View Single Post
Old 05-23-2019, 08:19 AM
Half Man Half Wit's Avatar
Half Man Half Wit is online now
Join Date: Jun 2007
Posts: 6,897
Originally Posted by wolfpup View Post
In an effort to keep it brief, I'm omitting those points where I would simply be repeating myself.
You should have instead kept it longer in an effort to reply to the points you keep omitting...

This can ONLY be interpreted as "no cognitive processes at all can be computational", since ANY such computation would, according to your claim, require an external interpretive agent. If true, that would invalidate CTM in its entirety.
The computational theory of mind is the statement that computation is all the brain does, and, in particular, that consciousness is computational. This, I indeed have shown to be in error.

That does in no way imply that no process that goes on in the brain is computational. I've been careful (to no avail, it seems) to point out that my argument threatens solely the interpretational abilities of minds: they can't be computational. Using these interpretational powers, it becomes possible to assign definite computations to systems---after all, I use the device from my example to compute, say, sums.

Furthermore, even systems that aren't computational themselves may be amenable to computational modeling---just as well as systems that aren't made of springs and gears may be modeled by systems that are, like an orrery, but I suspect where these words are, you just see a blank space.

Only when challenged on it are you now offering creative re-interpretations. But perhaps you'd like to take on the creative challenge of re-interpreting what you meant by CTM having been "dismantled".
I hold consistently to the same position I did from the beginning: computational modeling of the brain is useful and tells us much about it, but the mind is not itself computational. I have been very clear about this. Take my very first post in this thread:

Originally Posted by Half Man Half Wit View Post
But if that's so, then computation can't be what underlies consciousness: if there's no fact of the matter regarding what mind a given system computes unless it is interpreted as implementing the right computation, then whatever does that interpreting can't itself be computational, as otherwise, we would have a vicious regress---needing ever higher-level interpretational agencies to fix the computation at the lower level. But if minds then have the capacity to interpret things (as they seem to), they have a capacity that can't be realized via computation, and thus are, on the whole, not computational entities.
There, I clearly state that whatever realizes the mind's interpretational capacity can't be computational, and thus, minds can't be computational on the whole. That doesn't entail that nothing about minds can be computational. That would be silly: I have just now computed that 1 + 4 equals 5, for instance.

Also, I have been clear that my arguments don't invalidate the utility of computational modeling:

Originally Posted by Half Man Half Wit View Post
Also, none of this threatens the possibility or utility of computational modeling. This is again just confusing the map for the territory. That you can use an orrery to model the solar system doesn't in any way either imply or require that the solar system is made of wires and gears, and likewise, that you can model (aspects of) the brain computationally doesn't imply that the brain is a computer.
Originally Posted by wolfpup View Post
"Wholly computational" was manifestly never my claim, and I was clear on that from the beginning.
Then why take issue with my claim of demonstrating a non-computational ability of the mind?

A Turing machine starts with a tape containing 0110011. When it's done the tape contains 0100010. What computation did it just perform?
As such, the question is underdetermined: there are infinitely many computations that take 0110011 to 0100010. This isn't a computation, it's rather an execution trace.

But of course, I know what you mean to argue. So let's specify a computation in full: say, the Turing machine has an input set consisting of all seven bit strings, and, to provide an output, traverses them right to left, replacing each block '11' it encounters with '10'. Thus, if produces '0100010' from '0110011', or '1000000' from '1111111', or '0001000' from '0001100'.

This is indeed a fully formally specified, completely determinate computation. You'll note it's of exactly the same kind of thing as my functions f and f'. So why does a Turing machine execute a definite computation?

Simple: a Turing machine is a formally specified, abstract object; its vehicles are themselves abstract objects, like '1' and '0' (the binary digits themselves, rather than the numerals).

But that's no longer true for a physical system. A physical system doesn't manipulate '1' and '0', it manipulates physical properties (say, voltage levels) that we take to stand for or represent '1' or '0'. It's here that the ambiguity comes in.

If you were to build the Turing machine from your example, then all it could do would be to write 1 or 0 (now, the numerals, not the binary digits) onto its tape. Anybody familiar with the Arabic numbers could then grasp that these ink-blots-on-paper are supposed to mean '1' and '0' (the binary digits, again). But, somebody grown up on Twin Earth that's identical to ours except for the fact that 1 means '0' and 0 means '1' would, with equal claim to being correct, take the Turing machine to implement a wholly different computation; namely, one where every '00' string is replaced by a '01'-string.

That's why I keep asking (and also, why you keep neglecting to answer): what computation is implemented by my example device? You're backed into a corner where you'd either have to answer that it's a computation taking switch-states to lamp-states, in which case the notion of computation collapses to that of physical evolution, or agree with me that it can be equally well taken to implement f and f'.

Although I note that you seem to have shifted your stance here somewhat---or perhaps, it hasn't been entirely clear from the beginning: you've both argued that the two computations are the same (which amounts to accepting they're both valid descriptions of the system, just equivalent ones, which starkly conflicts with you singling out a function of the same class as individuated computation in this post), and that multiple interpretations become, for reasons vaguely tied to 'emergence', less likely with increasing complexity. So which is it, now?

Perhaps for one last time, let me try and make my main point clear in a different way. Symbols don't intimate their meanings on their own. Just being given a symbol, or even a set of symbols with their relations, doesn't suffice to figure out what they mean. This is what the Chinese Room actually establishes (it fails to establish that the mind isn't computational): given a set of symbols (in Chinese), and rules for manipulating these, it's in principle possible to hold a competent conversation; but it's not possible to get at what the symbols mean, in any way, shape, or form.

Why is that the case? Because there's not just one thing they could mean. That must be the case, otherwise, we could just search through meanings until we find the right one. But it just isn't the case that symbols and rules to manipulate them, and relationships they stand in, determine what the symbols mean.

But it's in what their physical properties mean that physical systems connect to abstract computation. Nothing else can be right; computations aren't physical objects, and the only relation between the physical and the abstract is one of reference. So just the way you can't learn Chinese from manipulating Chinese letters according to rules, you can't fix what computation a system performs by merely having it manipulate physical properties, or objects, according to rules. An interpretation of what those properties or objects mean, what abstracta they refer to, is indispensable.

But this reference will always introduce ambiguity. And hence, there is no unambiguous, objective computation associated with a physical system absent it being interpreted as implementing said computation.

Last edited by Half Man Half Wit; 05-23-2019 at 08:20 AM.