Hello all,

I have learned at the Univ. Neural Networks and Artifical Intelligence.
That was complicated ! But I never used it again...

Do you think AI has many uses in programming or is it useless and just a "nice theory" for brain masturbation ?

For instance AI is used in tools against spam like Bayesian filter. Do you think AI works and is efficient ? Or is it just a scientific toy ?

Thanks

If you look to the industry I'm sure you'll find some principals of AI in practice. My professor mentioned that genetic algorithms are used in engineering to find optimal parameters in a simulated system.

Since neural networks differ so much from the computing paradigm we are familiar with (learning rather than raw number crunching) it’s more difficult to find every day applications for them. They work on a higher level.

Bayesian filtering, like you said, is a great one. Streamlining human interfaces would be another great application.

AI would obviosuly be very useful if it really was Artificial Intelligence, but the current state of the art isn't all that great. No-one really understands how brains work. Neural network simulations are misguided IMO, because we need to work out how the brain works at the software level, not at the hardware level. You can't understand how a web browser works by looking at the patterns of electrical activity in the circuits of a computer, and it's equally naive to think that human intelligence reduces to simple patterns of connections between neurons. What's more likely is that neurons in the brain form highly complex structures which are able to perform computations rather like an ordinary computer, once you abstract away from the hardware differences.

Alex

Artificial Intelligence is no match for natural stupidity.

forget about AI's

just make something that can talk and see and has emotions will you!

and yes, our brain is rather nothing more then a blob of semi random circuits that change them selfs constantly to adapt to what you need (based on how you use the brain) so it becomes more efficient

Mostly they provide fodder for writing papers. You know, if you are a prof, you basically have to write. Junk papers are obligatory.

NNs are ok if you pick the right parameters for your particular problem. Otherwise, your problem might not converge to a solution, which is a tremendous limitation.

Depends on how broad your definition of AI is. It could be as simple as having an intelligent, usable interface for your software, or as complex as learning things and adapting to unexpected situations. There are a lot of things you probably learned in University that you'll never use again; that doesn't make them useless.

I know next to nothing about AI, but one concrete practical computing application of it is in games, especially games with strategy involved. I'm sure there will be plenty more applications as the field advances.

Here's an application for neural networks that might be of interest to this crowd:

http://www.eetimes.com/article/print...axonomyID=6385

Do you have any evidence for this? To use the same analogy as I did previously, a computer processor looks like a semi-random circuit until you understand how it works. Neural networks are reasonably good at noticing statistical patterns, but that's about it.

Alex

yes, but actually to simulate a brain you require a lot of computer, as our brain is divided into 3 main parts, the biggest one can be divided into 2 half's

each on does something different and prefers a different way to look at things (the right side is more about art, laziness, colors, .. all in all more object oriented ... while the left prefers numbers, language ,, kinda boring stuff.. more procedural stuff).

the first part of the brain (the brain that first gets info from the nerves) is responsible for more reflex's, and just wanting to stay alive... the next part influences emotions, and the last part (you know, the big one divided into 2 parts) is for higher thinking.. like having fun!

each part of each part of the brain in its won acts like a separate computer, and rewires itself based on what it does... the more math you do, the more the mathematical part of the brain adapts, and thus the faster math will get done, and how much math you can understand

one interesting note is that as parts of the brain are used , they often shrink in size compared to a normal person (this is because its being used far more efficiently, and thus doesn't require as much activity to do the same stuff)



now to get to semi randomness.... as the theory of evolution goes (what i wont say as its not relevant,... and i cant think of what i want to say... but tis good to reference this as so no one kills me for not believing in what some story says is fact)

as animals evolved, theres really no way for a organism to know how to evolve its brain, or any part of itself ... and thus random mutations to the DNA allowing for different proteins, and different ways for the embryo to grow and such take form... the only way for anything to get better is that the random mutations can do 1-2 of 3 things .. it can be bad and kill the organism, it can do nothing at all, it or it can help the organism live better then it did

as brains evolves it was random mutations that caused it to grow advanced, any misfortunes would most likely have killed the animal (well, today those people are locked in there bedrooms,... and what might be the evolutions of the brain are locked in mental homes) ... so over time with RANDOM mutations the DNA structure the brain comes to where it is today, as it helped us live. and thus the mutation lived ..

