Artifical intelligence that you can talk to?

Posted by spike1 on July 11, 2013, 8:17 a.m.

Hello! I was just wondering what you guys think of artificial intelligence(natural language in specific). I was rather interested in AI a few years ago and found a few good books(AI – A mathematical approach is a great read by the way :D), but they all showed rather rigid ways of creating an AI, and they all appeared to fake our intelligence, not become it. So first off, I made an AI(it was pretty good actually :D, had some nice conversations(should I post it?)), and it was good, but once again, it just faked intelligence.

So that got me thinking, what is intelligence, and to be honest I cant find an answer. Although I would love to go by Alan turings idea of an AI, which is that if someone cant tell the difference between it and a human then it is therefore intelligent, but that dosn't seem right, at least not to me. So yeah, I was wondering what your opinion on artificial intelligence is? Will we create something that can pass for human one day(in a chatroom sense), will we ever create something that can learn? Would that still be considered intelligence, or would that still be fake(in your opinion)?

I honestly believe we will create something that can be confused for a human, but I have trouble with the real intelligence part. I really want to believe that all we are are chemical reactions, and therefore can be re-created in the computer, but it just seems incomplete. I don't believe in us having a soul or whatnot, but I think there has to be something more to us, more than just chemical reactions. For example, if I burn my hand, it hurts(I know, spilled boiling water on it a few months ago, crap that was horrible), and I have trouble believing that that is just my mind telling me to feel pain. Does anyone know of any theories that could explain things like this, or are we just chemical reactions and our current knowledge is correct?

Anyway, to gear it a little bit more towards gamemaker, I propose a challenge(as long as there arent any other ones happening here) which I think will be fun. Lets all have a go at creating something that we can talk to in a chatroom like environment. No deadline, I just want to see what we come up with :D.

Btw, if you have read all of this, you are crazy lol :D. Thanks for reading! :D

Comments

JuurianChi 10 years, 10 months ago

Hrm.

I think intelligence can be marked by the ability to learn and retain information and then be able to use that information later.

And it's not so much as being able to learn and use all the information in the world, but to put it all to use in a manner that is useful to the first party.

It's like how animals can learn to function in domestic society. Or how Cleverbot can use information given by other people to somewhat mimic the idea of having a conversation.

spike1 10 years, 10 months ago

That was quite an interesting video, I'm amazed they have figured things like that out :D. I am interested in your idea of intelligence, although im not sure I agree with it(then again, things like this have no answer :D). Dosn't that logic imply that a gamemaker array is intelligent? It can store information, although it fails at that second part as the programmer has to tell it what to do.. Sorry, went off on a fail there although ill keep that sentence as i can use it.

The problem with the array being intelligent is that another intelligent being has to tell it to use its information, and i struggle to see any AI that dosn't work like that. If its ok that the programmer tells it how to use its information, then arrays are intelligent, and that doesn't sound right. So by that definition, a true AI is not possible, only one that fakes being intelligent?

Its certainly an interesting idea, and seems to be the same problem i face when i try to imagine a real AI. How do you think a real AI would work? It would have to be able to process the information, and work out what to do with it, but a programmer would have to tell it how to work out how to proccess the information, therefore creating a loop which cant be solved. Any thoughts? Sorry for the terrible reply structure :D(Its 2am, so if this dosn't make sense then please forgive me lol :D).

F1ak3r 10 years, 10 months ago

Quote:
I think there has to be something more to us, more than just chemical reactions.
And if it turns out I'm wrong, well, yay for new computer friends!

twisterghost 10 years, 10 months ago

What you're talking about is creating something that will pass the Turing Test:

http://en.wikipedia.org/wiki/Turing_test

I've been working on a project for myself which acts as an AI-esque assistant for remedial tasks I don't feel like doing every day, and started off with the idea of making it understand (relatively) natural language. I didn't try too hard on that because natural language processing is itself an entire field of computing with people dedicating their lives to researching it. It really is outside of the scope of "I'm hacking on this project that passes the Turing test"

Still fun though.

