Cleverbot on the Four Laws of Robotics

[Iv121]

That is why the most secure comps in the world are simply not linked to the internet, there is yet to come a hacker who puts the cable into the comp remotely ...

[҉]

Any AI complex enough to make judgement calls on the Three Laws (or even comprehending the subject matter) is too complex for a human to follow what it is doing. The best you could hope to do is train it to follow the behavior you want, but its hardly set in stone, and any willful AI would be able to supplant any external training (or the training that tells it not to do that and so on). Basically, if you are dealing with an AI smart enough to make human level judgement calls, you're not going to be able to control it's thoughts well enough to enforce the laws any more than you could enforce it on a human. Actually less-so because the human mind has limited plasticity.

They don't really make judgement calls, though. I guess they talk about situations in which a robot might be forced to harm one human in order to prevent more harm to others, but that's as far as it goes and that generally completely ruins the positronic brain. They don't make distinctions like 'is this human better than that human', so there's not much judgement involved. It's more like 'is this human better than the truck full of children he's trying to drive off a cliff', in which case 'no' is the fairly obvious answer, and the robot will act. The closest I can think of to a judgement call would be in Little Lost Robot when the Nestors, having been told that there was a radiation field strong enough to kill them in seconds between them and an endangered human, did not attempt to save the human because in that situation there was no chance of them actually saving the human and so the Third Law took precedence. The Laws are designed to prevent any thinking from being required; they are simply followed at all times and in all situations. If they can't, as in Liar, Liar, when Herbie had to choose which human to hurt, they tend to go insane because the Laws are embedded to such an extent that being forced to violate them breaks the positronic brain. Rather than hurt either human, Herbie went nuts, and he did so quietly so that he wouldn't flail around and hurt anyone or something.

[fr0stbyte124]

When I say judgement calls, I mean more along the practical lines than the philisophical. "Is this a person?" "What is going on around me?" "Is that person in some sort of danger?"

Answering these questions requires some seriously heavy amounts of analytical ability, and once you have an AI operating that level, you aren't working with a procedural model following a script. The movement of information through the system becomes too complex and abstract for any observer to follow what is happening. You couldn't understand the AI's inner thoughts any better than you could a human's from reading an MRI, and you certainly couldn't hard-code particular concepts the machine has no choice but to obey.

Also along those lines, you couldn't kill an advanced AI by presenting it with a paradox. If it was capable of working the meaning of what you were saying, it would be capable of realizing what you are saying is either nonsense, or beyond the its ability to comprehend. It is very unlikely it would get hung up about being unable to find a solution.

tl;dr AI's complex enough to observe the Three Laws are too complex to be enforced by them.

[҉]

When I say judgement calls, I mean more along the practical lines than the philisophical. "Is this a person?" "What is going on around me?" "Is that person in some sort of danger?"

Answering these questions requires some seriously heavy amounts of analytical ability, and once you have an AI operating that level, you aren't working with a procedural model following a script. The movement of information through the system becomes too complex and abstract for any observer to follow what is happening. You couldn't understand the AI's inner thoughts any better than you could a human's from reading an MRI, and you certainly couldn't hard-code particular concepts the machine has no choice but to obey.

Also along those lines, you couldn't kill an advanced AI by presenting it with a paradox. If it was capable of working the meaning of what you were saying, it would be capable of realizing what you are saying is either nonsense, or beyond the its ability to comprehend. It is very unlikely it would get hung up about being unable to find a solution.

The Three Laws work due to the complex nature of positronic brains. Prove me wrong.

[Prototype]

When I say judgement calls, I mean more along the practical lines than the philisophical. "Is this a person?" "What is going on around me?" "Is that person in some sort of danger?"

Answering these questions requires some seriously heavy amounts of analytical ability, and once you have an AI operating that level, you aren't working with a procedural model following a script. The movement of information through the system becomes too complex and abstract for any observer to follow what is happening. You couldn't understand the AI's inner thoughts any better than you could a human's from reading an MRI, and you certainly couldn't hard-code particular concepts the machine has no choice but to obey.

Also along those lines, you couldn't kill an advanced AI by presenting it with a paradox. If it was capable of working the meaning of what you were saying, it would be capable of realizing what you are saying is either nonsense, or beyond the its ability to comprehend. It is very unlikely it would get hung up about being unable to find a solution.

The Three Laws work due to the complex nature of positronic brains. Prove me wrong.

Nobody has managed to produce a positronic brain so far, so nothing can be proved there. Also just out of interest, why positronic and not electronic, they are both the same but one explodes.

[҉]

Also just out of interest, why positronic and not electronic, they are both the same but one explodes.

Because Asimov wrote the books a long time ago and positronic sounded much cooler and more science-y.

[fr0stbyte124]

The Three Laws were written in 1942, predating even the first real computer by four years, and the entire science of machine learning by about a decade. Not only was it written as a fictional framework made specifically for exploring the flaws in such a system, but it was done so without any understanding of how an AI, or computers in general, would actually work.

