The Lounge is rated Safe For Work. If you're about to post something inappropriate for a shared office environment, then don't post it. No ads, no abuse, and no programming questions. Trolling, (political, climate, religious or whatever) will result in your account being removed.
Have you ever used neural networks i did, they dont learn they are just a mathematic code that adjust the output based on previows input-output examples and that is if the problem you want to solve can be mathematically modelled if not then neural networks is useless.
My friend real inteligence cant have a mathematicall model so it cant be programmed.
I agree to some extent with both of you Vasily and Collin.
I think(well is more like i hope or wish) humanity will be able to create intelligence as good as normal human being at least, posibly something beyond that, but im sure that it wont happen in our lifetime.
Collin Jasnoch wrote:
Vasily Tserekh wrote:
My friend real inteligence cant have a mathematicall model so it cant be
Collin Jasnoch wrote:
How are you so certain?
Show me inteligence that can not be "mapped" as you say, and I will show you how maybe you just are not inteligent enough to map it.
That one is easy, free will cant be mapped/modeled, an algorithm is based on a set of rules. You can say that you can create a/"set of" rule to create new rules, but then again is just another rule that cant be broken. free will is the part of intelligence that make us able to break any rule to adapt or get better.
OK, dont know if you are trying to confuse me, playing dumb or it is a genuine question, i will suppose the last one.
Lets stay with the example you gave of current AI, the social networks algorithms.
It is based on a set of rules, lets say(for the sake of simplicity) if the user if from north america and is male, the algorithm will "decide" to show a beer adverticing.
Now of course the algorithm may take tons of rules i used only 2 because i want to keep this simple.
Now if the one who had to decide what adverticing must show to the users was a human being, he may decide to show other adverticing although he was only instructed to only take into account the location and gender, he may also take into account new paramters withouth being told to do so, like age, politic, religion, etc.
So with this simple example(maybe not the best) just what im trying to say is that a part of our intelligence is the capability to break the rules, which i called free will(maybe not the best translation because english is not my native language).
I hope i had made my point at least little bit more clear.
OK i think i understand what you are saying, and yes maybe the term free will is not correct, but to me it seems the best way to call it.
Im not an expert in Neural networks, but i have used them several times at college and yes they can learn, but unless you are talking about a new type i dont know or heard about, the structure of a neural network is static, what changes with the training is the weight of the path the impulse runs through between 2 neurons, and yes after they are trained they can be trained again to "learn" something diferent.
But as far as i know what they cannot do is to change the input, i mean if you create a huge neural network that works with an input of lets say 25 parameters, it will always take into consideration 25 parameters, it cant grow bigger than that, in the example the NN takes 2 parameters into consideration, if it was designed that way, doesnt matter how many layers it has it wont take into account any other parameter, for that being posible you would need a new NN.
So what im trying to say is that what i dont see posible in AI as it is right now is the adaptability our human intelligence has to get new parameters into account(at least talking about NN, im sure there are plenty of other models/theories in AI that i havent heard about).
Well then all this was on my assumption that there is no NN that has the ability to mutate/evolve(with this i mean that the same NN, can change to be able to receive a new input it wasnt designed for) without being redesigned, if exists a model or something i dont know please just tell me the name of that model or theory so i can read about it and agree completely with you.
what you said only occurs because the NN has no way of getting the 3º parameter.
As a human, you can't get some parameters on your own. Parameters like, let's say, the radiation level on the ambient. you depend on a separate equipment that it's not "built in" your body, so you need a Geiger counter. If no one is available, then you can't take the radiation level in account.
If you can provide a way to the NN to get parameters that you don't predict that it will use, then i see no reason for the NN to be unable to learn to use them, even with the actual technology.
I'm brazilian and english (well, human languages in general) aren't my best skill, so, sorry by my english. (if you want we can speak in C# or VB.Net =p)
cmon people intelligence its not making something that can recognize a face, a fish can do that and fish are not inteligence, inteligence its to CREATE for example E=MC2, and i still say taht a for loop and if and switches cant achieve that, about the cocaine was a joke
Until we can break away from the Turing model, we will never be able to achieve AI. Although you could argue that we humans are also Turing machines. Very complex in nature but still Turing machines nonetheless and very predictable.