some other human species also evolved a while back... all where killed off (i forget if it was natural disasters, or if they just couldn't cut it that much)

so you see, the brain is a semi randomly wired biochemical computer capable of rewiring itself to adapt to our needs ... even the nerves adapt to our needed, when you do something over and over mailine (i think its misspelled) come over nerve cells, this stuff acts like a insulator, and signals are sent faster... i think the same can happen in the brain ... as nerve and brain cells are almost identical (well, nerve cells are long, where brains cells have lots of hooks to latch on to other cells as needed by it, but they are similar)

tho for all that brain power its amazing how little people actually learn ... don't study to hard in college or you wont learn a thing,... as most people cant remember more then 5% of anything they learn anyways, its best to study for a small time, then go play, then come back ... its proven to help people recall more ..... tho i don't have any experience as my memory is so bad i can forget something in seconds

now should we discuss emotions as well (this part can become very random, especially during PMS times for people (and yes, even guys can get PMS)

Sure, the brain rewires itself in response to how you use it, but it could do this in an extremely complicated way. We certainly don't understand how it does this, which is the important question. My computer changes patterns of electrical activity in response to its inputs, but we can't understand how this happens at the level of its electrical circuits.

Random mutations don't lead to random structures. The brain evolved through random mutations, but its structure is definitely not random. I suppose you could call it semi-random, in the sense that there are probably other ways it could have been wired up which would be equally effective, but all the evidence suggests that we are born with a lot of brain structure already in place. I agree that the brain rewires itself as it gets practice at certain tasks, but we have very little idea of how it does this; it's likely to be a highly complex process.

Anyway, I'm not sure how far we disagree. My point is that the brain is not just a semi-random network. Above the level of neurons, it's probably structured something like a conventional computer.

Alex

yes, we are born with all the basic structures in place (or else how would we live to pass genes on (you all know what that means )

and i do must feel the process for rewiring is a rather simple one at heart, don't forget theres about half a trillion brain cells (or was it half a billion?, i cant remember, and its only a quess as i cant remember the exact(tho enough) numbers very clearly, but half in half of about a trillion or billion was about right (i think it was trillion? tho, it was a few years or so ago (perhaps 3 years?))

anyways, at those high numbers, even a simple process would appear complex (don't forget chaos theory .. or just look at fractals, a simple design times a hew hundred times or so appears quite complex)

I completed a degree in AI a couple of years ago. With regards to this being any real use in the computing world - for me the answer is mixed.

Has it helped me get a GOOD programming job? the answer is NO. Had i stuck with simple old Comp Sci i feel this would be different...

As to uses of AI for commerce;

1) Every game on every console uses a massive amount of AI (and not just to control the opponent characters).

2) AI has transformed medical imaging - cancers, deformed blood vessels, etc can all be identified with AI Image Processing techniques.

3) The highest paid career bracket occupied by Mathematicians programming AI for Financial market prediction.

4) Ford style production lines are no run and work scheduled with AI techniques.

5) Genetic algorithms have been used to reduce the number of components in electrical circuits - the result was so novel that it defies the laws of physics!

6) Sony's albio dog robot, and Mitsubishi's walking robot...

7) Search engines are AI; document vector models, jaro fuzzy matching, etc.

8) Fuzzy logic has been controlling household appliances for over a decade. My washing machine proudly states its Fuzzy Controller on the front.

9) Government and military have invested heavily in AI, drive your car in/out of London (UK) and it'll be added/removed from a list of vehicles known to be in the capital. Speech recognition of keywords in telephone conversations flagging terrorism etc... Actually the BigBrother side of AI is very well funded.


We are talking mega bucks £$£ in the 1st point alone. AI really does have huge commercial application.

As a model of neurological processes - forget it. You won't be using any of the current AI models to construct HAL in your spare bedroom.

To understand why AI will probably never become mere simulation of brain function read Roger Penrose (Emperor's new Mind, and Shadows of the Mind).

for the experts here, is this AI?

http://alice.pandorabots.com/

AI has usefulness in Gaming, which, if you think about it, isn't really useful at all.