= 27d567bc2d48d9f40e9ccc20ea370b15)
ABrooks comments on Rationality Quotes April 2012 - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (858)
What is the difference between a thought you can't think and one you don't think?
Well, for example I don't think very much about soccer. There are thoughts about who the best soccer team is that I simply don't ever think. But I can think them.
Another case: In two different senses of 'can', I can and can't understand Spanish. I can't understand it at the moment, but nevertheless Spanish sentences are in principle translatable into sentences I can understand. I also can't read Aztec hieroglyphs, and here the problem is more serious: no one knows how to read them. But nevertheless, insofar as we assume they are a form of language, we assume that we could translate them given the proper resources. To see something as translatable just is to see it as a language, and to see something as a language is to see it as translatable. Anything which was is in principle untranslatable just isn't recognizable as a language.
I think the point is analogous (and that's no accident) with thoughts. Any thought that I couldn't think by any means is something I cannot by any means recognize as a thought in the first place. All this is just a way of saying that the belief that there are thoughts you cannot think is one of those beliefs that could never modify your anticipations. That should be enough to discount it as a serious consideration.
And yet, if I see two nonhuman life forms A1 and A2, both of which are performing something I classify as the same task but doing it differently, and A1 and A2 interact, after which they perform the task the same way, I would likely infer that thoughts had been exchanged between them, but I wouldn't be confident that the thoughts which had been exchanged were thoughts that could be translated to a form that I could understand.
Alternative explanations include:
These are, of course, not exhaustive.
You could call some these cases a kind of thought. Maybe to self-modifying programs, a blackbox executable algorithm counts as a thought; or maybe to beings who use the same information storage for genes and minds, lateral gene transfer counts as a thought.
But this is really just a matter of defining what the word "thought" may refer to. I can define it to include executable undocumented Turing Machines, which I don't think humans like us can "think". Or you could define it as something that, after careful argument, reduces to "whatever humans can think and no more".
Sure. Leaving aside what we properly attach the label "thought" to, the thing I'm talking about in this context is roughly speaking the executed computations that motivate behavior. In that sense I would accept many of these options as examples of the thing I was talking about, although option 2 in particular is primarily something else and thus somewhat misleading to talk about that way.
I think you're accepting and then withdrawing a premise here: you've identified them as interacting, and you've identified their interaction as being about the task at hand, and the ways of doing it, and the relative advantages of these ways. You've already done a lot of translation right there. So the set up of your problem assumes not only that you can translate their language, but that you in some part already have. All that's left, translation wise, is a question of precision.
Sure, to some level of precision, I agree that I can think any thought that any other cognitive system, however alien, can think. There might be a mind so alien that the closest analogue to its thought process while contemplating some event that I can fathom is "Look at that, it's really interesting in some way," but I'll accept that this in some part a translation and "all that's left" is a question of precision.
But if you mean to suggest by that that what's left is somehow negligible, I strenuously disagree. Precision matters. If my dog and I are both contemplating a ball, and I am calculating the ratio between its volume and surface, and my dog is wondering whether I'll throw it, we are on some level thinking the same thought ("Oh, look, a ball, it's interesting in some way") but to say that my dog therefore can understand what I'm thinking is so misleading as to be simply false.
I consider it possible for cognitive systems to exist that have the same relationship to my mind in some event that my mind has to my dog's mind in that example.
Well, I don't think I even implied that the dog could understand what you're thinking. I don't think dogs can think at all. What I'm claiming is that for anything that can think (and thus entertain the idea of thoughts that cannot be thought), there are no thoughts that cannot be thought. The difference between you and your dog isn't just one of raw processing power. It's easy to imagine a vastly more powerful processor than a human brain that is nevertheless incapable of thought (I think Yud.'s suggestion for an FAI is such a being, given that he's explicit that it would not rise to the level of being a mechanical person).
Once we agree that it's a point about precision, I would just say that this ground can always in principle be covered. Suppose the translation has gotten started, such that there is some set of thoughts at some level of precision that is translatable, call it A, and the terra incognito that remains, call it B. Given that the cognitive system you're trying to translate can itself translate between A and B (the aliens understand themselves perfectly), there should be nothing barring you from doing so as well.
You might need extremely complex formulations of the material in A to capture anything in B, but this is allowed: we need some complex sentence to capture what the Germans mean by 'schadenfreude', but it would be wrong to think that because we don't have a single term which corresponds exactly, that we cannot translate or understand the term to just the same precision the Germans do.
I accept that you don't consider dogs to have cognitive systems capable of having thoughts. I disagree. I suspect we don't disagree on the cognitive capabilities of dogs, but rather on what the label "thought" properly refers to.
