1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

News IBM announces neurosynaptic software system

Discussion in 'Article Discussion' started by Gareth Halfacree, 8 Aug 2013.

  1. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,130
    Likes Received:
    6,718
  2. edzieba

    edzieba Virtual Realist

    Joined:
    14 Jan 2009
    Posts:
    3,909
    Likes Received:
    591
    Sounds like one of Thinking Machines 'Connection Machine' supercomputers compressed down to a chip. I wonder if their chip has any significant advantage to simulating a SLNN using RAM and a modern fast sequential processor iterating over the net (other than the operations happening truly in parallel, which is irrelevant unless the entire system is asynchronous).
     
  3. dogknees

    dogknees Minimodder

    Joined:
    27 Dec 2008
    Posts:
    103
    Likes Received:
    5
    Does no one remember Turing? The structure doesn't matter, it still can't do anything that a "normal" computer can't. Though it could be faster at some things.
     
  4. dogknees

    dogknees Minimodder

    Joined:
    27 Dec 2008
    Posts:
    103
    Likes Received:
    5
    I think they're using a LOT more very simple cores/cells that the Connection Machine. Still, the connections become the biggest part of the chip when the number of cells goes up, particularly with a lot of connections between cells. Guessing they would be using some sort of multiplexing buses with local, on-chip and between-chip sets.
     
  5. Gradius

    Gradius IT Consultant

    Joined:
    3 Feb 2009
    Posts:
    288
    Likes Received:
    1
    They keep saying it "mimics the brain" and stuff like that, but in reality the nature did a way such it will never be possible!
     
  6. edzieba

    edzieba Virtual Realist

    Joined:
    14 Jan 2009
    Posts:
    3,909
    Likes Received:
    591
    Nonsense! Brute force method: molecular simulation of the volume of neural tissue you want. More elegant solution: parametric neurons with diffuse neurotransmitter simulation. Most SLNNs for the last few decades are just parametric models.

    That is, unless you subscribe to Dualism and think the brain has some sort of magic intangible goop that lets you think.
     
  7. desertstalker

    desertstalker What's a Dremel?

    Joined:
    7 Feb 2004
    Posts:
    73
    Likes Received:
    0
     
  8. dogknees

    dogknees Minimodder

    Joined:
    27 Dec 2008
    Posts:
    103
    Likes Received:
    5
     

Share This Page