Thread: Consciousness and AI

Originally Posted by Xaqaria

The internet.

*nod*

If we put our AI robot in the same room with Charlie Sheen would it become a douchebag?

Originally Posted by sloth

*nod*

If we put our AI robot in the same room with Charlie Sheen would it become a douchebag?

douchebags have a function.

Hackers are able to control nuclear powerplants?

Originally Posted by Mario92

Why do people always assume an AI will be hooked up to every machine gun, nuclear plant, and robotics factory on the planet?

Don't forget the deadly neurotoxin emitters.

Originally Posted by Philosopher8659

douchebags have a function.

Charlie Sheen = Douchebag
Douchebag has function
Tools have function
Charlie Sheen = Tool?

Originally Posted by Xei

Hackers are able to control nuclear powerplants?

Most hackers are not able to improve their own intelligence, make millions of calculations per second or stay online attempting to gain access 24/7. I don't know if a true AI would be able to do these things either but the chances are a little better than your average hacker pulling off similar feats.

Originally Posted by Xaqaria

Most hackers are not able to improve their own intelligence, make millions of calculations per second or stay online attempting to gain access 24/7. I don't know if a true AI would be able to do these things either but the chances are a little better than your average hacker pulling off similar feats.

Uh, I'm pretty sure reactor control mechanisms aren't hooked up to the internet in the first place.

Originally Posted by Mario92

Uh, I'm pretty sure reactor control mechanisms aren't hooked up to the internet in the first place.

Which is one of the main reasons for the uncertainty expressed in the post you quoted. You listed nuclear plants with all those other things, not me.

Obviously though, there are some dangerous things that an AI could potentially access through the internet and there are other things that are not connected and therefore it would not be able to access.

Originally Posted by Xaqaria

Most hackers are not able to improve their own intelligence, make millions of calculations per second or stay online attempting to gain access 24/7. I don't know if a true AI would be able to do these things either but the chances are a little better than your average hacker pulling off similar feats.

Hackers may not be able to improve their brainpower, but all it takes to hack is to keep up with the trends in computing and know how computer systems work. You'd be shocked at the amount of teenagers who have the ability to penetrate high security computer systems by spending large amounts of time finding small flaws in software and developing zero day exploits. Shit, the majority of self-proclaimed hackers don't even know how the exploits work; they just find them online and wreak havoc with them. Some teenagers have even been able to access secure offshore information concerning nuclear program intelligence. They gained access to officials' email addresses, all of their emails, classified documents and project information. With some planning they could have taken it even further. If a teenager can do that, there's no telling what an AI bot could do, given enough intelligence. It would certainly be more destructive than compromising a few nuclear programs, though. I can't say for sure that an AI bot would ever go to these lengths, but you can bet your ass it could if it ever needed to.

Originally Posted by Mario92

Uh, I'm pretty sure reactor control mechanisms aren't hooked up to the internet in the first place.

I'm sure you don't know the first thing about how nuclear facilities operate. I would only assume that the facilities operate alongside a DMZ, which could potentially be penetrated. There are also ways to gain internal access to facilities without even requiring any internet access, such as flash drives and other media. Even offering bribes could get people to plant worms.

You could put the virus on an employers mobile phone. And then set it to copy over to their network when they connect to the internet in the plant. If they even have it there. Not sure if they do. But point is you can do it. And a super intelligent AI could do it even more easily, if it so desired.

There are no computers that are hooked up to the internet that are immune to being hacked into. If it's on the net, it's vulnerable. Luckily, defense networks, nuclear power control systems... are on private networks. You can't hack into those remotely. That's not to say that someday the machines will overthrow us, it's just not any time soon.

Originally Posted by ninja9578

There are no computers that are hooked up to the internet that are immune to being hacked into. If it's on the net, it's vulnerable. Luckily, defense networks, nuclear power control systems... are on private networks. You can't hack into those remotely. That's not to say that someday the machines will overthrow us, it's just not any time soon.

Private networks can be hacked into without outside internet access, which is why I briefly listed some methods that can be used to do so. But yes, I think we can all agree that an artificially intelligent supercomputer isn't going to overthrow humanity and destroy us all anytime soon.

Originally Posted by Xaqaria

Most hackers are not able to improve their own intelligence, make millions of calculations per second or stay online attempting to gain access 24/7. I don't know if a true AI would be able to do these things either but the chances are a little better than your average hacker pulling off similar feats.

My point was that no sane person would connect a nuke to the internet in the first place.

Originally Posted by Xei

My point was that no sane person would connect a nuke to the internet in the first place.

And your point was proven wrong by several people already.

Well, I wouldn't say it was proven wrong, just that even if a nuke weren't hooked up to an internet an artificially intelligent bot would still be able to activate it. However there is a chance that there is an open, hackable link from nukes to the internet.

Woops I misread that post. Most likely a sane person would not hook a nuke up to the internet. But yes, it's still possible for an AI or even a human to hack in to it without the internet.

Yeah I was only clarifying my point which still stands. MindGames showed it could be possible by more general means, which is the important point really.

Take the example of a chess-playing robot. A robot programmed to play chess would respond to its opponents moves while at the same time choosing from a range of possible moves it can make. We automatically assume that the robot is not conscious because it's decision making process depends on how it was programmed (to assess all the possible moves it can make and choosing the best one) but couldn't we say that humans use the same decision making process?

A computer is programmed with a range of symbols and rules that it follows in order to produce output. As language users, humans use the same process to output information in a coherent way so that other humans can understand.

I think Dub mentioned before me that a computer operates on a binary sequence of on/off switches which in turn affect other on/off switches that form a coherent output. The human brain operates in a similar way but with neurons. I'm not sure yet if AI can become self-conscious or not but I think the similarities (at least in function) between the mind and computers are closer than we imagined when we first invented the computer.

Also if you want to give a Turning test to AI go to cleverbot.com

I don't recall saying that.

Originally Posted by Xaqaria

The main issue from a hardware standpoint is that our computers utilize binary logic gates (this or that) while neurons fire in much more complex structures (if this then that and that and that and...) and that only covers the electrical aspect of how neurons communicate. There are even still more levels of communication within the chemical transmissions and then the electrical signals also release chemical transmissions and the chemical transmissions cause electrical signals to fire...

Ahh sorry it was Xaqaria

Cleverbot is a terrible example of serious AI.

Yeah it's shit. And it's called a Turing test. And it would also fail because you know it's a bot.

Most researchers who are serious about AI don't even try to pass the Turing test, they develop AI which is very good at specific applications.

Originally Posted by tommo

And it would also fail because you know it's a bot.

Explain..? Probably just misunderstanding you.

I just mean you can't do a Turing test when you know you're talking to a bot.
The point of it is to see if you can tell if you're talking to a human or an AI.