Can machines think?

This is just beautiful

A: Do you experience love?

J: No. I’m 53 years old.

A: Can you explain how it feels to be 53 years old?

J: How old are you?

A: I am 9 years old.

J: Nine? I don’t believe you : ).

A: : -) I don’t have any reason to deceive you.

J: Yes, you do, you’re a woman.

That's also a good one, hehe

I think that we are going to see much smarter AI in the near future - especially from Google. It's pretty interesting that they have created the best Arabic <-> English translation without knowing Arabic or its rules. [link]

With a lot of computing power and LOTS of data, it's possible to create some amazing things. An idea could be to use AI to display some better contextual ads or to understand quires even better - quires like "show me posts on amix.dk that's about JavaScript".

me: Jabberwacky, of course no
--
you know, Jabberwacky and A.L.I.C.E. are definitely the best chatter-bot competitors.
here is another scene of Jabberwacky with human this time:

human: How much is 10 apples plus 10 apples?
jabberwacky: 100.

human: Why can a kettingzaag be dangerous if not used properly?
jabberwacky: You are not a nice person.

translation from AI... may be, but that may be hard to understand for the robot.

look at those two conversations:
User One: Hi
Jabberwacky: Hi
User One: What is your favourite television program?
Jabberwacky: I like running
User One: Mine is Southpark

User Two: Hi
Jabberwacky: Hi
User Two: Do you have any hobbies?
Jabberwacky: Yes, lots
User Two: I like running
Jabberwacky: Mine is Southpark

he (Jabberwacky) thought Southpark itself a hobby!


but, i didn't know google arabic translation was built on AI, interresting.

also i know a trial to teach a (baby) chatterbot from scratch (without any previous knowledge of any language), visit Ditto and say hello.

unknown:
Google's Arabic translation isn't built on "AI", but on statical analysis.

My point (which was unclear ;]) is that you can fake AI and grammatical understanding by having LOTS of data and lots of computing power (i.e. using statistics). Google's Arabic translation is so good that it looks like they have understanding for Arabic, but they don't :)

Maybe the discussion is what's really AI? If a machine can "fake" intelligence through statical analysis - - is it really intelligence?

Nothing has come close to "real" artificial Intelligence yet anyway, it is a deep philosophical question wether or not it is possible at all. Intelligence in the human definition demands a conscience, that is; it needs to understand what it is talking about.

The thing is that we don't even know why we have one, so simulating it is a bit hard...

What all the bots+google's translator are doing is essentially the same, it an expert system build around either a static or dynamic ensemble of relations.

The question is how does 100 billion neurons with loads of interconnects spawn our persona? Could we do the same with say, all the CPU's on the Internet? naturally we would have an impossible slowdown simply because the pathways are way longer then the nerves and synapses in our body, but could we build a slow but yet thinking machine if we just put enough virtual neurons together?

It would be interesting for sure.. Do you know Ray kurzweil? I guess you do, but he is thinking a lot about what he calls the Singularity, where not only real AI would spawn, but we would be able to download our minds into the Internet... He loves to write, so his page is quite stuffed with this kind of stuff.. http://www.kurzweilai.net

Another interesting blog post I stumbled upon:

• How I Failed the Turing Test

Morten:
When I say AI, I don't mean intelligence in the human sense. I mean AI in the sense to be able to decode queries like "show me posts on amix.dk that's about JavaScript" or similar light-weight intelligence.

Google's translator isn't an expert system in the sense that it knows rules of a language - the translator uses statical analysis to translate between languages. Now, Google's got the best translator around by having a system that has no understanding whatsoever. My guess is, that one could fake AI in a similar way. I.e. fake understanding :)

Thanks for the link (I haven't heard about Ray Kurzweil). It looks interesting, but a bit "unreal".

i think you may be interrested to know little about AIML, the markup language ALICE concists of.


about kurzweil, do you know about the chalenge about AI future ?
make your bet here

That's the entire point, aslong as nobody figures out that the program is only faking, it is intelligent.

If it can keep a conversation interesting and sane for let's say 10 minutes, then nobody would care if it was merely creating the illusion.

But I can understand your comments, and it are these same concerns that have split up AI-makers everywhere into two camps. On the one hand there are programs such as ALICE.

ALICE works by 3 simple steps; first all input is reduced by omitting everything not relevant, possibly even verbs. Secondly the remaining input is searched into a relativly small database (46000 possible patterns), and the best possible match is found. Lastly, the output is generated according to the template in the pattern.

And then there is Jabberwacky.
Jabberwacky is the brute force approach, it saves every conversation it has ever had, and when a sentence is given, it searches this massive database for a match.

Compare it to breaking into a house, ALICE picks the lock, and Jabberwacky smashes the window. But humans just take their keys out.

DANmArk styrer