The "Freeola Customer Forum" forum, which includes Retro Game Reviews, has been archived and is now read-only. You cannot post here or create a new thread or review on this forum.
Had a lecture yesterday about neural nets. Some of you may know what these are, some may not. For those that don't, here's a quick run-down - based on my understanding.
Nueral nets are a network of neurones. Funny that. A neurone has many inputs, and only one output. Each input is weighted, in that input 1 may have a weighting of 10, and input 2 has a weighting of 1. Therefore, input 1 is much more important than input 2. With me so far? Right.
If you have a neurone with 10 inputs, the threshold level may be, say, 50. This means that when the inputs are multiplied by their respective weightings, if the results is over 50, then the output is active. Otherwise the output is not active.
Here's a slightly simpler example:
A neuron has 3 inputs, with different weightings:
Input 1 has a weighting of 11.
Input 2 has a weighting of 6.
Input 3 has a weighting of 5.
The threshold level is 10. So, if input 1 is active, then the output is active as the threshold level is reached. If input 1 is not active, then the threshold level can still be reached if both input 2 AND input 3 are active.
If you're not with me, go back and read it again. It does make sense. Promise :-)
People are trying to write computer programs that can represent complex neural nets, as these nets can learn - essentially true Artificial Intelligence.
So what happens in the future when (and it WILL happen, just maybe a few hundred years from now) we create proper AI. Will they have legal rights? Would it be right to give them dangerous jobs? Only allow them certain thoughts and feelings?
Were is the line drawn between man and machine?
This is a difficult question, when you consider that the average human brain is comprised of around 100 BILLION neurones. But if neurones merely react to their inputs (like a computer program), then we have no control over them. Therefore we have no control over our actions.
Because of this we are not guilty of anything. Nor can we take credit for anything. We are just machines. Effectively.
Thoughts? (if you can call them that...)
Nueral nets are a network of neurones. Funny that. A neurone has many inputs, and only one output. Each input is weighted, in that input 1 may have a weighting of 10, and input 2 has a weighting of 1. Therefore, input 1 is much more important than input 2. With me so far? Right.
If you have a neurone with 10 inputs, the threshold level may be, say, 50. This means that when the inputs are multiplied by their respective weightings, if the results is over 50, then the output is active. Otherwise the output is not active.
Here's a slightly simpler example:
A neuron has 3 inputs, with different weightings:
Input 1 has a weighting of 11.
Input 2 has a weighting of 6.
Input 3 has a weighting of 5.
The threshold level is 10. So, if input 1 is active, then the output is active as the threshold level is reached. If input 1 is not active, then the threshold level can still be reached if both input 2 AND input 3 are active.
If you're not with me, go back and read it again. It does make sense. Promise :-)
People are trying to write computer programs that can represent complex neural nets, as these nets can learn - essentially true Artificial Intelligence.
So what happens in the future when (and it WILL happen, just maybe a few hundred years from now) we create proper AI. Will they have legal rights? Would it be right to give them dangerous jobs? Only allow them certain thoughts and feelings?
Were is the line drawn between man and machine?
This is a difficult question, when you consider that the average human brain is comprised of around 100 BILLION neurones. But if neurones merely react to their inputs (like a computer program), then we have no control over them. Therefore we have no control over our actions.
Because of this we are not guilty of anything. Nor can we take credit for anything. We are just machines. Effectively.
Thoughts? (if you can call them that...)
Page:
VenomByte wrote:
> Ah, but neural networks can be trained. You are responsible for the
> conditioning of your own brain, aren't you? After the age of eighteen,
> at least. If you chose not to give yourself any morals, you're
> punishable for that.
Punishable for not giving yourself any morals? Surely life would be SO much easier to live if you didn't have any morals?
> This is a potentially very dangerous way to make AI, don't you think?
Possibly. But it will happen someday. It'll just be about 4 hundred years after I've died...
> Ah, but neural networks can be trained. You are responsible for the
> conditioning of your own brain, aren't you? After the age of eighteen,
> at least. If you chose not to give yourself any morals, you're
> punishable for that.
Punishable for not giving yourself any morals? Surely life would be SO much easier to live if you didn't have any morals?
> This is a potentially very dangerous way to make AI, don't you think?
Possibly. But it will happen someday. It'll just be about 4 hundred years after I've died...
It's not the sentimentality that annoys me, or the Robin Williams bit, it just seemed like a film I'd seen a million times before, but with robots in it.
Its not a lie - its schmaltzy and sentimental, as are most Robin Williams films.
RM18 wrote:
> It's a classic Robin Williams film
Lie! Lying is bad! But I am gonna nick the Msn addy...
> It's a classic Robin Williams film
Lie! Lying is bad! But I am gonna nick the Msn addy...
Humans are complex beasties, I'll give us that.
Instinct (if that is the neurone network etc) and all that.
I just like what happens when they work, not why.
:)
Instinct (if that is the neurone network etc) and all that.
I just like what happens when they work, not why.
:)
Ah, but neural networks can be trained. You are responsible for the conditioning of your own brain, aren't you? After the age of eighteen, at least. If you chose not to give yourself any morals, you're punishable for that.
Another interesting thing about neural networks, it that it is impossible to understand why they reason the way they do. You can build one, train it, and test it. You can harp on all you like about the weighted values, but at the same time you haven't the foggiest why this input really gives that output. You can't follow their reasoning.
This is a potentially very dangerous way to make AI, don't you think?
Another interesting thing about neural networks, it that it is impossible to understand why they reason the way they do. You can build one, train it, and test it. You can harp on all you like about the weighted values, but at the same time you haven't the foggiest why this input really gives that output. You can't follow their reasoning.
This is a potentially very dangerous way to make AI, don't you think?
No problemo. It's a classic Robin Williams film, but well worth a watch. Panned by the critics, but I thought it was ok. Tell me what you think of it when you watch it - my MSN is [email protected]
Ah-ha.
I may have to look into it. May get it out over the weekend. Cheers.
I may have to look into it. May get it out over the weekend. Cheers.
Basically, its a story of the future. A robot droid, highly intelligent (played by Robin Williams) and capable of living a life far surpassing that of humans, becomes a servant for a wealthy family. He talks and can perfom technical feats, and over time (200 hundred years, hence the film title) he gradually becomes more and more human like, getting human skin, learning to love, even forming a relationship with a human, to the point where there is a court case over what consitutes being human.
Not really what the post is about, but I recommend renting it out. Underneath the sentimental mush, there's some thought-provoking issues directly related to some points you drew up, like the divide between man and machine, and how that may be blurred in future as AI is developed.
Not really what the post is about, but I recommend renting it out. Underneath the sentimental mush, there's some thought-provoking issues directly related to some points you drew up, like the divide between man and machine, and how that may be blurred in future as AI is developed.
Nope.
Page: