= 27d567bc2d48d9f40e9ccc20ea370b15)
DeVliegendeHollander comments on Open Thread, May 11 - May 17, 2015 - Less Wrong Discussion
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 (247)
In what conceiveable (which does not imply logicality) universes would Rationalism not work in the sense of unearthing only some truths, not all truths? Some realms of truth would be hidden to Rationalists? To simplify it, I mean largely the aspect that of empiricism, of tying ideas to observations via prediction. What conceivable universes have non-observational truths, for example, Platonic/Kantian "pure apriori deduction" type of mental-only truths? Imagine for convenience's sake a Matrix type simulated universe, not necessarily a natural one, so it does not really need to be lawful nor unfold from basic laws.
Reason for asking: if you head over to a site like The Orthosphere, they will tell you Rationalism can only find some but not all truths. And one good answer would be: "This could happen in universes of the type X, Y, Z. What are your reasons for thinking ours could be one of them?"
Depends on how you decide what truth is, and what qualifies it to be "unearthed."
But for one universe in which some truth, for some value of truth, can be unearthed, for some value of unearthed, while other truth can't be:
Imagine a universe in which 12.879% (exactly) of all matter is a unique kind of matter that shares no qualities in common with any other matter, and is almost entirely nonreactive with all other kinds of matter, and was created by a process not shared in common with any other matter, which had no effect whatsoever on any other matter. Any truths about this matter, including its existence and the percentage of the universe composed of it, would be completely non-observational. The only reaction this matter has with any other matter is when it is in a specific configuration which requires extremely high levels of the local equivalent of negative entropy, at which point it emits a single electromagnetic pulse. This was used once by an intelligence species composed of this unique matter who then went on to die in massive wars, to encode in a series of flashes of light every detail they knew about physics, and was observed by one human-equivalent monk ascetic, who used a language similar to morse code to write down the sequence of pulses, which he described as a holy vision. Centuries later, these pulses were translated into mathematical equations which described the unique physics of this concurrent universe of exotic matter, but no mechanism of proving the existence or nonexistence of this exotic matter, save that the equations are far beyond the mathematics of anyone alive at the time the signal was encoded, and it has become a controversial matter whether or not it was an elaborate hoax by a genius.
What do you mean with "Rationalism"?
The LW standard definition is that it's about systematized winning. If the Matrix overlords punish everybody who tries to do systematized winning than it's bad to engage in it. Especially when the Matrix overlords do it via mind reading. The Christian God might see it as a sin.
If you don't use the LW definition of rationalism, then rationalism and empiricism are not the same thing. Rationalism generally refers to gathering knowledge by reasoning as opposed to gathering it by other ways such as experiments or divine revelation.
Gödel did prove that it's impossible to find all truths. This website is called Lesswrong because it's not about learning all truths but just about becoming less wrong.
That's misleading. With a finite amount of processing power/storage/etc, you can't find all proofs in any infinite system. We need to show that short truths can't be found, which is a bit harder.
I don't think that's correct. My best understanding of Godel's theorem is that if your system of logic is powerful enough to express itself, then you can create a statement like "this sentence is unprovable". That's pretty short and doesn't rely on infiniteness.
The statement "this sentence is unprovable" necessarily includes all information on how to prove things, so it's always larger than your logical system. It's usually much larger, because "this sentence" requires some tricks to encode.
To see this another way, the halting problem can be seen as equivalent to Godel's theorem. But it's trivially possible to have a program of length X+C that solves the halting problem for all programs of length X, where C is a rather low constant; see https://en.wikipedia.org/wiki/Chaitin'sconstant#Relationshiptothehalting_problem for how.
I'm not sure how much space it would take to write down formally, and I'm not sure it matters. At worst it's a few pages, but not entire books, let alone some exponentially huge thing you'd never encounter in reality.
It's also not totally arbitrary axioms that would never be encountered in reality. There are reasons why someone might want to define the rules of logic within logic, and then 99% of the hard work is done.
But regardless, the interesting thing is that such an unprovable sentence exists at all. That its not possible to prove all true statements with any system of logic. It's possible that the problem is limited to this single edge case, but for all I know these unprovable sentences could be everywhere. Or worse, that it is possible to prove them, and therefore possible to prove false statements.
I think the halting problem is related, but I don't see how it's exactly equivalent. In any case the halting problem work around is totally impractical, since it would take multiple ages of the universe to prove the haltingness of a simple loop. If you are referring to the limited memory version, otherwise I'm extremely skeptical.
That's only if your logical system is simple. If you're a human, then the system you're using is probably not a real logical system, and is anyway going to be rather large.
See http://www.solipsistslog.com/halting-consequences-godel/
DeVliegendeHollander post didn't speak about short truths but about all truths.
If we're talking about all truths, then a finiteness argument shows we can never get all truths, no need for Godel. Godel shows that given infinite computing power, we still can't generate all truths, which seems irrelevant to the question.
If we can prove all truths smaller than the size of the universe, that would be pretty good, and it isn't ruled out by Godel.
While Gödel killed HIlbert's program as a matter of historical fact, it was later Tarski who proved the theorem that truth is undecidable.
There's no guarantee we should be able to find any truths using any method. It's a miracle that the universe is at all comprehensible. The question isn't "when can't we learn everything?", it's "why can we learn anything at all?".
Because entities which can't do not survive.
