• 1
    • 2
    • 3
    • 4
    • 5

You are here:HOME >> News >> Industry News News

The future of artificial intelligence will be towards a generalization

[Date] 2016-10-18 14:20:09   [Come From]www.powerlinksz.com  

  Definition of artificial intelligence is not difficult, but it is not possible, it is not because we do not understand the human mind. Oddly, the progress of the artificial intelligence will help us to define more human intelligence is not what, rather than defining what is artificial intelligence?
  But no matter what is artificial intelligence, in the past few years we have indeed from the machine vision to a lot of progress has been made in a broad range of areas, such as playing a game. Artificial intelligence is from a study subject to early enterprises to adopt. Companies such as Google and Facebook have invested a lot of bets on artificial intelligence, and the application of this technology in their products.
  But Google and Facebook just started: over the next ten years, we will witness the spread of artificial intelligence in one product. We will communicate with the Bot - they are not scripted robot dial-up program (robo - dialer), we can‘t even aware that they are not human. We will rely on automobile for route planning, to respond to dangerous road.
  It is no exaggeration to estimate that over the next few decades, we have contact with each kind of application will be integrated into some artificial intelligence function, and if you use the application, we will not be able to do anything.
  In view of our future will be inevitably is tied with artificial intelligence, we must ask: how is our development now? What is the status quo of artificial intelligence? Where are we heading?
  Now the power and limitations of artificial intelligence
  Description of artificial intelligence that surrounds the centre or the following: the intensity (how smart), width (solve the problem of range is narrow, and the problem of generalized), training (learning), capacity (what can solve the problem) and autonomy (artificial intelligence is assistive technology can only act on their own). There is a range of these each center, and each point of the multidimensional space are represented the understanding of artificial intelligence system goal and the ability of a different way.
  In the center of the intensity (strength), can easily see the results of the past 20 years, and realize that we have created some very powerful program. Deep Blue (Deep Blue) in chess defeats Garry Kasparov. Watson (Watson) beat the Jeopardy of winning champions; AlphaGo beat is arguably the world‘s best go chess player lee se-dol.
  But all this is limited. Dark blue, Watson and AlphaGo are highly specialized, the purpose of a single machine, can only on one thing done well. Dark blue and can‘t play chess, Watson AlphaGo can‘t play chess or participate in Jeopardy, even the most basic level. The intelligence of their range is very narrow, also cannot be generalized.
  Watson has made a lot of progress in various applications such as medical diagnosis, basically, but it is still a must for quiz machine specially modulation in specific areas. Dark blue with a large number of specialized knowledge about chess strategy and encyclopedic knowledge of open. AlphaGo is built with a more general framework, but it still has a lot of manual coding knowledge in your code. I am not despise or underestimate their accomplishments, but realize that they have not do is also very important.
  We haven‘t been able to create can solve the problem of a variety of different types of artificial general intelligence (artificial general intelligence). We haven‘t listen to the tape of a year or two human dialogue machine can speak for themselves. Although AlphaGo through the analysis of thousands of matches and then more game and go under the "society", the self but the same cannot be used to master chess program.
  The same general approach? Maybe you can. But we are the best achievement is far away from the real general intelligence also -- true general intelligence can flexibly unsupervised learning, or can be flexible enough to choose what you want to learning content, whether it‘s playing board games, or PC board design.