Artificial intelligence - the rise of machines

[WiFO215]

I think the key here is control. If I control the machine with no brainpower limit in/of me, I'm still on solid footing. And I would have, of course, no reason to surrender that control.

Actually, from a personal standpoint, I think a safer thing to do would be to let that thing connect to my brain. The damn thing would become addicted to PF and sit in front of the computer the whole day. I'd have saved the world from being overtaken by machines. :D

 

[Kronos5253]

Yes, but then you can only do whatever your neural network is computing. So, if I make a robot that is controlled by a computer such that the computer simulates a neural network and uses the result of the computation to control the robot, then there can be no significant difference in behavior between the robot and a real person.


Of course, this is not a actical way to go about creating an intelligent machine. I think a practical way is to simply copy nature: Use genetic programming to create a brain ab initio without attempting to understand how a brain works. This is how our brains arose. There was never a programmer who set pout to design a brain at all. It is all a consequence of random mutations and selections.

So, all we need is a dumb but powerful computer.

For the first part, you still have to have a program that mimics our neural network, and that program needs to be scripted by SOMEONE. It's not just something you have at hand.

For the second part, you can create an intelligent machine that way, but it won't mimic a human brain/mind. And it still won't be able to do everything that the human brain/mind can do, mostly because of the way that the brain/mind receives information and stores it. Everything goes through the 6 senses (<- that's not a mistake).

I just personally don't see it as possible. But hey, we shall see :) lol

 

[mgb_phys]

You are comparing personal computers of today with the human brain. PCs as they exist today cannot even simulate the brain of a spider.

They are doing better in some AI fields.
They beat human players at chess - but not Go.
When I did this stuff, 15 years ago neural nets were better than humans at galaxy classification, 10 years ago they beat humans for cervical cancer cell recognition.
I don't know what state of the art is now but with terraflop GPUs it should be pretty good.

 

[Raizy]

Consider this about that statement though: When you use a computer to do mathematical calculations, you only have whatever programs you're using to do the calculation, and whatever else you might have running be it a music player, AIM, surfing the internet, etc. When you do mathematical calculations in your head, you have to sift through sensory information being constantly sent to your brain, all the processing to move your hands and eyes and every part of your body, and your own thoughts about what's going on around you and your current task. It would be the equivalent of running every ".exe" on your computer at once, then trying to have it do the calculation. Most likely the computer would shut down or freeze up because it would be too much. And that amount of information that you'd be running on the computer is nothing compared to what our brains do and can do.

I wonder if it's possible for a group of mad scientist to kidnap some genius or average joe, enslave him, hook up sensors onto his brain, install some mechanical parts in there and [insert some sci-fi gibbersh here], and get him to start predicting the weather.

 

[jobyts]

Human brains are not way too smart either. If you are reading a book on brain anatomy, does the brain know that it's learning about itself? no. it's just like learning anything else. If the brain were way intelligent, we didn't have to read from a book how brain works!

 

[b.shahvir]

Is it possible, or practical for a software programme to self evolve?

 

[Count Iblis]

Is it possible, or practical for a software programme to self evolve?

Yes, the genetic programming technique has been developed in the 1980s and has been applied with a lot of success in many different fields.

 

[b.shahvir]

Yes, the genetic programming technique has been developed in the 1980s and has been applied with a lot of success in many different fields.

Then in a way my concern is valid!