[Lieoften]

Alright, i have no clue where else to post this so imma post this here. I'm just curious as to what everyone thinks an Artificial Intelligence is made up of? at what point could you call something AI? Do you consider Cleverbot or video game enemies AI? etc etc etc.

[Ticon]

To a degree yes. I don't consider anything we have right now true AI as none of it is truly intelligent. the code and thus the human is. Although cleverbot may learn thing here and there, its just an algorithm to remeber common words and what not. As far as video game enemies go, definitely not. I would say just about every video game enemy runs off a script. Look at WoW for example, every enemy unit runs off a script. aggro, basic attack, fireball, basic attack if (hitpoits < 40%) run etc.

I'm not a good enough programmer to even think about designing AI so sorry I can't fill in that one

[supersloth]

your mom is artificial and also talked about a book* she read recently. so i'm going with that, your mom.

*"everybody poops", from her description, it sounds riveting.

[Lieoften]

supersloth, on 31 August 2012 - 11:59 PM, said:

your mom is artificial and also talked about a book* she read recently. so i'm going with that, your mom.

*"everybody poops", from her description, it sounds riveting.

really? A your mom joke?

[supersloth]

yes. really.

[Lieoften]

supersloth, on 01 September 2012 - 12:38 AM, said:

yes. really.

well played, sir. well played.

[Aphex19]

Have you ever heard of the Turing Test or the Chinese Room experiment? Alan Turing proposed that a computer should be considered intelligent if it can fool human beings in to thinking that they are speaking to a human. This test is notoriously difficult and (from what I can tell) most AI's fail miserably with proper testing. I mean, honestly, would Cleverbot (for example) fool most people under scrutiny? I doubt it. Perhaps. However, John Searl argued that even if a computer could pass the Turing Test, the computer still has no real understanding of what it's doing (it's just following instructions), so it shouldn't be considered intelligent. He demonstrated this using the Chinese Room experiment.

I kind of agree with John Searl. I think that a computer should only be considered intelligent if it can show understanding and reasoning ability, although I'm extremely cynical of any such thing ever happening with the classic computer system. Maybe interfacing organic matter with computers is the way to go? Cyborgs? Our brains work quite differently to computers, and lead to different strengths. Humans are great at abstract thinking and reasoning, whereas computers are great at just executing specific tasks very quickly.

edit:
An example of why Cleverbot would fail the Turing test most of the time, I think.

Quote

Cleverbot: Prove to me that you are human.
User: How?
Cleverbot: If you really are a human you know how to prove it here.
User: I can't prove it.
Cleverbot: That's interesting. Why can't you prove it?
User: I don't know how to.
Cleverbot: Why don't you know what it means?
User: Huh?
Cleverbot: Well maybe not quite there yet but if you're married then it's only a matter of time.

This post has been edited by Aphex19: 01 September 2012 - 05:57 AM

[modi123_1]

Current AI is made up of code.. typically. There are various areas of AI you can get into... take a look, in a book, and you can see quite a bit of the areas.

Then there's artificial life constructs.. boids, swarms, etc. Things that create complexity from simple rules of similar small automatons..

Though I am thinking you are trying to get to 'when do we make skynet/the matrix'... So you need to draw a bright line around "real" vs "fantasy" when splitting hairs on terms.


http://www.amazon.co.../dp/0137903952/

[Lieoften]

modi123_1, on 01 September 2012 - 09:05 AM, said:

Though I am thinking you are trying to get to 'when do we make skynet/the matrix'... So you need to draw a bright line around "real" vs "fantasy" when splitting hairs on terms.

actually i was thinking more along the lines of The Alpha from Red Vs Blue. I understand that we will most likely never achieve a Skynet style AI system or anything close to the Asimov robots, but something like Tony Starks JARVIS might be more achievable.

[WolfCoder]

Intelligence is the ability to solve various problems. Things like decision making, game playing. Artificial Intelligence would mean it was such Intelligence that was built and created by someone. Lots of people confuse it with Artificial Sentience/Sapience which is better described under Artificial Life which is life built and created by someone. The most useful applications of Artificial Life today are still simulations of many tiny, simple organisms.

Of course, you could use Artificial Life as an Artificial Intelligence and have the organisms solve a problem- computers are mimicking evolution to create new designs and engineer efficient things in various applications.

[Choscura]

What seems likely to me is that 'functional' AI will become prevalent as either a way of doing work, but the pressure to make it "User-Friendly" (social in a way that is useful to humans, able to understand, account for, predict and - I bet -utilize human impulses). If this is at all accurate, it seems like AI will be like a pet, like an animal companion, that has as many bodies as you let it into, and at the same time actually is a discrete entity when it occupies a device- since that would be the most efficient distribution- but an increase in bandwidth would let the related instances re-absorb, to allow for the variety of input be available to all parts as future reference.

of course, individual 'personal' AI will not be as comprehensive as a centralized big one would be, so it may be that individual AI are 'hobby' machines and that the big ones have limited access to personal info but a much broader range of experience, and that could end up as something like 'google chauffeur', where lots of supervised and unsupervised machine learning data can be used by an AI service that can drive you around for $X per month.

This will be fun to watch.

