NLF theorems are stated in terms of the average performance of evolutionary algorithms, but ID proponents must be mindful whenever the word AVERAGE is used, because it implies there may be above average performers, and I’m surprised Darwinists have been slow to seize refuge in the possibility of above average outcomes.
To illustrate, the house edge (casino edge) in the game of dice (craps) is a mere 1.41% for the “passline” wager. So on average we expect the casino to win, but not immutably. I asked one pit boss, “what was the longest winning streak by the players?” He said something on the order of 15 wins in a row, and the casino lost over $140,000 in a few hours as a result. 😯 Just because the casino on average has the edge over dice players, doesn’t mean there will not be above average performers. There will be, it is guaranteed!
All this to illustrate, that even with the NFL theorems saying evolution will not increase CSI on average, it does not preclude the possibility of pathological examples where CSI increases above average, in fact it is guaranteed to happen whenever we use the notion of average (with some deviation from average).
What is the best pathological example I can think of. How about at thought example first. Consider a robot whose sole mission is to take existing materials and build Rube Gold machines.
In principle, this robot can build something far more complex than itself. That is to say, the final CSI doesn’t have to be as intelligent as the robot, but it could have more interlocking parts than the robot. The only task of such Rube Goldberg machines might be to turn on a light. But if the robot builds enough of them with sufficient variety, then there will be more CSI in the end than what we started with. At the very least, the CSI of the Robot combined with the CSI of the Rube Goldberg machines it builds is greater CSI than the initial CSI the Robot by itself before it started on its task.
We see small illustrations with this if we start out with a small population of beavers. Let them loose, let them multiply, and CSI will increase as they build more and more Dams. Or how about Bees with honeycombs? But what matters is not the quantity of CSI, it is the quality. Building more honeycombs (and hence more CSI) is not the sort of CSI that the ID/Evolution debate is really interested in. Unfortunately the NFL theorems do not distinguish quality of CSI from quantity.
For that matter, take a robot in a room full of coins that have random heads-tail configurations. The robot orders them all heads. The final CSI inside the room (the Robot’s CSI plus the coin’s CSI) is now greater than what we began with!
Because biology has agents that can be deemed to possess weak AI (Aritifical Intelligence), they can in principle increase CSI from their initial state if they are pre-programmed to do so. Front-loaded evolution, or the evolution James Shapiro envisions may take advantage of this fact. Do I believe this is what happened? No. IMHO, the evidence indicates we do not live in world that is occupied by such pathological examples. But I highlight these issues for the sake of completeness. Care should be taken when arguing ID using NFL theorems. I illustrate how I argued it in The mootness of anti-NFL arguments.
Perhaps the moral of this essay, NFL theorems do not distinguish the quality of CSI from the quantity. In the case of the robot increasing CSI in a room of coins, it doesn’t improve the CSI of the robot. Be careful using NFL arguments to defend ID.