The AI Engine

Isaac Asimov 1st law of robotics states " A robot may not injure a human being, or, through inactions, allow a human being to come to harm.'
The Second Law of robotics is: "A robot must obey the orders given it by human beings excpet where shuch orders would conflict with the First Law."
The Third Law of Robotics is: "A Robot must protect its own existance as long as such protection does not conflict with the First or Second Law."

Now change all of the words Robot to AI. That would have to be its core, or built in there any ways. Now as for it doing more that what its program for, well thats not Ghost in the Machine (which I have not see btw) And if you read my first post when I described the AI it is quite with it its progrmaing to do what ever it wants to. U see it can rewright its self so there for There is no programing limitations. Once the AI is created and is told the basics so it can get an knowledge base it is free to do as it pleases.

That is where the 3 laws come in. They will make it so that it is not and I quote u "An Evil Genius AI".

See the one big thing about this AI over everything else is that it CAN program itself. This makes it so that it has the ablity to adadpt so that its not stuck to what it was started out at. Kinda like Genic Codeing. It alows the AI (this is not a bot is moved beond that) can change itself to what ever the siuation is.

I dunno if the negative side of AI should stop us from doing more research into it. Every technology has it's +ve and -ve uses. Consider electromagnetism. You have engines running cars,stations providing you with electricity, electric appliances and that same technology is also being used to drive tanks etc.
If there are negative uses , there are definitely huge positive uses as well. Btw, do we reprogram ourselves too??? I mean no one knows the internal structure of our brain and maybe that structure is changing (which I think it does, that is how the weights in the neural networks get changed). Thinking about the negative uses and stop pursuing the ultimate goal is not a good idea. This universe has been made for us and we are to conquer it(that is how Islam puts it). you really cannot lose hope and stop working as there is still a LOT to know!

I agree, Immadn, that we should not let any such fears get in the way of continued development. People are always afraid of new ideas and changes. Take, for example, the establishment of time zones. To our mind it makes perfect sense, but at one time the US was all in one time zone. An hour was declared where everyone would adjust their clocks to fit the time zone scheme. There was uproar. People thought the world would end. And when the moment came, tick-tock-tick-tock-TURN-tick-tock, life just went on as it had.

I know I am not saying that we need to make one, What I am saying is we need to be carful about it. What if they had set the timezones backwards of what they are. That would be hell. No deaths or something would have comefrom that but it still would have made life hell. But and True AI with access to the internet could do some major damage. We need to be carful with the AI. Maybe make it on an IntraNet first before going and letting it have access to the out side world... I'll Learn C++ in the next cupple of days and see if I can comeup with something. I will run on my network. That should be enough space for it to grow. I will leave this system off for it so that it cant get out but I can tell you about it.

But you're missing the point- how could an AI possibly get into systems that a person couldnt? You're imagining some energy-based life form that walks the walls of etherspace as if they werent there.

An interesting thought you inspired in me was that of an AI moving.. At this point, an AI couldnt move it's code off of the place where the file is based. But letting it do so isnt so impossible. Imagine it worked like a virus or worm- it could copy itself to another server, activate itself, then turn itself off on the server it came from. Or it could even jump onto an end-computer somewhere. This would take some ingenius programming, and I wouldnt know what the point would be, but it is possible.

Ok, well... It will be an energy-based life form. think about it if its fully seniant then its a life form. So it sees its world like we see ours. It comes up to a gate way and sees it as a door. It comes up to a password it sees it as a locked door. and just like a locked door all it has to do is pick the lock.

God I hate this because it sounds bad but... Imagian your self in the Matrix you just need to see something to move it. Well the AI sees it as data and just moves it.

Now as for the moveing its self around ya it would work just like a worm. It copys its self over then sends the "whats happends so far" update to its self then deletes its old self. Or it could Just copy it out there and runn it back to its self. makeing the whole internet or what ever a large neralnet. That way it could incress its computioal power and space while remainging one orginsiam.

You need to stop seeing this as Programing and see it as a large Org. It can do everything we can do and then some since it knows programing way better than we ever can since it is the programing. That is why it can work on its self so that it can make its programing better. Think of it like us doing genicet enginering on our selfs to make us faster, stronger, and smarter.

Quite nice discussion so far...

