Results 1 to 6 of 6

Thread: Genetic Algorithms, anyone?

  1. Top | #1
    Content Thief Elixir's Avatar
    Join Date
    Sep 2012
    Location
    Mountains
    Posts
    16,529
    Archived
    707
    Total Posts
    17,236
    Rep Power
    61

    Genetic Algorithms, anyone?

    If anyone here would read or has read this now ancient account on the work of Dr Adrian Thompson, I'd love to hear your thoughts.

    For those unwilling to read the whole history, Dr Thompson used a simple (10 sectors x 10 sectors) FPGA (field programmable gate array) and tried to get it to discern a 10kHz tone, and to distinguish it from other signals. To do this. he started with a random configuration and made random changes to the program, saving those iterations that showed any different output when fed a 10kHz tone vs a 1kHz tone.

    Finally, after just over 4,000 generations, test system settled upon the best program. When Dr. Thompson played the 1kHz tone, the microchip unfailingly reacted by decreasing its power output to zero volts. When he played the 10kHz tone, the output jumped up to five volts. He pushed the chip even farther by requiring it to react to vocal “stop” and “go” commands, a task it met with a few hundred more generations of evolution. As predicted, the principle of natural selection could successfully produce specialized circuits using a fraction of the resources a human would have required. And no one had the foggiest notion how it worked.
    It's that last bit that fascinates me. The random mutations not only eventually solved the problem, but did so in a manner that could not be understood by those who designed the genetic algorithm, even in retrospect.
    Dr. Thompson peered inside his perfect offspring to gain insight into its methods, but what he found inside was baffling. The plucky chip was utilizing only thirty-seven of its one hundred logic gates, and most of them were arranged in a curious collection of feedback loops. Five individual logic cells were functionally disconnected from the rest⁠— with no pathways that would allow them to influence the output⁠— yet when the researcher disabled any one of them the chip lost its ability to discriminate the tones. Furthermore, the final program did not work reliably when it was loaded onto other FPGAs of the same type.
    Curiouser and curiouser.

    evolution had not merely selected the best code for the task, it had also advocated those programs which took advantage of the electromagnetic quirks of that specific microchip environment. The five separate logic cells were clearly crucial to the chip’s operation, but they were interacting with the main circuitry through some unorthodox method⁠— most likely via the subtle magnetic fields that are created when electrons flow through circuitry, an effect known as magnetic flux. There was also evidence that the circuit was not relying solely on the transistors’ absolute ON and OFF positions like a typical chip; it was capitalizing upon analogue shades of gray along with the digital black and white.
    In conclusion, the article tentatively posits that "adaptable alternatives may one day strip away ... limits and lay bare the elegant solutions that the human mind is reluctant⁠— or powerless⁠— to conceive on its own."

    I'd say that's already happened. When I first read of this decades ago, it gave me an almost religious reverence for the process of evolution, as a transcendent force with powers vastly exceeding that of the human brain -so much so that it actually can perform "miracles" in the sense that the solutions it finds right in front of scientists in the full light of day, cannot be understood even by humans' most rigorous examination.

    Thoughts?

  2. Top | #2
    Veteran Member
    Join Date
    Nov 2017
    Location
    Layton, UT
    Posts
    2,505
    Rep Power
    17
    Yeah, there have been a number of fascinating, and mind boggling, results of using genetic algorithms to solve problems. Real world physical solutions that, like your example above, we don't really quite understand and would have not likely designed on our own. One of my favorite examples is a tiny, lightweight antenna that wound up being use on at least one satellite/spacecraft (I forget the details).

    I am also fascinated by the flip side of this field. Similar in practice an effects, simulating evolution by code or computer models that are then let to run. Some very unexpected results have occurred. One of my favorite (now very old by internet standards as well) examples is from the Secular Web library: https://infidels.org/library/modern/...oretierra.html

  3. Top | #3
    Veteran Member
    Join Date
    Dec 2017
    Location
    Land of Smiles
    Posts
    1,068
    Rep Power
    15
    Wow!! I did a little work with straightforward software GA before I retired many years ago, but never anything as exciting as these two linked articles. (Isn't the weirdness in the FPGA example just explained by negative feedback which will, of course, be device dependent?)

    It's very "old hat" by now but I thought Danny Hillis' 1990 work on predator-prey GA was interesting: This GA operates to optimize TWO species concurrently, predator and prey. Hillis used this to find a new optimal sorting network (which no human had found) for N=13. Googling "Danny Hillis predator prey" gets lots of hits. e.g. https://arxiv.org/pdf/1506.05082.pdf . (For a short while I hoped to find an improved N=17 sorting network, but made no progress.)

  4. Top | #4
    Contributor
    Join Date
    Nov 2017
    Location
    seattle
    Posts
    6,613
    Rep Power
    20
    I'd have to understand the starting point and what random variations mean in the context. Is it purely logic or is it a processor.

    I Understand FPGAs, don't know what n x n sectors mean. Linux runs on an FPGA configured as a processor.

    In engineering genetic algorithms have been around since at least the early 90s used as a form of optimization.

  5. Top | #5
    Content Thief Elixir's Avatar
    Join Date
    Sep 2012
    Location
    Mountains
    Posts
    16,529
    Archived
    707
    Total Posts
    17,236
    Rep Power
    61
    Quote Originally Posted by Worldtraveller View Post
    Yeah, there have been a number of fascinating, and mind boggling, results of using genetic algorithms to solve problems. Real world physical solutions that, like your example above, we don't really quite understand and would have not likely designed on our own. One of my favorite examples is a tiny, lightweight antenna that wound up being use on at least one satellite/spacecraft (I forget the details).

    I am also fascinated by the flip side of this field. Similar in practice an effects, simulating evolution by code or computer models that are then let to run. Some very unexpected results have occurred. One of my favorite (now very old by internet standards as well) examples is from the Secular Web library: https://infidels.org/library/modern/...oretierra.html
    That article gets to what I was thinking (sort of)...

    A sophisticated computer program that is way too complex for humans to understand can now be evolved from random numbers! This completely destroys the Creationist argument that complexity cannot arise without a designer.
    I think the creationist would point to the programmer and say "See! The Designer - one that vastly exceeds the complexity of the program it writes - is required!"
    Of course, accepting that premise leads to the conclusion that their "God" could be the equivalent of Trump's "fat guy in his Mom's basement", who has neither any intent to design anything in particular, nor any knowledge of what he is doing.
    IOW, not a god worth worshiping; not even one that gives a damn about drowning almost all its reproductive iterations in a "fludde".

  6. Top | #6
    Contributor
    Join Date
    Nov 2017
    Location
    seattle
    Posts
    6,613
    Rep Power
    20
    You can write self modifying code. In systems there are adaptive algorithms but they are bounded. The problem is always unpredictable unbounded results. A potential serious problem with autonomous AI based weapons without human oversight. I believe autonomous battle robots are being looked at.


    There are programs that are well beyond the ability of any one person to grasp the whole. tential serious problem with autonomous

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •