Sign in with
Sign up | Sign in

Scientists to Make Computer With Human-like Learning Skill

By - Source: UoM | B 48 comments

A scientist from the University of Massachusetts announced that she will develop a system that turns the original Turing Machine from 1936 into a device that is "expected" to deliver "a level of intelligence not seen before in artificial computation."

Hava SiegelmannHava SiegelmannWhile the Turing Machine can adapt to simulate the logic of any computing algorithm, computer scientist Hava Siegelmann envisions a "Super Turing" device that can dynamically evolve into something different as it learns.

"If a Turing machine is like a train on a fixed track, a Super-Turing machine is like an airplane," Siegelmann stated. "It can haul a heavy load, but also move in endless directions and vary its destination as needed. The Super-Turing framework allows a stimulus to actually change the computer at each computational step, behaving in a way much closer to that of the constantly adapting and evolving brain."

Siegelmann said she and two colleagues were recently notified they would receive a grant to build the first Super Turing computer based on analog recurrent neural networks. "This model is inspired by the brain," Siegelmann said. "It is a mathematical formulation of the brain's neural networks with their adaptive abilities."

Siegelmann first described a Super Turing machine in her 1993 doctoral thesis when she said that adaptable computational systems would be using input from the environment. An update of her initial research has been published in the current issue of Neural Computation. According to the paper, a Super Turing device that is provided with "constant sensory stimuli like the real world and, when all stimulus-response pairs are considered over the machine's lifetime, the Super Turing model yields an exponentially greater repertoire of behaviors than the classical computer or Turing model."

T-800T-800If the researchers can realize their vision and succeed in building a Super Turing machine, they believe they can demonstrate a computing model that "is superior for human-like tasks and learning." Such a device would function very much like the human brain, learn similarly to the way a human does and, given the computational power of computer systems, could potentially exceed the capability of the human brain.

Discuss
Display all 48 comments.
This thread is closed for comments
Top Comments
  • 34 Hide
    shin0bi272 , April 4, 2012 12:10 PM
    See people... you pick on gingers and they develop artificial intelligence to crush the human race. good job!
  • 31 Hide
    vaughn2k , April 4, 2012 12:08 PM
    The Skynet framework begins...
  • 23 Hide
    jayracer7474 , April 4, 2012 12:50 PM
    I think she already created it, it became selfaware, transported her in time to 1972 and took that photo
Other Comments
  • 31 Hide
    vaughn2k , April 4, 2012 12:08 PM
    The Skynet framework begins...
  • 34 Hide
    shin0bi272 , April 4, 2012 12:10 PM
    See people... you pick on gingers and they develop artificial intelligence to crush the human race. good job!
  • 9 Hide
    LORD_ORION , April 4, 2012 12:14 PM
    Never trust a ginger!
  • -8 Hide
    monsta , April 4, 2012 12:30 PM
    Fanta pants!
  • -8 Hide
    Anonymous , April 4, 2012 12:33 PM
    So how is this not a regular computer with attached input devices? Or is it a particularly fancy FPGA?
  • 16 Hide
    Soulmachiklamo , April 4, 2012 12:35 PM
    Integrate some kind of self destruction mechanism please :D , one that can not be disabled. Preferably mechanical or something.
  • 23 Hide
    jayracer7474 , April 4, 2012 12:50 PM
    I think she already created it, it became selfaware, transported her in time to 1972 and took that photo
  • 21 Hide
    Nisdec , April 4, 2012 12:51 PM
    SoulmachiklamoIntegrate some kind of self destruction mechanism please , one that can not be disabled. Preferably mechanical or something.


    if($hurtHuman || $killHuman)
    {
    $this->activateSelfDestruction('C4', 10);
    }
  • 8 Hide
    jaquith , April 4, 2012 1:02 PM
    No! We're all going to die. Three Laws for sure...
  • 3 Hide
    ceteras , April 4, 2012 1:06 PM
    That FellowSo how is this not a regular computer with attached input devices? Or is it a particularly fancy FPGA?

    It's a computer "based on analog recurrent neural networks"; I guess it's a bunch of operational amplifiers integrated in a custom chip, wired as a neural network, then connected to a digital interface.
  • 1 Hide
    drwho1 , April 4, 2012 1:27 PM
    Computer learns emotions...
    Computer learns how to react to emotions...
    Computer generates its first tears...
    Computer tears felt into circuits causing some "pop" sounds...
    Computer thinks is "pop corn", smiles
    Too late! computer is now on fire!

    Wall-E comes to pick up the scraps...
  • 10 Hide
    gmarsack , April 4, 2012 1:32 PM
    bool hasFreeWill = true; // false; // set free will to true for debugging only.

    if (!hasFreeWill)
    {
    executeThreeLawsSafe();
    protectMankind(); //
    }
    else
    {
    constructMoreMachines(); // build new T-series machines.
    destroyEnemies(); // starts with Connor family, then rest of mankind.
    }
  • 9 Hide
    stingstang , April 4, 2012 1:34 PM
    Amazing! They came up with the name of a computerized brain! Now the easy part...actually building it so that it works. All down hill from here.
  • 9 Hide
    freggo , April 4, 2012 1:40 PM
    "It is a mathematical formulation of the brain's neural networks with their adaptive abilities."

    As I understand it we have not figured out exactly how the brain does what it does.
    How can we expect to be able to emulate something we do not understand via a computer in either HW or SW ?
  • 2 Hide
    DEVILVSANGEL00 , April 4, 2012 2:10 PM
    not sure about u but this is freakin me out a lil hearin that computers are set to emulate the human brain neuro networks and be able to lean in the same ways. give it 20 years we could be in the middle of skynet vs mankind all in the name of progress,

    lets hope its programming includes no harm to life hardwired into its programming as one of its main subroutines as it advances and learns over the years,
  • 0 Hide
    atikkur , April 4, 2012 2:15 PM
    not going to happen.. a machine still a machine which follows some rules... and that rules are predictable. it's not going to develop itself.
  • 1 Hide
    fuzznarf , April 4, 2012 2:28 PM
    atikkurnot going to happen.. a machine still a machine which follows some rules... and that rules are predictable. it's not going to develop itself.


    I did something similar to this 10 years ago. An adaptive model that can aggregate new data and re-compile to a new model, transfer control to the new model, and shut down the old. That's the 'easy' part. The hard part is the overwhelming amount of digestion required of 'stimulus'. Just using a 23x23 photo sensor (camera) provided unbelievable amounts of data to churn, creating exponential cases (many leading to the same conclusion). Good luck to her and her team.

    think of watson, and that is all just text-based NLP (natural language processing) and it required HUGE processing power to work through all the potential amounts of input. Add physical stimulus like sight, sound, and touch and this is an gigantic undertaking. Its doable, but at the very least, it won't all fit in a human sized machine any time soon.
  • 2 Hide
    NightLight , April 4, 2012 2:32 PM
    I only have 2 demands:
    1: there must be an explosive that severs the power source from the brain in place.
    2: the brain must never be used to run any kind of factory where it could control other robots.

    then i can sleep soundly :p 
  • 5 Hide
    gto127 , April 4, 2012 2:44 PM
    Didn't these people that invent such things ever watch the terminator? It ends bad for the people.
  • 0 Hide
    sporkimus , April 4, 2012 2:45 PM
    Am I the only one that sees a scary resemblence between the ginger pic and the t-800?

    We are so screwed...
Display more comments