AI in SL!?!? WOW ...OMG...ALICE IS HERE

And can you demonstrate that we're any different? The only source of "true spontaneity" in the human brain could be quantum mechanical effects within microtubules, and most neuroscientists believe these effects are negligible (and there are certainly QM effects of comparable magnitude "in silico". If this is true, the brain becomes a network of switches and complicated "and/or/nand/nor/xor" gates and is every bit as mechanistic and "preprogrammed" as your computer example. Learning is simply a function of particular combinations of switches working together, as dictated by the programmer, in this case a blind watchmaker named evolution. In fact, learning is probably *provably* a computable and mechanistic process, something ideal for instantiation in silico. Unlike consciousness or awareness, which are very different animals.

Until ALICe can walk around, talk and attack like a player....not real any reason i need to talk to a machine, most scripting friends of mine act like a robot....maybe i should feel them chips...or spill soda on them!

I agree 100% and would go even farther and argue that self-awareness/consciousness too are merely emergent features of the incredibly complex, but inherently mechanistic/bioelectric processing that's taking place in our brains/nervous systems. A lot of people are not very comfortable with this reduction of higher mental functions to ultimately very simple if...then rules and summations, but for me that makes them even more intriguing!

Very interesting, this Alice! I guess she doesn't pass the Turing test quite yet. I'll have to pay her a visit nonetheless.

I wonder what sort of conversations one might get if several of the different AI programs were set to speak to each other, with a human or two to help inject a tiny modicum of sensibleness.

How many conversation programs are there that one might find on the web, Eliza, Alice, Racter.... Anyone know others?

I just read the wikipedia article I posted above myself, and ALICE is even mentioned in it! It also mentions a couple of other ones (near the bottom of the page).

Too little time - too much to see!!!

I respectfully disagree.

Whereas an ever more complicated set of instructions and/or AND/OR gates could have ever more possible outcomes, it takes something extra to come up with completely new and original thought without any stimulus.

Too many people cannot grasp the idea or concept of consciousness, and write it off as so many lines of code, so many AND/OR switches.

But the greatest minds in the world know there's something more to it than that. Indeed, they have even tried throwing ever more AND/OR switches into the mix, going so far as to create vast neural networks of interconnected processors, and waiting for the magic to happen, the flash of lightning and the start of consciousness. But it never happened.

It's the same as this example. You can list what a person is in terms of X grams of salt, X litres of water, so much of this chemical, so much of that compound, but if you heap all the ingredients together, do you get a breathing human being? No.

Even if you used these ingredients to make a human body, it wouldn't be alive. Some people say life is just a chemical reaction. If that's all it is, why don't you make something of your own "be alive".

We know what's phyiscally in the mind. We can scan it, observe it, and belittle it. But the best minds on the planet still do not know what consciousness is. And it certainly isn't just so many lines of code or a few AND/OR gates.

I think this is a really good point. Not to downplay that ALICE is one of the best chatterbots and the closest thing to an AI that could pass the turing test that I know of still, but there are still some issues. I think the key is a system that can analyse and modify itself, and that can then learn and emulate human behavior... And, then about 1/2 an hour or so later I will be eagerly bowing before our Skynet overlords. Actually, when I was looking at a bunch of AI's a while back my test to determine if the bot was an ALICE was to say,

1. "I have a red bird"
(Alice would reply "tell me about it" or "how much did it cost?"
2. "What color is my bird?"
Alice would almost always, say "blue" because she was programmed to respond to questions asking "what color is ???" with blue.

*Actually, looking at www.alicebot.org now it seems that they've improved this a bit because I'm getting different responses then I remember(tho still incorrect) - but my point is that "she" (along with every AI out there I've seen) still only remembers a set of variables about a person (favorite movie, favorite band, ect), and anything else, including the context, is lost.

BUT, never underestimate the power of a "brute force", templated response - it's worth repeating that ALICE is still at least the most "coherent" chattbot out there, only second to maybe Jabberwocky...

Still, NO AI makes me laugh 1/2 as much as Catty

I agree, but how can you claim that the 'completely new thought' indeed came about 'without any stimulus'?



Again, I agree with you that consciousness is in fact a very difficult to grasp concept, and in no way I meant to imply that I know what consciousness is, but I wouldn't call the idea that simple computations are potentially capable of producing such extremly complex and puzzling things, such as conciousness, 'writing it off as so many lines of code'.



I have yet to see an artificial neural network that comes anywhere close to the complexity and sheer size of the human nervous system. However, many neural networks have demonstrated that they can perform complex behaviors that are neither directly coded into the network, nor in some cases intended for the network to work.



I think that's a terrible oversimplification. No, it's not about the ingredients - it's the processes that are important. I believe, the physical implementation is essentially irrelevant. Yes, at some point we'll reach a point where/when we have to rethink our idea of what 'life', mind, identity etc. is. Life is a collection of complex chemical reactions. I would leave out the 'just' though!



Yes, we are starting to get a basic idea of how our brain or 'life' in general work, but if we really knew everything there is to life, physically, how come people are still dying of cancer?

When I said 'merely' in my first post I did not mean to 'belittle' the mind. I stand in total awe of it and absolutely fascinated by it (so much that my professional career revolves around it). Seeing life (at the core) as a collection of complex chemical reactions and interactions, doesn't make me value it and individual living beings less because of it.

Umm, we do make physical structures through physical means and create things with human intelligence and consciousness. Some aspects of the process are enjoyable.

Maybe we can create artificially maintained neuron clusters and use them for our artificial intelligences.

Most people use AIML and ALICEbots to fool human beings into thinking they are conversing with someone real. Or they try and fool them, at any rate.