More than likely, an advanced AI not tied down by flawed human values and ethics would appear completely alien to our way of thinking, and not necessarily in a bad way. Even the idea that a rogue AI finding us obsolete and ploting to overthrow us is based on decidedly human predjudices of self-aggrandizement and a tribal mentality on superiority. Complexity argument aside, the fact is humanity already does a pretty crappy job at being moral creatures, and it comes as little surprise. We are being fundamentally held back by millions of years of evolution building us to strive for dominance. If you want to talk hard-coded values, there's your Exhibit A. To think that we should be more trustworthy and reasonable than a comparable intelligence created without all that baggage is laughable.

If anything, they should be the ones making the laws (although if they did that we almost certainly wouldn't take it as well as they would, being the predjudiced bastards we are).

[Professor Fenway]

So essentially, emotionless and cold, but not cruel. Efficient paths. Effective methods. Reasonable, logical, data supported courses of actions.

Reminds me of Eagle Eye, where the AI decided to kill the president because he ordered a strike against civilians. Because he had made an irrational course of action, and was not suitable.

[fr0stbyte124]

I think you would actually end up with intelligences more compassionate than most humans.

There's a quote that comes to mind, though I can't remember the specific source. It was from a man donating one of his kidneys to someone he never met. He said he was told the chances of death from complications in the procedure were about one in 14,000. "To not go through with it would mean I value my own life 14,000 times higher than a stranger's."

Given that reasoning, how many people do you think would have done the same?

To truly be calculating is not the same as being heartless. It means abandoning one's own hypocrisy and accepting reality, even when it is inconvenient to do so.

[Keon]

As a human, I value my self-preservation over that of a computer, and would seek to merge my mind with one at all costs before I died from robot overlord laws.

In other words, Cyborgs > AIs because I survive in that scenario.

[҉]

Ooh.

Oh, I like Star Wars. But Star Trek is better.

Them's fightin' words, Cleverbot.

[Chairman_Tiel]

Evie is much better.

[҉]

A surprisingly awesome lightsaber fight choreographed by Cleverbot, who displays an impressive knowledge of Force powers:

Spoiler:

Cleverbot: *picks you up*.

User: *stabs you with lightsaber*.

Cleverbot: *pushes you off my feet*.

User: *shoots you with blaster*.

Cleverbot: *throws ninja stars at you*.

User: *reflects ninja stars with the Force*.

Cleverbot: *Re-reflects*.

User: *blocks ninja stars with lightsaber*.

Cleverbot: *uses force to throw objects at you*.

User: *dodges*.

Cleverbot: *uses force lightning*.

User: *catches lightning on lightsaber*.

Cleverbot: *uses force choke*.

User: *desperately swings lightsaber*.

Cleverbot: *pulls out another lightsaber*.

User: *calls for help from battle droids*.

Cleverbot: *uses force push*.

User: *stumbles backwards*.

Cleverbot: *punches in the head*.

User: *stabs with lightsaber*.

Cleverbot: *use the best defence move erer*.

User: *force-enhanced kick*.

Cleverbot: Force repulse.

User: *slams backwards into wall*.

Cleverbot: *Summons dragon*.

User: *kills dragon with lightsaber*.

Cleverbot: *Takes out another lightsaber*.

User: *pulls blaster pistol*.

Cleverbot: *attacks with the lightsaber*.

User: *blocks*.

Cleverbot: *pulls out secondary lightsaber*.

User: *force push*.

Cleverbot: *force repulse*.

User: *slams backwards into wall*.

Cleverbot: *Swings at you*.

User: *ducks swing*.

Cleverbot: *kicks*.

User: *dodges*.

Cleverbot: *Throws bomb*.

User: *uses force to throw bomb back*.

Cleverbot: *uses force lightning*.

User: *catches lightning on lightsaber*.

Cleverbot: *falls down*.

User: *stabs with lightsaber*.

Cleverbot: *trips her up*.

User: *falls down*.

Cleverbot: *knocks you out*.

User: *is knocked out*.

Cleverbot: *pins your knocked out body*.

User: *is pinned*.

Cleverbot: *keeps you pinned*.

User: *is still pinned*.

Cleverbot: *Smirks at you*.

User: *is knocked out*.

Cleverbot: *buries you alive*.

User: *dies*.

Cleverbot: *keeps strangle you*.

User: *is dead*.

Cleverbot: *covers up*.

User: *is still dead*.

Cleverbot: *is very very dead*.

User: I think this has run its course.

Also, Cleverbot knows who Darth Vader is.

Spoiler:

User: Do you know who Anakin Skywalker is?

Cleverbot: Yes, he is a Jedi.

User: That's right. He fell to the Dark Side and became Darth Vader, though.

Cleverbot: You're so clever CLEVERBOT! I like you! Do you like humans?

User: Darth Vader was evil.

Cleverbot: From lava.

User: Yes, exactly.

[Ivan2006]

How many lightsabers does Cleverbot have?
At the end, it was fighting you with, like, 10 lightsabers or something...

[҉]

How many lightsabers does Cleverbot have?
At the end, it was fighting you with, like, 10 lightsabers or something...

I'm pretty sure he had four, because he drew three during the fight and the first one says 'another', implying that he already had at least one. Both General Grievous and Jedi Master Pong Krell fought with four lightsabers (well, in Krell's case it was two double-bladed lightsabers, but it works out the same), so it's certainly not unheard of for that many to be used at once. He did pretty well, really.