Computers As Good As Monkeys At Image Recognition

Discussion in 'Computer Science | IT | Networking' started by Smriti Jha, Jul 5, 2014.

  • by Smriti Jha, Jul 5, 2014 at 1:51 AM
  • Smriti Jha

    Engineering Discipline:
    Computer Science
    Computers are excellent at most tasks but there is one such thing at which they are consistently beaten by humans, Pattern Recognition. But it seems our trump card to supremacy may be soon slipping away. Brain Scientists like Charles Cadieu have been pitting computers against monkeys at key image recognition tests for a decade. Up until now, monkeys were doing at least ten times better than a machine, at what Cadieu calls the “neural substrate that makes recognition possible.” This was done by wiring sensors to a monkey's head and measuring how the neurons in the inferior temporal cortex reacted to images flashing on the screen.

    Siri-in-a-year.

    But Scientists at Massachusetts Institute of Technology have changed the game by using "deep learning" algorithms to improve the machine's guess at recognizing image. It is reported that the reported techniques have brought about almost a tenfold increase in the performance making computers “pretty equivalent” to monkey brains. Apparently, the answer to every other computational roadblock of the last decade comes in the form of a "deep-learning" algorithm, thanks to Geoff Hinton (who is now at Google). With its various contributions to consumer market as well as research, I think "deep-learning" deserves to make a certain Top 10 list.

    Source: Wired
     
    • Like Like x 3
    • Informative Informative x 1
    #1
  • Categories:

Comments

Discussion in 'Computer Science | IT | Networking' started by Smriti Jha, Jul 5, 2014.

Share This Page