so what the hell is AI, anyway?

[exdeath]

When your programming only describes the unconcious low level electrical/chemical functioning of a neural node (inputs, outputs, data storage), and you build a computer from these nodes and have to teach it like a child higher level functions that are not pre-programmed or known in advance by a single node... that is AI.

Anyone who thinks it's possible any time soon underestimates the power of a human brain or overestimates our technological state. I mean seriously, we *still* rely on moving substrate covered in metal dust to store and retrieve data... our technology really sucks.

A human child of age 3 can glance across the isle at a page of a magazine he/she has never seen before with a picture of a tree and exclaim "kitty!" at a blob of dithered pixels 2 cm across in an instant. Something a multi trillion dollar super computer still cannot do even with purposefully written algorithms. One is two feet tall and powered by a pint of applesause and some carrot sticks, and the one that can't do it is the size of a house and powered by half a city's electric grid and is good only for the one thing it can't even do.

The non volatile data storage and retrieval speed requirements alone trump our current technology. We sure as hell aren't getting there with our current primative serial 100 MB/sec hard drives. We will need massively parallel instant access multi gigabyte per second FeRAM/MRAM type technology before we can even think about coming close to creating anything resembling an artificial brain.

 

[Cogman]

I would say we have reached the peak of AI when we can teach AI to do stuff the same way we teach children to do stuff.

In otherwords, once a computer can learn how to sculpt just by reading about sculpting, observing sculpters, and trying sculpting itself, that is when we have truly created AI. It is the point where programmers make themselves obsolete.

 

[iCyborg]

Are you two talking about artificial human intelligence/brains or just artificial intelligence? Computers can do some pretty intelligent stuff nowadays, often seeing patterns where even people have problems noticing them.

The ultimate goal would be creating something smarter than us which could then also make smarter and smarter versions of itself. This is known as technological singularity.
http://en.wikipedia.org/wiki/Technological_singularity

 

[Paul98]

Just look how far computers have come in the last 20 or 30 years. In another 20 or 30 years we will have computers thousands if not millions or billions of times more powerful. With some of the breakthroughs we are having in quantum computing, vertical transistors, graphene, we might have another huge jump in computing power.

If we are able to make the computer self aware, or able to improve on it's self and solve problems. After that the growth should come even faster, after that things could change in a major way.

 

[Aikouka]

A human child of age 3 can glance across the isle at a page of a magazine he/she has never seen before with a picture of a tree and exclaim "kitty!" at a blob of dithered pixels 2 cm across in an instant. Something a multi trillion dollar super computer still cannot do even with purposefully written algorithms. One is two feet tall and powered by a pint of applesause and some carrot sticks, and the one that can't do it is the size of a house and powered by half a city's electric grid and is good only for the one thing it can't even do.

You gave an example similar to this in Off Topic, and I disagreed with it there as well. It sounds odd, but I think you're over-analyzing the situation. When you give me a situation like that, I actually attempt to simplify the problem at hand, which honestly... is how you write computer programs to begin with. So, you start off with asking yourself, how does the child go from looking at the magazine cover to exclaiming, "kitty!" to begin with?

The question itself is quite loaded, because I'm asking what is sort of an impossible question. The reason is because while we don't think of our brains like functions, but that kid had to have used some piece of data to formulate his response... some sort of variable. But what makes it hard, is... what is that variable that would define that response? Also, if you were to present that child with a similar variable (and never the original), would the same response occur? With that questioning, I'm essentially asking what occurs within the function.

I'm pretty sure that I talked about it above (it's been awhile), but I believe that we're heavily associative in our processing. We tend to associate shapes, colors, smells and other stimuli to events and even to other stimuli. Using the same example of a tree, could you confuse a child by using a peach tree and a red apple tree. I'm fairly certain that you could. I can't be 100% sure, because I've never tried it. 😛

But anyway, what I'm trying to get at is these lower associative functions that are mostly involved with processing stimuli are probably the "easiest" bits of functionality to implement. I mean... what would it take to create an AI that could formulate the ideas in this post that I just wrote? I still believe the idea of step-wise decomposing the way we think is probably the most abstract thing we could ever do!