No announcement yet.

Alpha Go

  • Filter
  • Time
  • Show
Clear All
new posts

  • Alpha Go

    Documentary about the Deep Mind Challenge match from 2016

    Deep Mind is a computer program that plays the board game go.

    For a while it was thought that it would be a long time before a computer could beat a human at this game.

    I certainly was surprised to find out how short that long time turned out to be.

    Google bought the english company Deep mind that pioneered the program

    Fascinating watch.

    Originally posted by youtube comment
    "Move 78", that will be the name of my resistance movement when the machines takes over.

  • #2
    To grasp the gravity of this consider that chess, played on a 8 x 8 board could be brute forced to defeat Gary Kasparov back in 1997.

    The trouble with using that approach with Go that uses a 19 x 19 board is the number of possible configurations exceeds the number of atoms in the universe so > 10 to the power of 38

    A million computers would be at the job for a million years.

    Clearly, a different approach is required. One that can learn from expert players and create openings on it own.


    • #3
      I really enjoyed the documentary.

      But now the team have gone onto to make a breakthrough in the CASP competiiton, A bi-annual comp on predicting 3d structures of proteins from their amino acid sequences. Its been one of the holy grails in biology for decades. We have had limited success with human driven methodologies, but it been taking years, is expensive and doesnt always work. Aphafold can now do it quickly, cheaply, and with a higher success rate. Its one of those breakthroughs were it is difficult exactly to predict the implications but they could be truly massive and many.... in developing and directing research for humans.

      Another thing to ask yourself is rate of improvement. How much farther and how quickly can alphafold progress from here? Can alphafold eventually start making preidictions about how proteins interact with each other and in their environement....

      The chinese also definitely achieved quatum supremacy last week at room temperature. Looks like there is little doubt unlike the debated google efforts last year. They have also developed a second platform using photons/light and mirrors. We are going to have multiple chances at finding a workable quantum computing platform that could work with A.I. to solve previously unsolvable problems.


      • #4
        Here's Alphafold

        Protein folding problem is how to figure out the structure or shape of the protein given its sequence of amino acids.

        Knowing the structure means you understand its shape and this gives an insight into its function

        Proteins can be seen as machines carrying out specific tasks. Thousands of them in the body. These machines are noiseless.

        Last edited by Double Edge; 25 Dec 20,, 10:38.