Wednesday, July 13, 2005

Menuge replies to Carr II

Carr writes: Menuge wrote 'What's puzzling about the passage below is how we can specifiy what the information that is being transmitted is without talking about its meaning.'But a CSI detecting machine can, according to Menuge,pattern match the specification of the information without knowing about the meaning of the specification.Is *specified* information something that cannot be detected by naturalistic methods, or can only human beings detect specified information (as only agents can understand the meaning of the specification)?

Menuge replies:

Well, here's an example: a mindless computer program might be programmed withcertain information that in fact contains CSI, such as, to use Dembski's example, a long sequence of prime numbers. Now suppose that in the Search forExtra Terrestrial Intelligence (SETI) project, a radio signal is detected from deep space that consists of a large sequence of these prime numbers. Theprogram could match the received sequence with a subsection of the sequence ofprimes in its database, and also determine if the sequence was long enough to be"complex" (in the sense that it is too unlikely to have been generated by chance). It seems to me, so long as "detect" is used in a minimalistic way (athermometer detects the temperature in this sense), there is no reason to denythat the program detects CSI.  The machine can detect (reliably register /signal) this CSI even though there is no reason to suppose it knows what a primenumber is, or why the sequence it received consitutes CSI. The way I see it, detection does not require understanding, and the detection of CSI that has a meaning does not require that the detecting device can access the meaning. This is rather like the way a voice-activated computer can be programmed to obey thespoken words "Open Word" without knowing what it means to open the program. So it seems that, at least in some cases (I'm not sure in all, because some information is not so easy to represent computationally: there might be problems raised by Turing's halting problem or Godel's incompleteness theorem, where a human can detect CSI that is not computationally tractable---but this is endlessly debated), CSI can be detected by naturalistic methods. However, just as there is distinction between non-conceptual seeing (seeing red) and conceptual seeing (seeing red as danger), it seems to me that the program thatdetects CSI is not, like the scientist, seeing it as CSI, because it lacks the relevant concepts.. Angus

4 comments:

Steven Carr said...

You mean the program which detects CSI in a sequence of prime numbers has to be programmed beforehand with the specification that it is matching, and that only humans can recognise the specifications (although once having done so, they can tell a machine what pattern to look for)



How then can CSI be science, if it boils down to 'I know it when I see it'.?


There is no scientific way of detecting CSI , as detecting specification is a subjective process.

Victor Reppert said...

Steven: I got this reply from Menuge:

In response to Carr,
No,
I made it perfectly clear that detecting CSI is perfectly objective, and normally a machine cannot (I think he means can-VR) do it. Detectign CSI just does not require understanding what it means.. This is just the same in other areas of science. Siesmographic instruments detect earthquakes without understanding what they are. No-one argues that seismography fails to be obejctive and hence is not scientific on these grounds.

Steven Carr said...

Dretske wrote ' Once this distinction is clearly understood, one is free to think about nformation (though not meaning) as an objective commodity, something whose generation, transmission, and reception do not require or in any way presuppose interpretive processes."

Was Dretske right to say that information is an objective commodity - one that can be detected by something that cannot comprehend the meaning of the information, and presumably generated by the same machine running in reverse? (If a machine can say that a particular pattern contains information, then a machine can work out what patterns contain information and then generate them)


Can we take a machine , give it a 'something' and tell it to tell us whether that something has CSI , without having to tell the machine exactly what particular specification (or specifications) it is matching against?

Is there a general algorithm which will detect a specification, for all specifications (or even just for many specifications)?

If CSI can be detected by a machine, then can information be detected by something that cannot understand the meaning of the information?

Steven Carr said...

It seems to me like you could use Menuge's arguments to say that art is scientfically verifiable,but that only humans can recognise art.

Menuge can take individual works of art (such as he takes individual specifications , such as his prime number example), then program a machine to recognise these works of art when it seems them.

'Hence' art can be objectively recognised.

Of course, this does not follow ,because you have programmed the machine with the resulty you want - If you see something that resembles the Mona Lisa , print out 'art'.

Similarly with his claim that CSI is scientifically objective.

He has just programmed the machine with the result he wants. He claimed himself that he hard coded the specification the machine is matching against.

If the machine see prime numbers , print out 'CSI'.

If the machine sees DNA, print out 'CSI'. etc, etc.

CSI would only be scientifically objective if Menuge did not have to program the computer with each and every specification it has to match against before it prints out 'CSI'.

Just as 'art' would only be scientifically objective if Menuge did not have to program the computer with each and every work of art it has to match against before it prints out 'art'.