How do we learn? And in the same way can a machine learn?

[Moni]

How do we learn? And in the same way can a machine learn?

 

[dpm]

Do you mean can a machine learn at all, or can a machine learn in the same way that we do? If it's the first, then the answer is yes, up to a point. There's a whole subsection of the field of AI dedicated to machine learning.

 

[Mk]

How is sleeping related to how quickly one can attain a new skill? I remember a Time Magazine article saying that was the new theory or something. What sleep does? Ugh I'm tired.

 

[Moni]

Umm...I want to know about the edge technology or development made so far!

I need that to do some paper works :)

 

[selfAdjoint]

There are (at least!) two meanings to the word learning. One is the simple kind of learning that simple animals do, like conditioning. For this we have a pretty good understanding, it's based on the strengthening of a neural connection when it it is excercised repeatedly. A scientist named James Hebb had this neural idea back around 1950, and this kind of learning is called hebbian. The neural networks idea in computer software is based on it.

The other kind of learning is the complex and subtle kind we do every day. On that we have some good research but no deep understanding yet.

 

[Genecks]

Check this out: http://en.wikibooks.org/wiki/Intoduction_to_Psychology:chpt7"

I never finished the wikibook.

 

[RunDMC]

A computer can learn but not comprehend.

There's a theory that a complete and accurate simulation of an average brain would naturally produce intelligent thought, but there's no way to test this with current computers. (There's another theory saying there's bugger all intelligent thought on Earth anyway, see the PF Lounge ;) )

 

[Moni]

haha...Yeah RunDMC .. but who's that Bugger? ;)

thanks Bio-Hazard :)

 

[Moni]

There are (at least!) two meanings to the word learning. One is the simple kind of learning that simple animals do, like conditioning. For this we have a pretty good understanding, it's based on the strengthening of a neural connection when it it is excercised repeatedly. A scientist named James Hebb had this neural idea back around 1950, and this kind of learning is called hebbian. The neural networks idea in computer software is based on it.
The other kind of learning is the complex and subtle kind we do every day. On that we have some good research but no deep understanding yet.

Hmm, I'm talking about that other kind of Complex and Subtle idea :)
How's that can be studied in machine? I think I've to look on Psychology journals for that :(

 

[Intuitive]

A computer can learn but not comprehend.

That depends on how well the programming routine is (tree'd)

The only limit is ones own programming ability.

 

[neurocomp2003]

no you need to look in ALife/AI journals and Neural Net journals for what you want. It also depends on the learning..pseudo learning (ie Game AI-liek deep blue) or realistic learning liek the Blue Brain PRoject by IBM and swiss lab.
MIT has some really neat Computer vision as well as york university.
And there are industrial learning algorithms that use principals of th brain for industrial stuff like Cryptography and Satellite imaging and medical imaging.

Also look at steve grands Lucy its kinda cool or look up gary flakes book.

 

[Moni]

That depends on how well the programming routine is (tree'd)
The only limit is ones own programming ability.

how about an intelligent computer system...who itself can write program for itself ;)

 

[Moni]

no you need to look in ALife/AI journals and Neural Net journals for what you want. It also depends on the learning..pseudo learning (ie Game AI-liek deep blue) or realistic learning liek the Blue Brain PRoject by IBM and swiss lab.
MIT has some really neat Computer vision as well as york university.
And there are industrial learning algorithms that use principals of th brain for industrial stuff like Cryptography and Satellite imaging and medical imaging.

Also look at steve grands Lucy its kinda cool or look up gary flakes book.

can you provide some link on them?

Cause, I'm about the design any new learning model...so, my approach is to first know the human/animal learning process and find something new...so that it can be any new model...what you suggest?

Do, I still have to look into NN and Fuzzy stuffs?

I've gone through few AI books...and their learning algorithms are too specific and purpose specific...why don't they talk about something general?

 

[Mk]

can you provide some link on them?
Fuzzy stuffs?

http://en.wikipedia.org/wiki/Fuzz" 's some Fuzzy stuffs.

 

[Genecks]

A computer can learn but not comprehend.
There's a theory that a complete and accurate simulation of an average brain would naturally produce intelligent thought, but there's no way to test this with current computers.

Yet comprehension could be described as an emotion triggered by new neural passage ways being opened or else altered. It is a sensation. The thing is, robots can't feel emotion which makes them different than us. They might be able to display the neurological change, but to feel it.. hmm...

Coming back to this post later I think about how computer learning is more of a conditioning more than of an actual learning process. Else and if statements I propose.

The greatest thing in the Universe is want, and if you can program a computer to want something, or create an objective, it will learn to complete that mission and then cease to exist until given a new objective. They thing one must do is have the A.I. want everything. We as humans can't simply define everything for we don't know everything. To tell a robot to learn everything would probably be process of elimation from old knowledge to new knowledge in which it goes forth trying to learn new material and constantly redefine its world.

 

[Moni]

Nicely said man :)

WANT!

That means QUESTIONING :D

 

[Intuitive]

how about an intelligent computer system...who itself can write program for itself ;)

That depends on how well the programming routine is (tree'd)
The only limit is ones own programming ability.