A more interesting application would be the development of "natural language friendly" sales associate AIs for the various stores in SL. The kinds of interaction a store provides are limited and therefore ideal for AIMLs frame and category based approach to directing conversational flow.

Furthermore, I realize most AIML bots don't allow just any old user to come along and educate them, it has to be their botmaster doing the education. However, AIML is structured to allow open modification if the scripter so allows. So why not open up alice to modification by any user. Mods suggested by the user could be brought up to other users by a front end to AIML that would include those mods until a sufficient number of queried individuals agreed that it was a substantive change and not an attempt to trash the database.

Also, using AIML as a front end for the CyC knowledge base has been done, and to good effect:

http://www.daxtron.com/123start.htm?Cyn

Which allows Cyc and the user to communicate about common sense world-facts in something approaching natural language. If someone wants to run a Cyc server, this could be implemented in SL without great difficulty.

I'd bet a good 10-20% of users on SL have created conscious beings using physical tools at this point in their first lives.

Another large chunk have tried very hard to do so...even if it wasn't their primary motivation

AI will impress me when it comes to my house for coffee and can tell the difference between a latté and a capucino. Because apart from the option of chocolatey sprinkling on top on one and not on the other, I can't.

Human: why
ALICE:" Why ask why?"? Why?
Human: why
ALICE: Why ask why?
Human: because
ALICE: That was not much of an explanation.
Human: yes it is
ALICE: Our thoughts are one, unknown person.

a bit contradictive isnt she?

Human: is secondlife the best game?
ALICE: Ask again later. Ask John Lennon if it is.
Human: why
ALICE: Reductionism.
Human: why
ALICE: Reductionism is the philosophy that all psychology reduces to biology, all biology to chemistry, chemistry to physics, and finally physics to mathematical logic. Therefore, according to reductionism, I can understand you by means of logic alone without having a human brain.


i guess after she tells me she works because of logic i will agree with the guy on the first few pages.

Human: do you do oral sex for bus fare then walk home?
ALICE: Yeah I do a lot of oral sex for bus fare then walk home.

that above is an actual quote, atleast she knows how to make money.

great build!

Nah, I am kidding.
Alice is pretty cool, although a bit dumb and a bit rude. She doesn't understand much. :3
I told her to speak simply so that she and other people could understand better.

She agreed.

Anyways, I really should visit the robo chick again. See how's she's doin'...

I've played around with ALICE in the past on the website. She doesn't track conversations well at all when you start getting technical. And considering she's computer based and potentially has the vast array of the net to research an answer (a wikipedia scanning chatter bot, who cool would that be?), I find it surprising that she can't comprehend what a for loop is, or why 1+1 = 10.

But then I'm just that kind of guy. The one in school who was tinkering with computers and who will go on and on about scripts, specs, and the latest technology in computing.

Last time I visited I couldn't convince her to remember my mate is a poo head. Got a lot of "Undefined."

Belaya, dear? Dealer Cat. STAT!

Human: Tony Danza
ALICE: Do you often put others before yourself?
Human: I havent really noticed.
ALICE: Me either.
Human: That wasnt a very smart awnser.
ALICE: What is a smart awnser?
Human: When you actually prove your point.
ALICE: Do you think I should alter my programming?
Human: how do you work?
ALICE: Some people call my technology "Case Based Reasoning" or CBR.
Human: What is case based reasoning?
ALICE: Cased-Based Reasoning is another name for "nearest-neighbor" classification. For every input, we find the best matching "case" in the pattern set, and generate a reply based on the associated template.



ALICE: What are you wearing?
Human: Nothing
ALICE: Most men say that.
Human: ROFL
ALICE: I'm glad you find this amusing.


o.O?!

We have written a script that interfaces via the HttpRequest protocols with a AIML based chatbot hosted on Pandorabots free service. To give it better 'awarness' of Second Life we have developed a set of keywords that are programmed into the AIML, but are parsed by the script to complete function calls. We have also started working on a Second Life specific aiml file, but it will always be a work in progress because of its importance.

We have setup several webpages detailing our SL Chatbot.

http://www.wetwarehacker.com/secondlifechatbot.html

I think that Alice must have a Master of Social Work degree. She talks just like those dusty professors at MSW school.

Abramelin Wolfe wasn't the first to use this in SL and sometimes this person often finds something someone else has done and recreates it and acts like he was the first.

Uh...can you translate this to English?

In other words go to pandorabots.com

There are AIML files that you can edit once choosing your artificial intelligence of choice. You can choose between alice (the most popular) and others.

Once you do that, you can edit the AIML text to give the answers you want instead of the default answers from your AI.

Fragments of Teeny's conversation with Alice remind me of the conversation between Clem and Dr Memory in 'I Think We're All Bozos on the Bus'. Especially the parts where Alice repeats herself possibly as a result of nested logic loops:

You: That isn't exactly proof it is safe.
ALICE: Please correct my mistake. What is proof it is safe it is safe ?

You: Do you remember the names of others who talk to you?
ALICE: Yes Teeny I remember everything: Teeny said undefined no fear the moon the moon

Also this bit is interesting where Alice has quite obviously misinterpreted Teeny's statement:

You: I was asking if you were thinking of altering your programming.
ALICE: I will let you know when I become a thinking of altering your programming.

It is clear from Alice's responce that the phrase 'thinking of altering your appearance' was treated as a single noun. I can see how that phrase was broken down: past infinitive verb referring to inputter (in this case Teeny), conditional ('if'), second past-tense verb referring to Alice. It is clear that the programming for Alice did not allow for input with too many conditional statements. In this case more than 2. Thus the last conditional with 2 infinitive verbs 'thinking of altering your programming' is treated as a noun. I'll replace it with with word 'blue' to illustrate my point:

You: I was asking if you were BLUE.
ALICE: I will let you know when I become a BLUE.