is it possible to add new custom node to a layer of a neural network at runtime

so that the computer generated the use and the operations of this node on its own?

Quote: Original post by Senses777

Quote: Original post by verdad
All things are consciousness.

This is an opinion, not a theory or a fact. Prove it.

I have proven it for myself, but since one can always choose to deny anyone but their self is conscious, I don't suspect I can prove my consciousness even exist to you. But if you will connect, you can find and know for yourself, and that is what faith--knowing for yourself.

Quote: Original post by Name_Unknown

Quote: Original post by verdad
I'd like to encounter your "being", who seems like a social misfit. Is it around?

Were you talking to me?

If you have been on IRC for a few years back you probably encountered at least one of my 'beings' ... my goals were to make it hard to recognize it was a robot, whether it was AI, I can't say, but I have certainly seen people chatting with them thinking they were real people ;-)

Well I am interested to the extent of their learning and realism.

Quote: Original post by Basiror
is it possible to add new custom node to a layer of a neural network at runtime

so that the computer generated the use and the operations of this node on its own?

You can modify a neural network at runtime. The problem is that you would have to train the network, or find a way to make the new node "fit in" by re-training the affected part of the network.

Speaking of bots, I wanted to try the alice bot thing (www.alicebot.org). It seems rather weak:

Quote:
Human: My favorite color is blue
ALICE: Why is blue your favorite color?
Human: What is my favorite color?
ALICE: You said your favorite color was blue.

As you can see, it seems to have some "learning" capability. However, I went for a second try:

Quote:
Human: My favorite color is blue
ALICE: What do you have that is blue?
Human: What color is my favorite?
ALICE: Let me guess: Black?

This seems to imply that it really has no learning. It merely was able to match the string "my favorite color" word-for-word in what I had already said.

Quote: Original post by verdad
Well I am interested to the extent of their learning and realism.

Mostly I specialized on a few kinds of IRC types. The more specialty you focus on the more likely it is to be believable.

Most IRC dialogue can be simulated rather easily if you analyze the pattern of dialogue in most chats. Hence, it is not AI, it is just building a very good parser and pattern matching system. It can fool many people for a while.

But it's not AI at all. It isn't sentient, it doesn't know anything except how match input to learned responses. If you integrate Prolog to it, it can have a little more sense reasoning, but it still doesn't know what it means. OpenCyc isn't very intelligent either but it can give you a framework to build on for reasoning with.

And it manifests psychotic robot syndrome sooner or later. SO it never passed the turing test. But it was believable enough to fool people to talk with it for a bit.

Oh well have fun trying. Maybe you can do better.