Basic things like "in 20 minutes, notify my phone the server response time"

Itll essentially run that through a series of parsers to figure out what needs to be used to break it down. Just a shitton of regex, really.

"in 20 minutes" will be parsed to set up a timer with the rest of the command executing then.

"notify my phone" will set up a task to be executed where it will make an API call to NotifyMyAndroid

"the server response time" is a bit more complicated. "the server" will default down to a knowledge dictionary of terms to places and types. It will see that "the server" is located at a given IP and is a server. So then, response time can be found as a function of a server type and can be executed to send back results to the notify my phone execution which will be called after 20 minutes has passed based on the timer that was set up.

All the stuff comes together in ways like that. Still is limited though, it wont be able to do certain tasks until I add in the knowledge of how to complete them properly.

Cesque 10 years, 10 months ago

Quote:
and they all appeared to fake our intelligence, not become it

Yep, that's pretty much the modern approach to AI - although "simulate" is a better word than "fake".

This has also been a source of controversies within study of artificial intelligence, but in the end, a. there are no solid means of building or even defining "real" artificial intelligence, b. even if it were possible, it would not be as usable as an efficient simulation of it, c. there's always a chance a simulated process can lead to better modelling of the idea behind it.

Quote:
So first off, I made an AI(it was pretty good actually :D, had some nice conversations(should I post it?)), and it was good, but once again, it just faked intelligence.

What do you mean? A chatbot, basically?

Quote:
Although I would love to go by Alan turings idea of an AI, which is that if someone cant tell the difference between it and a human then it is therefore intelligent, but that dosn't seem right, at least not to me.

You're totally right - it's a theoretically useful "benchmark", but it has nothing to do with actually thinking (and you might just as well replace use of language with playing videogames - can you write a TF2 bot so good people will assume it's a player, etc.).

These days, you can generally "pass" the Turing test using modern approaches (Cleverbot does just fine), because technology has taken directions Turing probably couldn't have predicted.

Anyway, I wrote a rather long blog about AI before you had joined, so maybe you'll find something relevant there.

Quote:
Does anyone know of any theories that could explain things like this, or are we just chemical reactions and our current knowledge is correct?

The best non-religious approximation you can get is the difference between software and hardware. A computer requires physical elements (integrated circuits, etc.) to run video games, but a video game is not, in any useful sense of a word, a physical construction. It's an abstract program that's run by it.

JuurianChi 10 years, 10 months ago

We talk about this often, don't we?

What is it about Artificial Intelligence that is so appealing to us as human beings?

spike1 10 years, 10 months ago

If no one likes you its very useful to have an AI :D. Nah, seriously though its a good question. I wonder if it has anything to do with (Cant remember what its called, anyone know? :D). When we see something doing something, we try to project<?> human emotion and behavior onto it, which is why we might think a dog is happy, or a cat is looking at you funny, when really the dog is licking itself and the cat wants to attack you. Maybe its this phenomenon that makes AI so interesting, because in essence we are trying to project our emotions and knowledge into an inanimate object.

Another reason i can think of is the idea of living forever inside of an AI, because as we all know, humans don't want to die and if an AI can be created that can allow them(or a copy of them) to continue to live, they will be interested.

That's all i can think of, but for me the reason is that i want someone who i can talk to at any time and ask their opinion on a project, or have a discussion with them about programming or some obscure topic that most people know nothing about. For the short time i was in highschool(I homeschool now(Year 9 :D)), i made almost no friends because the only people there were sporty people, who know nothing about what I'm interested in(and vice versa). An AI for me would just mean someone/something that i can be friends with without having to limit what i talk about.

And cesque, yes it was a chatbot :D using a combination of a symbolic neural network, pattern matching an a few of my own ideas thrown in :D. It fooled my mum actually, told her it was an online chat system :D. Thanks for your ideas on AIs, i'm checking out your blog now :D. Would you be able to expand on your last paragraph though, because it seems like an interesting idea, but im having trouble drawing parallels between software and hardware, and our brain and our knowledge. I think i just forgot what i said in my original post though, so ill check back and see if i can make sense of it :D.