I know, I know, but I still want to share the link to Josh Parsons’ home page. I found it by the normal method of leaping randomly from tussock to tussock through the Internet’s hyperspace and I’ve had so much fun this evening playing with the Vorld Vide Wittgenstein – the Tractatus Generator. This delightful program will generate sections of what should have been in the Tractatus. Here’s a sample gem:
From the Tractatus Logico-Randomus, section 12:
12 Just as the object is total, so a possibility of a picture can lie in the nature of the internal fact. (No such possibility exists independently of a structure).
12.1 We cannot think that every question must be a substance, for otherwise we would think unlogically.
12.2 I conceive of an analysis as a sign.
More difficult, perhaps, than writing a program to generate counter-factual Tractati is a program to read them and distinguish the genuine. There are several levels, all syntactic rather than semantic, at which such a program could work:
- Giving the program the whole genuine Tractatus and having it scan for the candidate text—too easy and no fun
- A trivial word-counting such as is used to prove that the Earl of Rutland wrote Shakespeare’s plays when we all know that it was Marlowe
- A supervised learning program (for example, a neural network)
- A clustering program based on some normed distance between the texts
- An unsupervised learning program (for example, Re-inforcement Learning)
Before putting pencil to coding sheet (a reference that no one under the age of 50 will presumably understand), I was wondering which level was closest to a human’s activity—and in particular could a non-philosopher distinguish semantically? I sat back and tried the above section 12 out on a random passer-by (my wife actually). Would she be able to distinguish it from the genuine article? The answer was ambiguous. Perhaps this should be a practical experiment rather than a programming exercise: print out genuine snippets and paragraphs from the Tractatus Logico-Randomus and ask non-philosophers to distinguish the two. What about this, for example?
3.261 Every sign that has a definition signifies via the signs that serve to define it; and the definitions point the way. Two signs cannot signify in the same manner if one is primitive and the other is defined by means of primitive signs. Names cannot be anatomized by means of definitions. (Nor can any sign that has a meaning independently and on its own.)
3.262 What signs fail to express, their application shows. What signs slur over, their application says clearly.
3.263 The meanings of primitive signs can be explained by means of elucidations. Elucidations are propositions that stood if the meanings of those signs are already known.
I fear that this could be a really depressing exercise.