EvilZone
Programming and Scripting => Projects and Discussion => : gh0st February 14, 2012, 04:21:36 AM
-
what about a program that will help you in tasks as memorize important dates, that marks the number by saying the name of the person phone, and it will learn stuff not by touching its source code but by speaking and while the time passes it will learn more complex stuff :) you could carry it on your iphone or laptop what else I dont know but its a good begining
-
and what use would that be?
-
It would be like your secretary
-
Hm...I don't really understand what your saying. I think its something like this: you have a program which you can talk to and as you talk to it, it will learn, right? :P Well either way, I had an idea like this, where the program would learn through conversation and not by you feeding it with dictionaries and the like :D
-
idea as an idea is kinda good I suppose, but looking from a practical viewpoint it looks like a useless and not worth the struggle to make it...
-
maybe once I get my compter scientist degree I will be able to aford a humanoid robot like nao robot and then I will upload all the data learned that would be epic
-
This notion is kind of pointless, get a piece of paper and a pen and write your shit down. Eventually you'll learn to remember it all without being lazy a lazy ass. Not only that but how would you even code the network needed for it to learn what's important to you.
I say you should think this over.
-
So something similar to iphone's Siri? (http://www.apple.com/iphone/features/siri.html (http://www.apple.com/iphone/features/siri.html)) It's not AI though, but doing almost the same thing.
-
maybe once I get my compter scientist degree I will be able to aford a humanoid robot like nao robot and then I will upload all the data learned that would be epic
Yes, once you finish getting your 'compter' degree you will be able to afford siri by that time something better than this will have been made. Stop making excuses if you like the project try and do it. You probably won't succeed but you'll get a good lesson from it.
-
yeah I need a degree of electronics at least a master to make a project like that but its not impossible
-
well this doesn't sound like 'impossible' to me at all, in fact, if you use the appropriate libraries it might be a very achievable task.
You need:
Voice Synthesis
Data Management
And for learning 'more complex stuff' (I don't exactly know what you mean, but I'm assuming that you mean changing its own source code and behavior) you need Artificial Neural Networks.
And a bit about Smart phone app development if you wanna develop one for iPhone/android phones.
I wasn't sure about what you meant, but if it's simply like a secretary that does some certain stuff such as storing contacts (something like Siri), it's done through Voice Synthesis + Data Management. However, it's already made: Siri!
-
And for learning 'more complex stuff' (I don't exactly know what you mean, but I'm assuming that you mean changing its own source code and behavior) you need Artificial Neural Networks.
ANNs do not change their source code and whereas they are good for speech recognition in this case I don't think they are suitable for the rest and they are definitely not the general purpose tool for machine learning.
However, this thread is four month old, so I won't go too deep into it. Just this: Start with a chatbot and improve it over time. The Natural Language Processing part is the most difficult for this project and research still has to do a lot in this field. So you will definitely get to your limits. A poor assistant can be done fast, a really smart one can be a project for a life.
I don't see how the nao robot would be of any help with this.
-
and what use would that be?
I can see incredible use of such a program... think about the possibilities, you have facebook account with 200 "friends"... what you do is tell your secretary to manage a list of connections and then send a "happy birthday" message to all those that are not important enough to do it personally....
how about the secretary keeping a list of whos who... like you heard it today that a friends uncle is a lawyer and you ask the secretary to note it down... later it searches the online profile of said friend notes the name of said uncle and then look it up on the net or site like linkedin...
let us say you have to buy something from the market but you keep forgetting it when you visit the place... whenever you are at the market the secretary pops up and reminds you to buy the thing...
how about the secretary keeping tabs on the stocks you have bought and then keeping tabs on the targets and selling or buying more whenever needed..
how about the secretary noting all conversations you have on a date and then deducting what the other person would like to do on the next date? like lets say the person you date says that she likes the color black and you might not remember it but the secretary suggests you to wear black when you are going to meet her..
how about the secretary analyzing the facial ticks of a person live or online during chat and determine whether what he is saying is true or false...
