An anti-Kantian moment from Turing

Yeah, I know I’ve been rubbish at posting. Well I’m just going through a massive thesis writing week – I’ve somehow managed to write about 13,000 words in 4 days. Part of this is down to a new regime of writing location, I’ve moved from my desktop computer to… a laptop at the top of stairs. I don’t know how these things work, but it’s working.

Anyway here’s Turing again in his famous 1950 article:

“We like to believe that Man is in some subtle way superior to the rest of creation. It is best if he can be shown to be necessarily superior, for then there is no danger of him losing his commanding position. The popularity of the theological argument is clearly connected with this feeling. It is likely to be quite strong in intellectual people, since they value the power of thinking more highly than others, and are more inclined to base their belief in the superiority of Man on this power.” (Copeland 2004: 450)

You can probably tell where I’m heading with this. Anyway more soon.

2 thoughts on “An anti-Kantian moment from Turing”

  1. I really don’t think this is anti-Kant. In fact, I think Kant and Turing have scary levels of affinity with eachother. This is something I’ve been intending to write a paper about for a while. The crucial thing to remember is that Kant is interested principally in ‘rational agency’, not ‘humanity’. The only thing that indexes the transcendental psychology to humanity is our particular forms of sensibility, which Kant does admit could be otherwise, but the account of understanding and reason (the faculties dealing with the intelligible rather than the sensible) are universal (and I would argue, essentially computational). This is a very brief summary, and the sensible/intelligible split is much more complicated than this suggests. Despite indexing the forms of sensibility to humanity Kant still says some very interesting (and potentially universalisable) things about how any possible form of sensation would relate to the inferential capacities that constitute understanding and reason. Leaving this aside though, the moral is that Kant’s problem is nothing other than the hard problem of AI: how would a causal system have to be configured in order to compose a functional rational agent? This is transcendental psychology, and it really should be thought of as a species of computer science, a species which predates the genus by nearly 200 years.

  2. Hey Pete,

    Sorry, but I’d disagree with most of that. If there is any affinity with Turing and Kant, I’d argue that it exists only on the opposite level that you’re talking about, namely the fundamental unknowability of what an effective procedure will actually do. Unfortunately this reply is off the stinging end of a month writing about this very issue. So three things to bare in mind as a shortish summary;

    1.) Universality in a Universal Turing Machine is not about universalising functions as such (although it’s often been interpreted in this way as some sort of convergent generalised machine, (a.k.a John von Neumann), but simply means the discovery of a machine which is maximally sophisticated in the ‘platform’ hardware sense. To this end, there are an infinite number of UTM’s existing in programming languages as well as Turing machines and cellular automata. ‘Universal’ in this instance, requires a proof of equivalence that two machines can execute the same effective procedure, but this only consists with appropriate input in the program in the first place, it’s not a totalising rational thing. It doesn’t really have anything to do with universalising reason, it’s still subject to the same level of undecidability as with all procedures, and not something akin to universal values of rational agency. There is no higher level of computability which Turing sufficiently demonstrated, and it isn’t universal.

    2.) You say that Kant’s problem is nothing other than the hard problem of AI. No problems there as I understand Kant. The trouble is, I don’t think Turing was particularly preoccupied with AI in the first place. This is a controversial position admittedly considering the wealth of literature, but I do uphold that the Turing test you refer to, has little to do with understanding machines as thinking rational thoughts and mimicking rational agents. For a start, Turing called it an imitation game but he never actually specified what the criteria for imitating intelligence was (he actually thought the question ‘can machines think’ as “too meaningless to deserve attention”). His 1950 article came off the back of a report done for the NPL in 1948-49 whereby Turing speculated that intelligence in the computational sense was failing to decide on the outcome of a machine’s behaviour. This is why the imitation game was devised – it’s thoroughly consistent with his undecidable proof of 1936 – putting the interrogator on the level of the observer (all participants were of universal sophistication), who would be unable to decide on either outcome A or B. To be fair Turing never states this explicitly, but I’d argue pretty hard that it was there in the 1950 article.

    3.) Furthermore I’d argue that part of your understanding of Turing comes from a common functionalist and formalist import (usually from Hilbert, but also from Laplace), which undermines what the continuity of undecidability constituted for Turing. Part of this comes from privileging the mental act of rational mental computation in the first place, which is strictly Hilbertian, that the remit of rational computability is indebted to reason despite who or what is executing it. But this preys fowl to the Hilbertian formalist distinction between ideal mind and physical matter (which has to be accounted for and not presupposed), the usual interpretation of which demands that one must generalise the discrete state machine model, having started from the so-called ‘ideal’ level of mathematics and logic.

    The problem is that Turing’s 1936 paper transformed this Hilbertian paradigm by flipping discrete state machines and mathematics and logic the other way around, so that discrete machines became more important than Hilbert’s ideal rational reducible dream. The written symbols of the discrete state machine generated the mind and not the other way round. Same goes for the so-called ideal status of rational mathematics and logic, as demonstrated by Chaitin’s constant, then Wolfram’s research into the mass enumeration of computable axiom systems. Turing is pretty much the opposite of what you find interesting in Kant – rather Hilbert’s your man, considering he was the one who heralded the step-by-step construction of formal meaningless symbols of strings as the basic modus operandi of Mind and thinking itself.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>