Dennett World

There are machines which take large amounts of data (e.g. weather observations) and produce human-like summaries, such as the literal string [There will be a south-westerly breeze of about 15 knots in Fair Isle tomorrow].

Let's imagine such a machine. It is able to take inputs from various sources (e.g. cameras, microphones, the internet, etc.) and produce text strings. I'll restrict the output to the printable ASCII characters plus carriage return.

After a period of collecting data and running some internal processes, the machine produces the following string:

(A) [Bedford Road is a street which runs downhill from the top of Powis Terrace to a junction on St. Machar Drive very near the Zoology building.]

(I will put strings in square brackets, to distinguish them from sentences.)

The statement which an English speaker might presume the machine was making when it produced this string is (more or less) true.

There is clearly a causal chain of some kind from the concrete fact of Bedford Road and the output of the ASCII string.  It involves the material of the street and it's environs, the 'sensory' input from cameras etc., and the physical states of any other machines and objects with which our imagined machine has interacted.

(Remember that the production of an ASCII string can be rendered in entirely physical terms - e.g. states of bi-stable circuits, or of magnetic domains, or patterns of ink on paper etc.)

This machine is a creature in a Dennettian world. I will even allow it some 'emergence', in the sense that it may be far too complex a machine for any human being, or any group of human beings, to be able to follow its processes from input to output.

We might imagine that such a machine could be developed to the extent that it could answer any purely descriptive question about the world outside itself, such as 'Is it raining in Lima?'

There would be things it would never be able to do - such as produce comprehensive descriptions of all of its own processes of description. Turingesque limitations would prevent this. In particular, it could not produce the string:

(B) [The machine which produced this string can accurately describe the world]

as a result of its normal descriptive processes, such as the ones which produced the Bedford Road string.

This is because it would need to incorporate a reliable test of its own empirical accuracy in order to be able to do this. This would generate an open question paradox for the machine: either it is presently able to accurately describe the world, or it is not. However it was enhanced, the same question would arise.

Either the machine itself is the standard of empirical accuracy, or it has no general standard of empirical accuracy.

This means that there are 'facts' about the world which an English speaker would never see 'represented' in the strings this machine produced - no matter how long they waited, or how cunningly they questioned it. String (B), after all, represents such a fact.

This has a consequence for Dennett's picture of 'reality', which represents people as machines of the type I have described - only with a potentially wider range of inputs and outputs.

For Dennett's picture to work, all of the inputs and outputs would have to be potentially representable in text strings (since this is, in this picture, how we produce scientific and philosophical theories).