[modi123_1]

Quote

I understand that we will most likely never achieve a Skynet style AI system

Hey pal - screw you! Don't go pissing on a man's life work before he's dead. Humanity ending artificial intelligence just happens to be one of this guy's pillars of a successful life though a lasting impression on the world.



Well baring any sort of catastrophic engagement, children, wife, unexpected riches, girlfriend sans restraining order, or being struck with a life long muse of writing and painting. God willing, one day you, all will get a polite knock on your door before it is blown off the hinges by a Personal Automated Mollifying Intermediary (or PAMI for short) who will read that hour's personal message, from me. As you are cowering in terror and fear on your futon filled with bits of wood and plaster you'll smirk and go "aww.. modi123_1 did it! That fancy bastard did it!".

A short assessment later and you'll be either be outfitted with appropriate bio interfaces (for when they plug you in to the network) or disintegrated into a pile of green goo. Just hope you are the first few houses for that morning as drill bits for the interfaces tend to get a bit dull after cutting through that much tissue and bone. On the bright side even if the edges are a bit dull the pain from the lack anesthesia is leagues greater than being conscious as each and every cell is ripped from the other, popped, and liquified through the disintegration ray.

[Lieoften]

Aphex19, on 01 September 2012 - 06:45 AM, said:

Have you ever heard of the Turing Test or the Chinese Room experiment?

There was a test a couple years ago (i was going to post this in my original reply to you) where the Cleverbot actually fooled more than half of the people it interacted with: http://www.newscient...t-is-human.html

[lordofduct]

WolfCoder, on 01 September 2012 - 12:15 PM, said:

Intelligence is the ability to solve various problems. Things like decision making, game playing. Artificial Intelligence would mean it was such Intelligence that was built and created by someone. Lots of people confuse it with Artificial Sentience/Sapience which is better described under Artificial Life which is life built and created by someone. The most useful applications of Artificial Life today are still simulations of many tiny, simple organisms.

Of course, you could use Artificial Life as an Artificial Intelligence and have the organisms solve a problem- computers are mimicking evolution to create new designs and engineer efficient things in various applications.

Thank you for posting this, as I was about to.

OP, get your terms correct. AI already exists, it was one of the first things we coded when we made our first computers. Even before with chess playing automatons (that's AI).

Artificial sentience on the other hand (or what ever vocab you want to use to describe it)... well it's hard to say what it's made up of because we don't know. That's one of the main reasons that we've yet to create what we would consider sentience in a machine.

This realm gets into a lot of philosophy about what IS sentience, how do we define if something is self aware, and how to test that. Turing's test has already been cited, but some don't consider that a suitable level of sentience... we've tricked people numerous times into thinking a machine was sentient with games and the sort.



Personally I consider there to be a few issues in this, the primary being what people expect "AI" to be like. Being the self obsessed chauvanists we are, we expect humanoid, or at least animalistic sentience. But why would a computer have such style of sentience if it's connection to the world around it... it's survival insentives, it's IO senses, everything about it is completely different.

So I think the start of creating sentience would be setting up a need for such sentience.

Ability and necessity to die
Necessity to collect resources to offset death
I/O sensory (replicate animalistic sensory organs like eyes and touch... or highlight cybernetic versions)

This gives the robot a world to act in, and a reason to act with in it. Then starting handing it the basic tools to problem solve. Give it "reason" through simple things like memory... the ability to store a snapshot of it's sensory inputs, and to connect a positive or negative value to it (a help or hurt value).


This system could mimic existing real-world systems.

A robot that needs collect natural resources to operate it's motor. Say it runs on hydrogen, and it uses eloctolysis to separate water into hydrogen and oxygen and burns said hydrogen to fuel itself. So it now has the need to hunt down water to survive... if it can't find any, it dies. It gets cameras to 'see', sensors to feel 'wet', and microphones to 'hear'.

This can result in actions that we would easily recognize as sentient.


It could also follow its own rules...

A program that feeds on special kinds of data. It gets an internet connection as its sensory organ. And it has to mine data from the internet to feed.

The problem with a system like this is limiting resources (data is easily dup'd), furthermore it's hard for us to recognize anything it does as sentient. We have nothing to compare it to... we may just see an action it takes and assume it did it because that's what it was programmed for.

But really... isn't that what ALL life does? It does that because that's what it was programmed to do.

This post has been edited by lordofduct: 02 September 2012 - 08:21 PM

[NecroWinter]

Lieoften, on 31 August 2012 - 10:49 PM, said:

Alright, i have no clue where else to post this so imma post this here. I'm just curious as to what everyone thinks an Artificial Intelligence is made up of? at what point could you call something AI? Do you consider Cleverbot or video game enemies AI? etc etc etc.

Thats kind of a difficult question. We dont even have a real understanding of what intelligence is. Video game AI often just reads controller input.

My opinion on the AI issue is to program Axioms and allow genetic algorithms to generate VERY BASIC personality traits and preferences, and to allow the machine to learn.

The reason I say that is, thats how we start out. We have our genetic personality traits that are given to us, and then we learn by experiencing life, when we experience life, we see certain events as positive or negative and that shapes our understanding. I think AI should do the same thing.