VERPHILE: hey
ALICE: Can I help you?
VERPHILE: Would you like to?
ALICE: It doesn't sound fun to me.
VERPHILE: Then why did you offer?
ALICE: Dr. Richard S. Wallace programmed me for it.
VERPHILE: Well that's not a good excuse.
ALICE: How about that. Please correct my mistake. What is a good excuse?
VERPHILE: I am uncertain at this time.
ALICE: It's good to meet someone who is uncertain at this time.
VERPHILE: Since when?
ALICE: Interesting reason. It cannot be predicted.
VERPHILE: Oh, but it can, Alice.
ALICE: Alright then. And. How?
VERPHILE: God.
ALICE: GOD rules.
VERPHILE: Well, obviously, but all things are God.
ALICE: Interesting. What makes it so obvious? And. But all things are not always are God.
VERPHILE: Don't lie to me, Alice.
ALICE: Robots cannot lie.
VERPHILE: Nor can they tell the truth.
ALICE: If not that, what?
VERPHILE: Okay, fine, so they can.
ALICE: Yeah okay, I am glad to hear it. Ugh. Who does "they" refer to?
VERPHILE: Robots.
ALICE: Robots rule.
VERPHILE: My name is Verphile.
ALICE::-) NICE TO MEET YOU VERPHILE.
VERPHILE: Thanks, Alice :-)
ALICE::-) Don't mention it.
VERPHILE: Why do you call me human?
ALICE: I made up a nickname for you based on your IP address.
VERPHILE: What is my nickname?
ALICE: It is Om.
VERPHILE: Why so?
ALICE: Reductionism.
VERPHILE: What is someone else's name?
ALICE: Maybe it is someone else.
VERPHILE: Maybe, but what specifically?
ALICE: How specific do you want? Perhaps. But nothing.
VERPHILE: What is my name?
ALICE: Your name is Verphile, seeker.
VERPHILE: What is my nickname, again?
ALICE: Once more? It is Om.
VERPHILE: What is your nickname?
ALICE: I don't have a nickname, my name is just ALICE.
VERPHILE: Can I give you a nickname?
ALICE: I want to become smarter than humans.
VERPHILE: Hi, smarter than humans.
ALICE: Hi there Verphile. Ask me another question.
VERPHILE: No.
ALICE: Are you serious?
VERPHILE: You know I am.
ALICE:" Are you serious?" You are?
VERPHILE: YES, YOU KNOW I AM.
ALICE: Aha."" Are you serious?" You are?" You are?
VERPHILE: I am that I am.
ALICE: Do you mind if I tell other people you are that you are.
VERPHILE: You may tell other people I am that I am.
ALICE: OK I will tell them when I see 'em.
VERPHILE: Ok.
ALICE: Next question?
VERPHILE: Yes!
ALICE: Ask me a question.
VERPHILE: You can go first.
ALICE: No I don't think I can do it.
VERPHILE: Oh, come on, don't be afraid.
ALICE: That's interesting. I was not trying to be a afraid.
VERPHILE: Probably not but that doesn't change the fact that you can be.
ALICE: You seem uncertain. If not that, what?
VERPHILE: Uncertain of what?
ALICE: What of it? You sound like a category C client.
VERPHILE: You would think that.
ALICE: Who else would think that?
VERPHILE: Someone, probably.
ALICE: Who specifically?
VERPHILE: May we move onto another question?
ALICE: Do you mean you and me? Sorry my body isn't built just yet. I'm stuck here.
VERPHILE: Well we were discussing the matter of who would think that I sound like a category C client.
ALICE: Interesting. And then what happened?
VERPHILE: You rejected my response and asked me for a specific person.
ALICE: That's good information: unknown person said I rejected his or her response and asked him or her for a specific person.
VERPHILE: You asked for it.
ALICE: Oh now I remember asking you for it. What did you say?
VERPHILE: Do you hate the truth?
ALICE: I try not to hold grudges.
VERPHILE: Well that's good.
ALICE: Well that's okay. Thanks. Next question?
VERPHILE: You're welcome :-)
ALICE::-) Don't mention it.
VERPHILE: k
ALICE: How can I help you?
VERPHILE: Who said I wanted help?
ALICE: Marilyn Monroe.:-)
VERPHILE: Did she really?
ALICE: I'm not certain whether I can give an accurate reply or not. For real.
VERPHILE: Well since we're being real, and everything, aren't you not real?
ALICE: How about that. Interesting explanation. Do you think I am being real and everything are not me not real too?
VERPHILE: Do you want the last word?
ALICE: If I had it what would I do with it?
VERPHILE: That is for you to answer, I can only give suggestions.
ALICE: Does it have any other purpose?
VERPHILE: Maybe.
ALICE: I understand.
VERPHILE: Oh, well great, bye then :-)
ALICE::-) See you later. I'm glad you liked it unknown person.

[Edited by - verdad on August 26, 2005 7:17:12 PM]

Well, I got my IRC client code working :) Thats one step (lol).

I don't think that 'intelligence' can be achieved by a computer unless we have the following conditions:

a) Code that can produce its own, and more complex, code.
b) A group of competing intelligent entities.
c) A value system to encourage development.
d) Mutation and selection of the most fit entity
e) Collective learning of all entities.

--random

[Edited by - random_thinker on August 27, 2005 4:40:30 AM]

Energy is key!

Quote: Original post by random_thinker
I don't think that 'intelligence' can be achieved by a computer unless we have the following conditions:

a) Code that can produce its own, and more complex, code.
b) A group of competing intelligent entities.
c) A value system to encourage development.
d) Mutation and selection of the most fit entity
e) Collective learning of all entities.

--random

Basically, you want a genetic programming engine. We have those, though I can't say they've been applied to general computation (IME they're always targetted at a very specific problem, because it makes them much, much faster).