Free will is not as free as you may think. Free will is a decision that is reached by analyzing your current environment (hormonal balances and current blood pressure etc taken into this account as well), processing the current data and measuring the outcome against the cost of achieving the preferred goal and making a decision based on this threshold.
Free will is an extremely complex mathematical algorithm.
Survival instinct. Base programming. Call it what you want, there is no doubt about it. When you are hungry, do you not eat? That is a goal. Do you communicate with others to better define requirements? That is a goal. Do you go to work every day to make a living? That is a goal.
If AI has the goal of 'please the master', that is the same as 'I need to procreate to ensure survival of the species'. Who put that goal in your instincts? Hmmmmm?
The very existence of free will is still hotly contested in neuroscience, psychology, philosophy, and theology. Even theoretical physicists weight in on the controversy every now and then.
If I take your definition of free will, which is not a universally accepted definition, what you are really describing is a system that is non-deterministic. While difficult, it is possible to create programs that are also non-deterministic so I would argue that, by your definition of free will, a program is indeed possible which has free will. A lot or research into emergent behavior has gown down this path. Even if one wanted to just simulate what you call free will, all one would have to do is insert a rule which says that all other rules can be broken (e.g. a statistical weight driven system).
Map this: "Human General Intelligence" (a.k.a Common Sense) , seriously, i believe he refers to the fact that we still don't have Terminator like intelligences running around us and i believe is not by the lack of tools, but of the lack of processing power.
Yes, young Padawan, neural networks are just a mathmatical model of the neurons that hopefully make up the grey mass between your ears. Training neural networks by the traditional feedback approaches has been found limited in many ways. Defining the topology and weights of a neural network to make it suitable for any task is an entirely different matter. Lazy people have simply tried to let one of the most powerful search algorithms[^] do that job and the results are really promising.
The gray mass between your ears has been configured by the same algorithm, with the tiny drawback that it had begun to do so many millions of years before any of our ancestors was able to climb a tree and pick some fresh fruits there. Once that you have understood all that, then you may come to realize that the problem is not finding an adequate emulation of neurons or how 'mechanical' they appear to you. It's the complexety of the desired result that will make us take a little more time than we would like.
Ok, here we go again. I forgot once more that you are the one and only authority on those things. You have looked at it, once again not seen the forest because of all those trees and therefore it's all just a myth.
At least artificial intelligence already is superior to natural stupidity
I'm of the opinion intelligence can be made without free will and even that free will be impersonated by applying the full set of human emotions to decisions, all the AI's experience, and a slight degree of randomness (possibly using a seed generated from the parents'). Having a full amount of override control in any decision is a good start to free will. However, people are always scared to admit free will can be created, as it seems to detract from the superiority of being human.
I think that applications COULD be intelligent, they certainly gain data naturally, however could this be considered knowledge? Knowledge is an acquaintance with facts, truths, or principles, as from study or investigation. Can current AI study or investigate a principle, truth or fact?
A social network could be seem to gain knowledge of a user's topical preference, however a social network does not naturally learn new skills. It is given new skills by a developer.
Though I certainly don't think that AI is a myth, I think that with current technology it is probably out of our reach (for now).
To what extent? I might have said the same thing after my one required undergrad class on the subject. Now that I'm just a thesis away from a masters concentrated on machine intelligence, I have a slightly different perspective.
First, I don't believe that the limits of AI are inherent to the techniques themselves. Rather, AI is limited by the ability of us humans to provide effective mathematical models and enough hardware to to evaluate them, both of which are constantly being improved. Consider that it took about three BILLION years for the human brain to develop its current algorithms and resources. I'd say AI is progressing at a fantastic rate by comparison!
Secondly, you denigrate the tradition program control structures (if, while, for, etc), but are you really so sure our brains work any differently?
Last Visit: 31-Dec-99 18:00 Last Update: 30-Jul-21 11:32