Ender:
The concept of AI changing its own execution code isn't such a new idea like you probably believe it to be. If I remember correctly it was even used in that old LISP AI Language. The Prof's right about AI facing the same limitations like you. Even if there would be some genius coding a worm-like AI which tries to hack into systems it can't do more than you as a hacker, it's just 1000 times faster in trying things out. So your superior AI Entity couldn't tell the stupid computer on the other end of the phone line to tell him all the User passwords for example or directly (physically) influence its actions. As long as we don't even understand the schematics of our own Brain, we're probably not able to do a really life-threatening (smart) AI.

The Prof etc.:
Where is the sense in having a large database of information without an AI actually using it in a creative way ? I'm with immadn here who said intelligence has to be evolved, not programmed. As Ender told you could have cross-references from the wordnet to some other database to let the Bots know about a wider context. But this still does not make them reason in any way. The only reasoning I see here is the pattern matching: Bot analyses the given sentence, does a pattern-matching with all the words and gives back a triggered response which doesn't change anything in the AI Engine, like it would when talking to humans (changing emotions, learning, whatever). I think using a neural network might be promising but probably won't succeed at the moment because of the complexity and the number of calculations/neurons one would have to use which is far too much now for a standard computer. Talking about some Ethics, Consciousness or even Laws of AI could be appropiate in some 10 or 20 years but not now...

Programming is like teaching. Each time I add new functions to the AI Engine, I'm teaching it. Since evolution involves random genetic mutations and survival of the fittest in a hostile world, I wonder how you would apply that to AI.

I dont think neural networking is going to make any strides until we understand what exactly neurons are doing. Otherwise we have a very complex and confused pile of spaghetti.

I think you're right about "self-applied" laws of AI ethics and behavior- that's for the future. Right now, it's up to the Personality-builder to exercise ethics when designing their Bot, lest the Bot be mean and unfriendly.

It's the old behaviorist vs. the world argument. The Prof is right to compare teaching (programming) to the effects of evolution. Of course, a neural net tries to more closely emulate the process of evolution, but that's still a very theoretical process and not without some plot holes, as it were.

Of course, the most satisfying AI is one where we think we understand what's going on in the AI's "head". Unfortunately, we don't sufficiently understand what's going on in our own heads. It's also worthwhile remembering the difference between AI and consciousness. A bot could be what we'd call intelligent without being conscious.

Possibly the best first step we can hope for is for an AI to pass the Turing test. After that the question remains of whether we can actually take AI any further or whether to do so is just to indulge rank speculation on intelligence and consciousness that may be totally unrelated to what goes on in our own little minds.

The Professor, you asked how can evolution be applied to Neural Nets. Well, it is possible. Such neural nets are called GenNets. They evolve over time. Our brain is an example of a hyper complex system that we cannot fully comprehend. Though we might be able to evolve a brain using genetic programming, that is genetic programming evolving neural networks. We might not be fully able to comprehend the resulting code as it might be too complex but it will surely function the way we want it to and this is the only promising (and economical) way if we want to understand the inner working of the brain. That is trying to understand the resulting code rather than doing costly experiments with the real brain which is perhaps still too complex for us to understand.

I'll be honest here- I think the theory of evolution is ridiculous. To think that we exist by a series of random mutations is the most shortsighted thing I've ever heard. As long as you dont think about it, you can believe in it. But imagine the nightmare of misshapen creatures that should be found in the fossil bed, or the incredible variety of people that would exist with non-fatal random mutations- patches of green hair on the hands, an extra eyeball, a dozen tiny failed useless organs in the abdomen that cannot be detected without surgery, and so on. There was a plan, a design, and a purpose to every part of us. Science exists by right of breaking off from the church, by dividing existence into the spiritual and the material. It's a millenium old habit, but science continues to embrace the material only, denying a clock-maker and a purpose, since those things are the realm of the spiritual.

I believe we should speak of learning and conditioning when it comes to AI.

I'd like to announce that I'm going to be putting together Gaia in the next week or so. It'll be the first AI based on "collective consciousness," if you will. I applied for the Loebner Prize, which awards a medal to the most believable AI each year, and if Gaia is successful, it shall be the participant.

Well random mutations do occur but are very very rare. I heard that there was one part in America i think , where the children that were born were very very genius. Probably a case of mutation. And talking bout evolution in AI, well it works!
I cannot say much bout it, but most of today AI machines are examples of 'logic engineering' than true intelligence.
Wish u good luck with Gaia, btw what is "collective conciousness"?