What if the secretary looks through all the websites of hotels in a city you wish to visit and determines the best one for you... by analyzing the distance, the prices and reviews of users on third party websites..
all this would require a very very intelligent program that I don't think is possible as of now...
-
So you mean like J.A.R.V.I.S. (http://marvel.wikia.com/Just_A_Rather_Very_Intelligent_System_%28Earth-199999%29)
-
So you mean like J.A.R.V.I.S. (http://marvel.wikia.com/Just_A_Rather_Very_Intelligent_System_%28Earth-199999%29)
something like that, yes. Something that acts as an extension of my own will, like a third hand which I do not have to think and move but works naturally like my own two do...
-
ANNs do not change their source code and whereas they are good for speech recognition in this case I don't think they are suitable for the rest and they are definitely not the general purpose tool for machine learning.
They don't change their source code, but they DO change their behavior, as if they're changing the source, since source code defines the behavior.
Machine learning is highly dependent on ANNs currently for they're the most known algorithms for developing a machine learning behavior. I'm open to other suggestions tho.
-
Machine learning is highly dependent on ANNs currently for they're the most known algorithms for developing a machine learning behavior. I'm open to other suggestions tho.
Just google and you find plenty of other concepts that are widely used and known.
This is just the tip of the iceberg: https://en.wikipedia.org/wiki/List_of_machine_learning_algorithms (https://en.wikipedia.org/wiki/List_of_machine_learning_algorithms)
Edit: Maybe I am blind, but I miss the genetic algorithms in this list.
-
Just google and you find plenty of other concepts that are widely used and known.
This is just the tip of the iceberg: https://en.wikipedia.org/wiki/List_of_machine_learning_algorithms (https://en.wikipedia.org/wiki/List_of_machine_learning_algorithms)
Edit: Maybe I am blind, but I miss the genetic algorithms in this list.
yes genetic algorithms are categorized under another field, so-called "DNA computing (http://en.wikipedia.org/wiki/DNA_computing)" (If I'm not mistaken)
I also found a book regarding this subject, called Machine Learning for Hackers (http://www.amazon.com/Machine-Learning-Hackers-Drew-Conway/dp/1449303714/ref=sr_1_3?ie=UTF8&qid=1339132954&sr=8-3), a quick look and I found Statistics as the main subject, and R language as the main tool being used, which is unlikely, in my point of view.
I'm used to using MATLAB for these purposes... I don't know maybe I should change the tools I use?
-
yes genetic algorithms are categorized under another field, so-called "DNA computing (http://en.wikipedia.org/wiki/DNA_computing)" (If I'm not mistaken)
No, they are not. DNA computing actually uses DNA to perform operations like splitting DNA strings, merging them and so on. It is not really something you can use, unless you have a laboratory. And it is not an algorithm, it is a kind of computer architecture.
Genetic algorithms on the other hand you can implement with every programming language you like. The algorithms mimic evolution, which is especially useful if you have to optimize something, that can't be done in the usual ways. You won't find an optimum with this approach, but you can get a good solution that is close.
I also found a book regarding this subject, called Machine Learning for Hackers (http://www.amazon.com/Machine-Learning-Hackers-Drew-Conway/dp/1449303714/ref=sr_1_3?ie=UTF8&qid=1339132954&sr=8-3), a quick look and I found Statistics as the main subject, and R language as the main tool being used, which is unlikely, in my point of view.
I'm used to using MATLAB for these purposes... I don't know maybe I should change the tools I use?
What exactly is it you want to do?
-
I have thought about this myself but not as a Secretary but as a programming partner. It would take lots of programming and a really advanced AI to start with so that while you are programming it can learn its own coding and therefor make it self more advanced when it learns something new. this is something that u can use to run a whole system by it self when your trying to run more then one computer the things it could do are endless.