Thursday, 4 April 2013

Design detection is a variant of the Turing test

It can be shown that purposive design detection in configurations of matter is a variation of the Turing test.
Fig.1. The Turing Test (courtesy Wikipedia). Receiver С is to detect the actual sender, a computer A or a human В, given that С has no knowledge of which of the other parties generates particular messages.

The Turing test is a thought experiment of information interchange, whereby the receiver is a human, the sender is either a computer program or a human. The receiver must be able to tell the computer from the human given only their messages, without any knowledge of which of the other parties actually sends a particular message. On the other hand, the computer program tries not to give itself away. This test was proposed by Alan Turing as a workaround which helps avoid the onorous task of modelling the process of human reasoning. The idea is simple: if we can write a program which behaves like a human in a certain sense, then we will be able to abstract away from human reasoning, given that all that matters for the receiver is actual messages, not the process of their generation.

John von Neumann demonstrated that biosystems can be modelled by cellular automata capable of sending information to and receiving information from other automata, replicating and maintaining their states (see here). To model biological processes, the rules prescribing how an automaton should change its state depending on received data are loaded into the system. The system is then set to its given initial conditions. What is important is cellular automata in certain conditions are Turing-equivalent. This means that they can be a universal model of computation similar to the Turing machine. So in this case biosystems are equivalent to computers.

Design detection poses the question, is it possible and, if yes, when is it so, to detect a posteriori traces of purposive intelligent agency in an observed arbitrary configuration of matter given this configuration only. In particular, design detection can be applied to biosystems.

Biosystems can adapt to evironments by computing the best (or near-best) states given code mutations, drift and recombinations (in cases of sexual reproduction) acted upon by natual selection. However, how statistically plausible this mechanism is, in terms of generating a particular bio-function, is an open question. In other words, ID maintains that based on statistical analysis of configuration patterns in some cases it is possible to detect traces of intelligent actions just by looking at properties of these configuration, similar to the Turing test:
  • The Turing Test:
    • Sender: a computer program or human;
    • Receiver: a human;
    • A positive (for the receiver) outcome: based on an analysis of responses it is possible to tell the human from the computer.
  • Design Detection:
    • Sender: a biosystem {random variation + natural selection} or an intelligent actor generating new functionality;
    • Receiver: the researcher;
    • Positive outcome: based on statistical analysis of possible configurations and on the properties of a given configuration, the researcher is able to detect (with a given probability) that the sender was an intelligent actor. The necessary condition for the positive outcome, for a given probability of the assertion being true, is that the amount of functional information, associated with the observed configuration (see here) exceeds an independently estimated threshold value. This threshold value corresponds to spontaneous/law-like factors acting alone and is computed based on an analysis of the probabilistic resources of the system. The justification for the introduction of an extra candidate source of biological novelty, an intelligent actor, is the statistical implausibility of non-intelligent factors contributing alone to the generation of large enough quantities of functional information (see e.g. here).

No comments:

Post a Comment