Perhaps we would do better to avoid the word "thought" altogether in this discussion in order to sidestep that communications failure. That said, I'm not exactly sure how to do that without getting really clunky, really fast. I'll give it a shot, though.
I certainly agree with you that if cognitive system B (for example, the mind of a Geman speaker) has a simple lexical item Lb (for example, the word "schadenfreude") ,
...and Lb is related to some cognitive state Slb (for example, the thought /schadenfreude/) such that Slb = M(Lb) (which we ordinarily colloquially express by saying that a word means some specific thought),
...and cognitive system A (for example, the mind of an English speaker) lacks a simple lexical item La such that Slb=M(La) (for example, the state we'd ordinarily express by saying that English doesn't have a word for "schadenfreude")...
that we CANNOT conclude from this that A can't enter Slb, nor that there exists no Sla such that A can enter Sla and the difference between Sla and Slb is < N, where N is the threshold below which we'd be comfortable saying that Sla and Slb are "the same thought" despite incidental differences which may exist.
So far, so good, I think. This is essentially the same claim you made above about the fact that there is no English word analogous to "schadenfreude" not preventing an English speaker from thinking the thought /schadenfreude/.
In those terms, I assert that there can exist a state Sa such that A can enter Sa but B cannot enter Sa. Further, I assert that there can exist a state Sa such that A can enter Sa but B cannot enter any state Sb such that the difference between Sa and Sb is < N.
Do you disagree with that? Or do you simply assert that if so, Sa and Sb aren't thoughts? Or something else?
I agree that this is an issue of what 'thoughts' are, though I'm not sure it's productive to side step the term, since if there's an interesting point to be found in the OP, it's one which involves claims about what a thought is.
I'd like to disagree with that unqualifiedly, but I don't think I have the grounds to do so, so my disagreement is a qualified one. I would say that there is no state Sa such that A can enter Sa, and such that B cannot enter Sa, and such that B can recognise Sa as a cognitive state. So without the last 'and such that', this would be a metaphysical claim that all cognitive systems are capable of entertaining all thoughts, barring uninteresting accidental interference (such as a lack of memory capacity, a lack of sufficient lifespan, etc.). I think this is true, but alas.
With the qualification that 'B would not be able to recognise Sa as a cognitive state', this is a more modest epistemic claim, one which amounts to the claim that recognising something as a cognitive state is nothing other than entering that state to one degree of precision or another. This effectively marks out my opinion on your second assertion: for any Sa and any Sb, such that the difference between Sa and Sb cannot be < N, A (and/or B) cannot by any means recognise the difference as part of that cognitive state.
All this is a way of saying that you could never have reason to think that there are thoughts that you cannot think. Nothing could give you evidence for this, so it's effectively a metaphysical speculation. Not only is evidence for such thoughts impossible, but evidence for the possibility of such thoughts is impossible.
I'm not exactly sure what it means to recognize something as a cognitive state, but I do assert that there can exist a state Sa such that A can enter Sa, and such that B cannot enter Sa, and such that B can believe that A is entering into a particular cognitive state whenever (and only when) A enters Sa. That ought to be equivalent, yes?
This seems to lead me back to your earlier assertion that if there's some shared "thought" at a very abstract level I and an alien mind can be said to share, then the remaining "terra incognito" between that and sharing the "thought" at a detailed level is necessarily something I can traverse.
I just don't see any reason to expect that to be true. I am as bewildered by that claim as if you had said to me that if there's some shared object that I and an alien can both perceive, then I can necessarily share the alien's perceptions. My response to that claim would be "No, not necessarily; if the alien's perceptions depend on sense organs or cognitive structures that i don't possess, for example, then I may not be able to share those perceptions even if I;n perceiving the same object." Similarly, my response to your claim is "No, not necessarily, if the alien's 'thought' depends on cognitive structures that i don't possess, for example, then I may not be able to share that 'thought'."
You suggest that because the aliens can understand one another's thoughts, it follows that I can understand the alien's thoughts, and I don't see how that's true either.
So, I dunno... I'm pretty stumped here. From my perspective you're simply asserting the impossibility, and I cannot see how you arrive at that assertion.
Well, if the terra incogntio has any relationship at all to the thoughts you do understand, such that the terra could be recognized as a part of or related to a cognitive state, then the terra is going to consist in stuff which bears inferential relations to what you do understand. These are relations you can necessarily traverse if the alien can traverse them. Add to that the fact that you've already assumed that the aliens largely share your world, that their beliefs are largely true, and that they are largely rational, and it becomes hard to see how you could justify the assertion at the top of your last post.
And that assertion has, thus far, gone undefended.