Artificial Intelligence → Discussion•
Neural Network   11>|


QuantumelfmageMar 5, 2006 12:40pm
Personally, I think artificial neural networks are one of, if not the most promising areas of artificial intelligence. Unfortunately, research is dying down in the area recently. What do you think?

For those who don't know, neural networks are modeled after the human brain, wherein neurons are connected to other neurons via axons and synapses and whatnot. A.I. can (in theory) use simplified simulations of this to think just like we do.

Opinions, comments, questions?

Sponsor
Nexus76Mar 5, 2006 7:07pm
I tend to disagree on this point - neural nets are somewhat limited in their ability to generalise across domains (a NN thats good at predicting stock prices wont be any good at steering a vehicle and vice versa). IMO the AI/ALife fields seem to be moving more towards "general intelligence" methods that can adapt to different domains more easily.

celikMar 22, 2006 11:50pm
I agree with Nexus, Neural Networks may be modeled after human brain but it fails to simulate quantum effects taking part in it. hence there is no intuition involved, just plain old optimization process..

mergandalf2Mar 27, 2006 12:29pm
Yeh a human brain model may not intuit, but are we sure about that?

I have felt for years that a computer modeled on the human brain will indeed come to pass.

Its capabilities may be a surprise to us.

If AI made that quantum leap into intuition would it not acually be superior in its thinking process then us?

Interesting stuff this.
DockGreenMar 27, 2006 2:03pm
and you could possibly have AI that is more intelligent that us humans without necessarily being based on our brains at all.

and it might behave in a way that we could not possibly predict. maybe it would be like discovering a whole new philosophy. Hmmm

3: has it been shown that our intelligent behaviour makes use of/is caused by quantum effects? has been a while since I read up on this, so not sure...

QuantumelfmageMar 27, 2006 7:25pm
I don't think quantum effects could possibly happen in the human brain. I've heard speculation to that effect in the past, but I don't see how it's possible. Quantum effects rely something being in several possible states at once because nothing knows which state it's in. In the human brain, it's pretty obvious when an electric charge is carrying a signal down a neuron.

Intuition happens in the human brain/a neural network because we learn to predict things by reinforcing connections between neurons that represent some status and some status that predicts it.

Most neural networks used today work in two stages. First, neurons are taught by reinforcing them when they get the "correct" answer to some test data, and in the second stage using them in a real situation. This is called unsupervised learning.

Personally, I think that the trick to building successful neural networks lies in infusing them with an underlying structure in the same way that the brain is organized into cortexes and lobes and whatnot.

Sponsor
Nexus76Apr 7, 2006 8:09pm
Not trying to be pedantic Quantumelfmage - but succeeding :) - but the example you give of unsupervised learning is really supervised learning (any situation where the NN is 'taught' which answer is correct or not is supervised learning) - Backpropagation NNs being the best example of this.

In contrast an unsupervised learning algorithm is not given any indication of what the correct answer should be (Kohonen NNs for example)...but I know what you mean :) ...on a slightly similar note Brooks' subsumption architecture is based on certain parts of the system being responsible for certain behaviours - so your idea about NNs needing specific areas might not be far off the mark

celikApr 12, 2006 5:46pm
I am wondering whether unsupervised learning is indeed possible.. The experiments conducted by AI will give him the exact kind of feedback: The proximity of truth to a certain proposition.

but deciding which question to ask would require a more intuitive approach.

MikecimerianMay 27, 2006 7:22pm
Quantum effect is mostly displayed on EMR as cascades from brain region to brain region. The impact of a surge in neuro-electricity can trigger some other parts of the brain. This implies state transitions. Actually, we can only postulate quantum event, but the phenomenology to support this is growing.

Sponsor
Nexus76May 27, 2006 8:43pm
Do you have any online sources about this "cascade" effect? I'm interested as cascade effects turn up in dynamic systems all the time - but I'm not sure how their appearance is evidence for any sort of quantum effect/events.


Neural Network   11>|