7 Replies to “Guided tour of no free lunch theorems

  1. 1
    DiEb says:

    Since when is a bunch of quotations a “guided tour”?

  2. 2
    Daniel King says:

    You have to read between the links.

  3. 3
    Querius says:

    While I don’t understand NFL learning and optimization to any depth, it seems that the primary critique against Dembski’s employment in ID has to do with misapplication regarding the evaluation of optimal solutions across multiple domains—Darwinistic evolution being considered a single domain.

    The objection seems plausible to me. Years ago, I had a conversation with a software engineer involved in machine learning for signature recognition on checks. He told me that the program created and applied rules to optimize success. Furthermore, he claimed that he had no idea what these rules were, but whatever they were, they seemed to work for about 30% of the signatures IIRC.

    What occurs to me is that the “learning to learn” is an algorithm in of itself.

    So, perhaps one can eventually find a scenario where a non-degrading* beneficial mutation occurs, perhaps even several that together result in a de novo gene or structure, but the learning-to-learn operation of DNA and epigenetic information also requires explanation.

    Considering the obvious and overwhelming complexity and fragility of DNA, this is simply asking way too much of Darwinian processes. Has anyone ever observed or even tried to observe the evolution of the mechanics of DNA? How about damaging one function by simplifying it and then observing how long it takes to evolve back?

    * Specifically mentioned to exclude antibiotic resistance where function is lost.

    -Q

    p.s. Glad to see you here Daniel King!

  4. 4
    Mung says:

    Since when is a bunch of quotations a “guided tour”?

    That was just to whet your appetite. You have to pay for the real guided tour, it’s no free lunch.

  5. 5
    Eric Anderson says:

    DiEB:

    I guess that should have been “self-guided” tour. 🙂

  6. 6
    the bystander says:

    Querius @ 3

    What occurs to me is that the “learning to learn” is an algorithm in of itself.

    For that algorithm to evolve, you need a ‘learning set’. So for DNA to gain intelligence to evolve its own instructions, it would have to ‘learn’ from a set of other DNAs and those other DNAs would either have ‘evolved’ on their own or would have been designed by someone/ something /Nature(?)

  7. 7
    Querius says:

    the bystander,

    Yes. The problem is where did the other DNAs get their “learning set”? I know . . . from still other DNAs. 😉

    I find the learn–how–to–learn issue a difficult obstacle. For example . . .

    Here’s a demo of a learning system.

    1. Take an empty chessboard and put a single black pawn on c3 (QB3). This is the guard dog. The guard dog moves clockwise in a square path, maintaining its adjacency to the four center squares.

    2. Take a white pawn and put it on one of the four center squares—the one furthest from the dog. This is the fox. Move the fox one square based on a card draw from a deck of cards. Hearts=north, Spades=east, Diamonds=south, and Clubs= west. The dog and the fox alternate moves.

    3. But to do this, draw five cards. If the fox survives, keep the five cards in order. If the fox or the dog ever occupy the same square, the fox is caught by the dog. Only those five cards are discarded and reshuffled. You start over, but use the previously successful sets of five cards.

    4. Eventually, you will be able to move the fox through the entire deck without being caught.

    Yay, this is a learning system. But how do you evolve the rule to use sets of 5 cards? Why not 3 cards, 1 card, or 10 cards?

    From this exercise, you can probably see that if the number of cards in a set vary occasionally, you might be able to “evolve” the optimum number of cards in a set for the fastest ordering of the cards in the deck.

    However, you can probably also see how this “deferent” will take way longer to evolve than the “epicycles”!

    -Q

Leave a Reply