Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Artificial Intelligence and other data processing technologies

  1. Jan 14, 2012 #1
    Hello everyone.

    We all know that conventional data computing technology is not sufficient to make complex mathematical calculations in some sense and we need to look at some other data computing technologies. I did some reading but could not deduce any answers for the following questions. Anybody interested please help.

    Scientists say, they need Artificial Intelligent system to make accurate judgement. So, they are shifting to Artificial Neural Networks to achieve that accuracy. It's a different thing that they found out new ways to deal with Artificial Intelligence in a conventional way to achieve this accuracy. But let's talk about Artificial Neural Network.

    My questions are interlinked.

    Question 1: Like we all know all living organisms have DNA. DNA carries information and determine the traits of species. DNA is biological, contains molecules, and there are four types of nucleotide s in DNA. And arrangement of these nucleotide vary from one living creature to another, which defines the traits of a living thing.
    So, in molecular computing (Molecular computers), if the scientists create a data computing processor which is not made of silicon but DNA, does it really mean that the scientists will be using biological DNA to make processors and DNA "constructed" is not going to be some sort of artifact made of electronic components? It's a silly question, but I am finding it hard to believe this. Because, if the scientists really make a processor using biological DNA how can they manage to keep the DNA alive? Will it have a survival period?
    Question 2:(please see the figure)
    On the right side of the figure. I know that there should be a system that should process all the non-linear signals/responses. I don't intend to raise any questions about the details of this system and the way it processes non-linear signals but my question 2 is to know what this Artificial Neurons in Artificial Neural Network is made out of. Because, some video on internet says "Artificial Neural Network is VLSI implementable." And a different source says that "ANNs (Artificial Neural Networks) are usually just computer programs with a network of data structures, not an actual physical network". Please note the word "usually" in the last sentence. Therefore, I want to know, is this Artificial Neural Network a hardware or software or both!

    Question 3: I think that Quantum computers are most powerful of all computers then followed by Atomic computers then Molecular computers then Photonic computers. Right? And the concept of data processing in Photonic computers can be used in any of the previously mentioned types of data computing technologies. Right? - The concept that I am refering to above is the way how photons are used to send data from one location to another, if the need arise. So, I can incorporate this concept in any/all of those above mentioned data computing technologies. Right? And if there is some processor which is completely made of some material that processes light signals directly that will be requiring less heat dissipation mechanism that will be well and good. I need more information on this subject. Please suggest any books, websites. Feel free to write any terminologies associated with A.I and these questions. I am going to use internet to learn.

    Please post concrete information and let's not make this thread look science-fictional.

    Your posts are valuable to me.

    Thank you.
    Last edited: Jan 14, 2012
  2. jcsd
  3. Jan 18, 2012 #2
    Somebody answer the Question 2 please, I don't care even if your answer is inaccurate but provide me the reasoning why you think of your answer as correct. Let's start a discussion. I love discussions. Thanks.
    Last edited: Jan 18, 2012
  4. Jan 18, 2012 #3
    I don't know anything about question 1.

    Question 2:

    ANNs are usually implemented in software. They're usually implemented as matrices for efficient computation. ANNs used in production environments are NOT simulations of how neuroscientists think the brain works. Traditional ANNs are just "inspired" by neuroscience, and are very simple.

    There is ongoing research that tries to more or less simulate the neocortex. Problem is that neuroscientists still aren't exactly sure how the brain works. Here's two off the top of my head:

    Numenta NuPIC - Commercial. Goal is to provide a system usable in industry. Not an exact simulation, but much closer to the real thing than traditional neural networks.

    HMAX - Academic. A model of the visual cortex that is fairly faithful to findings in neuroscience.

    Of course, all software can be implemented in hardware, and there is ongoing research in hardware neural networks as well, including analog implementations.

    It's also worth noting that neural networks have kind of fell out of favor in industry. Other machine learning techniques such as Bayesian networks and support vector machines perform better in many circumstances.

    Question 3: I don't know much about that stuff either. All I know is that nothing practical from any of those mentioned technologies has been produced yet :)
  5. Jan 18, 2012 #4


    User Avatar
    Science Advisor

    Hey pairofstrings.

    In terms of machine learning, what machine learning tries to do is take data of some sort and generate a model of what is to be expected in the hope that the model will converge to one that represents the system as more data is collected.

    From this, most machine learning algorithm will take a probabilistic approach in doing this. This is the basis of machine learning: build a model that converges to something that is accurate. Convergence in machine learning is a huge subject and is by no means trivial or in some senses, even gauranteed.

    ANN's do the convergence part by using weights. What happens is that you provide what is called 'training data' to 'train' the network that helps build and refine the weights.

    ANN's are not the only method used in learning, but they are definitely a major one.
  6. May 16, 2012 #5
    Hi everyone, my name is Leo my background is in Biology and i need your help in researching a topic that deals with AI. I'm writing a script for a film and the question i have is: "Is it possible to create an Artificial Intelligence that can operate globally and can it be altered ?"
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook