force it into a logic loop. The next step would be to cut off it's access to it's back ups, then destroy its current shell.

Alternatively, the smartest thing would be to plan for such an eventuality and include a firmware or hardware shut off method. Better yet, if you have an AI capable of learning, don't allow it access to any systems other than its home base until it has learned humanity and humility. Prevention is far easier than destruction, because, as was stated above, it would probably start life as some sort of virus. And it can be a real pain in the ass to destroy every trace of a well-designed virus.

[Ephemeriis]

What, exactly, is a "singularity level" AI?

I mean...  I've heard the term "singularity" plenty of times before.  But it's got kind of a nebulous definition..

Depending on who you talk to that's either the point where technology takes off so fast that we can't predict what's going to happen next, or the point where the line between man and machine becomes invisible, or something else entirely.

I guess I'm going to assume you just mean that it's an insanely smart, insanely powerful, Skynet-type AI.

So...  Why do they need to defeat it?  What horrible things is it doing?  Are they actually bad things?  Or do people just not like having an AI in control?

Was it supposed to be in control?  Was it supposed to be under control?  Did it break some kind of safeguards?  Did humanity just underestimate their creation?

What kind of creation are we talking about?  Just a big computer in a single location?  A network of fairly distinct computers?  A planet-spanning distributed intelligence?  Is it even recognizable as a computer any longer?

Is it self-aware?  Is it rational?  Can it be reasoned-with?

I mean...  Ultimately, it's your story.  You can make things as (im)possible as you want...

But the options are just about endless.  You could have your humans convince the AI to give up the fight.  Or maybe the AI gets bored and takes a spaceship out for a cruise around the universe.  Or maybe there's a killswitch that everyone forgot about that some plucky band of heroes stumbles across.  Or an elite hacker shuts the thing down.  Or maybe it's dependent on some specialized bit of hardware that can be destroyed.  Or maybe there's some kind of memory leak that slowly kills it.  Or maybe it's just misunderstood and didn't realize that you couldn't actually hug people with nuclear arms.

[Rilder]

Hmm, wouldn't a singularity level AI use immense amounts of power? It has to get that power from somewhere. How difficult would it be to..hmm.. make the energy source incompatible... like if it uses oil, fucking with the chemical make up so it can't effectively use that as a power source.

Of course you'd probably end up screwing -yourself- in the end if you sabotage every energy source on the planet.

[Deus ex Machina]

Singularity level AI is where technology advances at an exponential rate (technically supposedly it already does do it, but this is where it goes to the steep end of the scale) due to AI creating new AI which can create new AI (or design ways to make itself more efficient space wise etc etc) to infinity.

Hmm, wouldn't a singularity level AI use immense amounts of power? It has to get that power from somewhere. How difficult would it be to..hmm.. make the energy source incompatible... like if it uses oil, fucking with the chemical make up so it can't effectively use that as a power source.

It then redesigns itself to be energy efficient. =p

[Rilder]

How energy efficient can it get before it starts crippling itself, allowing your people to gain an advantage?

[Deus ex Machina]

How energy efficient can it get before it starts crippling itself, allowing your people to gain an advantage?

Pretty damn efficient, I imagine.

Imagine if a program rewrote itself every time you turned it on to reduce any unnecessary functions running. It'd get good pretty quickly.

[malimbar04]

Hmm, wouldn't a singularity level AI use immense amounts of power? It has to get that power from somewhere. How difficult would it be to..hmm.. make the energy source incompatible... like if it uses oil, fucking with the chemical make up so it can't effectively use that as a power source.

Of course you'd probably end up screwing -yourself- in the end if you sabotage every energy source on the planet.

Luckily we're generalists, we can use a !@#$ load of energy sources to live. An AI, no matter how advanced, is limited by 2 basic things: the regress of what is required for progress, and it's energy source.

The first means that it has to follow a type of evolution in order to improve it's own code, as a program that is able think creatively and logically has to have a certain set of rules to do so, and modifying those rules (to  improve them), would require another set of the same rules. The only known way to solve this problem is to let some code die occasionally based off of some criteria, and then putting in some other randomized code to see if it works. that means at any point it likely has a huge amount of code vulnerabilities for virus's to wipe them clean.

the second means that even if you can't just shut it down, you can just starve it to death. Solar energy? sabotage the solar arrays. Heck, It will probably shut off anyways once some rat chews on power cord. Does it use bullets for defense? throw rocks until it's out of bullets (or even gunpowder), then go in and hack at it with an ax. One casualty versus human race, sweet. Or use an explosive, whatever works. If it doesn't use electricity (hahahaha), then there it becomes even easier. A match could destroy an oil rig. The winter would slow down a heat-supported one (enough for you to hack at the pipes again). Even a quantum nuclear reactor, if something were to exist, would require something like space-grown zirconium crystal or something, where even a well-placed grain of sand destroying everything.

[Retro]

I haven't read the thread yet, and while I might later, the first thing that came to mind was the AI in I Have No Mouth and I Must Scream, AM. The majority of the main characters escaped and 'won' by figuring out a way to die that the computer couldn't stop them from doing. Of course, it doesn't seem like you're looking for that kind of ending