What Are the Key Components and Capabilities of AI Beyond If/Then Statements?

[Simfish]

Does AI consist primarily of if/then statements? I know that AI could also include random behaviors based on the "internal flip of the coin" as well. But what else can AI do?

 

[Math Jeans]

I think AI does function off of if/then statements, as well as different actions to repeated events. Basically if/then statements that count up. In fact, it would be hard to deny that humans function off of if/then statements. Basically, there has to be an event for there to be a reaction. It is fairly hard to think of an action that cannot be converted to an if/then statement.

"If I don't want to look straight ahead, then I look in a different direction."

"If my sister makes more sound, then I get more angry with her."

Just my opinion.

"If someone starts a thread on AI using if/then statements, then type opinion in reply."

 

[NateTG]

Does AI consist primarily of if/then statements? I know that AI could also include random behaviors based on the "internal flip of the coin" as well. But what else can AI do?

From a decision making perspective, all existing AIs can be implemented on touring machines. (Google/Wikipedia for more on what that means.)

AI is a relatively vague term, so it would be helpful if you could elaborate a bit on what you mean by AI.

 

[billiards]

From a decision making perspective, all existing AIs can be implemented on touring machines. (Google/Wikipedia for more on what that means.)

If yout meant Turing machines, I'm pretty sure I once read a book by Roger Penrose that would disagree with that. I believe it was called "The Emperor's New Mind".

 

[NateTG]

If yout meant Turing machines, I'm pretty sure I once read a book by Roger Penrose that would disagree with that. I believe it was called "The Emperor's New Mind".

You'll note that I said 'existing' AFAIAA Penrose argues that natural intelligences (such as some humans) require super-Turing quantum-mechanical properties in that book.

 

[Alkatran]

No, A.I. is not just a bunch of if-then statements, at least, no more than any normal program is. Look up neural nets, genetic algorithms, computer chess, etc.

As for consciousness depending on quantum mechanical properties... can't we model those on a computer?

 

[Aether]

As for consciousness depending on quantum mechanical properties... can't we model those on a computer?

Not on a conventional computer...at least not exactly for systems larger than about 30 electrons. --http://www.dwavesys.com/index.php?page=quantum-computing

...consider the modeling of a nanosized structure, such as a drug molecule, using conventional (i.e., non-quantum) computers. Solving the Schrödinger Equation (SE), the fundamental description of matter at the QM level, more than doubles in difficulty for every electron in the molecule. This is called exponential scaling, and prohibits solution of the SE for systems greater than about 30 electrons...This restriction makes first-principles modeling of molecular structures impossible, and has historically defined the boundary between physics (where the SE can be solved by brute force) and chemistry (where it cannot, and empirical modeling and human creativity must take over).

Quantum computers are capable of solving the SE with linear scaling exponentially faster and with exponentially less hardware than conventional computers...Results of these virtual reality simulations will be indistinguishable from what is seen in the real world, assuming that QM is an accurate picture of nature.

This type of simulation, by direct solution of the fundamental laws of nature, will become the backbone of engineering design in the nanotech regime where quantum mechanics reigns.

 

[Alkatran]

Not on a conventional computer...at least not exactly for systems larger than about 30 electrons. --http://www.dwavesys.com/index.php?page=quantum-computing

I wasn't talking about tractability, I was talking about computability. It doesn't matter, philosophically, if you can actually do the simulation. All that matters is if it is, in theory, possible. Consciousness slowed down by a factor of 10^10^100 is still consciousness.

 

[Aether]

I wasn't talking about tractability, I was talking about computability. It doesn't matter, philosophically, if you can actually do the simulation. All that matters is if it is, in theory, possible.

To which scientific theory of consciousness are you referring?

Consciousness slowed down by a factor of 10^10^100 is still consciousness.

Only to the extent that there is some objective measure of consciousness to begin with.