what is considered AI?

Started by
15 comments, last by Sneftel 18 years, 1 month ago
hi, im new here, im sorry, im just confused and i did not want to post a stupid question like it says in the forum rules, i was just wondering if, it is possible to lets say manipulate a program, like make some kind of program to do what "I" would do, like if im at work, and i make a program to do all my jobs, and it requires telling the program to press buttons and put in numbers and what not, __would this be considered artificial intelligence___?, because i was thinking AI was something that thinks on its own, this program would be doing exactly what i tell it, and all events are included, and no random activity, all activity solely depends on what has happend last... thanks, im sorry again, im a newbie
Advertisement
I could make an argument that you don't really think on your own. We're just machines whos external behavior is too complex to infer much about internal processes, but still totally deterministic.

As for an attempt at your actual question, artificial intelligence is just intelligent behavior shown by something, well, artificial (ie. a computer). Yes, it's vague, but such is life.
[size=2]
There's no clear and standard definition, even among researchers. AI tends to focus on making complex decisions based on real-world data, but this is by no means always or even usually the case. "Learnable" algorithms are common, but much of AI consists of non-learning algorithms, and of algorithms which are hand-tuned. The layman's definition of AI would usually be restricted to simulations of human activity such as chat-bots--and, in particular, simulations which encompass human traits such as randomness and fallibility--although modern researchers dislike this characterization.

I'd say that what you're talking about sounds more like automation than AI, because it doesn't seem to involve any complex decision-making.
One difficulty is that it's not the task that is artificial intelligence, it's the program. So you can't really qualify it as "AI" or not by the result or output of the program.

The problem is that a lot of real world tasks that are done by "artificial intelligence" programs, are currently better done by analytical methods anyway (perhaps what Sneftel says is automation) and so things are a little fuzzy.

What happens is, as soon as you can develop an AI with sufficient "reasoning" or complex decision-making, you understand enough of the task being solved that you can apply some analysis and simplify the AI. When does it stop being an AI and start being an automation or automated program? The line is hard to draw.
Think about it, a good AI program is a non-programmable one, which you neither give it orders nor goals to complete, and no pre-defined algorithms as well. Goals are defined and completed by itself.

When it comes to the real AI development, ask yourself some questions: What makes human human, and how comes our most advanced computers not as 'intelligent' as some basic creatures, does an arbitrary being needs any pre-defined solid 'algorithms' to exist ?.
--> The great thing about Object Oriented code is that it can make small, simple problems look like large, complex ones <--
Quote:Original post by Skeleton_V@T
Think about it, a good AI program is a non-programmable one, which you neither give it orders nor goals to complete, and no pre-defined algorithms as well. Goals are defined and completed by itself.

"Good AI"? Heh. Again, this is an area where there are as many definitions as you want. I'd note, however, that your definition of "good AI" does not really seem to be "useful AI".
That is an unproven theory that I'm working on, while quick-and-dirty AI algorithms produce acceptable results in some meanings, it againsts the learning nature of human. As a result, these algorithms may seem to be useful at first, but at harsh speech, they're incompleted imitations of human knowledge.

Without evidence yet though, I'll just have to agree with what you said [wink] and keep silent.
--> The great thing about Object Oriented code is that it can make small, simple problems look like large, complex ones <--
Why would the sole purpose of AI be to imitate human intelligence? Being good at things which people suck at is one of the main strengths of AI.
Quote:Original post by Sneftel
Why would the sole purpose of AI be to imitate human intelligence? Being good at things which people suck at is one of the main strengths of AI.


Because we(in general) make ourselves out to be the highest form of life, ever,(not including those who believe in a god nor those that really take the time to think about things realisticly), just the general opinion) we model thing off ourselves; all the time, why not AI also?


To the original poster:
There is a test I read about in a sci-fi book, which I am not sure whether it is an actual test or not, but the AI programmers test all their robots or programs by putting them on a computer and letting them chat with another person(say the janitor who knows nothing about the robot/program) and watch and see whether the Janitor or whoever realizes it's not a human. I realize that it isn't the best test in the world, I just thought it was interesting.
Quote:Original post by Servant of the Lord
There is a test I read about in a sci-fi book [...]

I think you mean the Turing Test.

This topic is closed to new replies.

Advertisement