Counterexample: Plants. Do they learn?
Of course. Leaves turn to follow the sun, roots grow in the direction of more moist soil...
Is that really learning, or just reacting to stimuli in a fixed, predetermined pattern?
Does vaccination imply memory?.. Does being warned by another's volatile metabolites that a herbivore is attacking the population?
(Higher) plants are organized by very different principles than animals; it is a never-ending debate on what constitutes 'identity' in them. Without first deciding upon that, can one speak about learning? I don't think they have it, but their patterns of predetermined answers can be very specific.
Also, there is an interesting study, 'Kin recognition, not competitive interactions, predicts root allocation in young Cakile edentula seedling pairs'. This seems to be more difficult to do than following the sun!
That just pushes the question back a step. Why can any entity learn?
In the spirit of Lumifer's comment, anything we would consider an entity would have to be able to learn or we wouldn't be considering it at all.
That would explain why all entities learn. Not why any entities learn. Ignoring things that can't learn doesn't explain the existence if things that can.
A more useful question to ask would be "how do entities, in fact, learn?" This avoids the trite answer, "because if they didn't, we wouldn't be asking the question".
I think if we follows this chain of questions, what we'll find at the end (except for turtles, of course) is the question "Why is the universe stable/regular instead of utterly chaotic?" A similar question is "Why does the universe even have negentropy?"
I don't know any answer to these questions except for "That's what our universe is".
I suppose what I want to know is the answer to "What features of our universe make it possible for entities to learn?".
Which sounds remarkably similar to DeVliegendeHollander's question, perhaps with an implicit assumption that learning won't be present in many (most?) universes.
The fact that the universe is stable/regular enough to be predictable. Subject predictability is a necessary requirement for learning.
For that matter, a world in which it is impossible for an organism to become better at surviving by modeling its environment (i.e. learning) is one in which intelligence can't evolve.
(And a world in which it is impossible for one organism to be better at surviving than another organism, is one in which evolution doesn't happen at all; indeed, life wouldn't happen.)
Don't need to posit crazy things, just think about selection bias -- are the sorts of people that tend to become rationalist randomly sampled from the population? If not, why wouldn't there be blind spots in such people just based on that?
Yes, but if I get the idea right, it is to learn to think in a self-correcting, self-improving way. For example, maybe Kanazawa is right in intelligence suppressing instincts / common sense, but a consistent application of rationality sooner or later would lead to discovering it and forming strategies to correct it.
For this reason, it is more of the rules (of self-correction, self-improvement, self-updating sets of beliefs) than the people. What kinds of truths would be potentially invisible to a self-correcting observationalist ruleset even if this was practiced by all kinds of people?
Just pick any of a large set of things the LW-sphere gets consistently wrong. You can't separate the "ism" from the people (the "ists"), in my opinion. The proof of the effectiveness of the "ism" lies in the "ists".
Which things are you thinking of?
A lot of opinions much of LW inherited uncritically from EY, for example. That isn't to say that EY doesn't have many correct opinions, he certainly does, but a lot of his opinions are also idiosyncratic, weird, and technically incorrect.
As is true for most of us. The recipe here is to be widely read (LW has a poor scholarship problem too). Not moving away from EY's more idiosynchratic opinions is sort of a bad sign for the "ism."
Could you mention some of the specific beliefs you think are wrong?
Having strong opinions on QM interpretations is "not even wrong."
LW's attitude on B is, at best, "arguable."
Donating to MIRI as an effective use of money is, at best, "arguable."
LW consequentialism is, at best, "arguable."
Shitting on philosophy.
Ratonalism as part of identity (aspiring rationalist) is kind of dangerous.
etc.
What I personally find valuable is "adapting the rationalist kung fu stance" for certain purposes.
Thank you.
B?
Bayesian.
[Edited formatting] Strongly agree. http://lesswrong.com/lw/huk/emotional_basilisks/ is an experiment I ran which demonstrates the issue. Eliezer was unable to -consider- the hypothetical; it "had" to be fought.
The reason being, the hypothetical implies a contradiction in rationality as Eliezer defines it; if rationalism requires atheism, and atheism doesn't "win" as well as religion, then the "rationality is winning" definition Eliezer uses breaks; suddenly rationality, via winning, can require irrational behavior. Less Wrong has a -massive- blind spot where rationality is concerned; for a web site which spends a significant amount of time discussing how to update "correctness" algorithms, actually posing challenges to "correctness" algorithms is one of the quickest ways to shut somebody's brain down and put them in a reactionary mode.
I don't think that's argued. It's also worth noting that the majority of MIRI's funding over it's history comes from a theist.
It seems to me that he did consider your hypothetical, and argued that it should be fought. I agree: your hypothetical is just another in the tedious series of hypotheticals on LessWrong of the form, "Suppose P were true? Then P would be true!"
BTW, you never answered his answer. Should I conclude that you are unable to consider his answer?
Eliezer also has Harry Potter in MoR withholding knowledge of the True Patronus from Dumbledore, because he realises that Dumbledore would not be able to cast it, and would no longer be able to cast the ordinary Patronus.
Now, he has a war against the Dark Lord to fight, and cannot take the time and risk of trying to persuade Dumbledore to an inner conviction that death is a great evil in order to enable him to cast the True Patronus. It might be worth pursuing after winning that war, if they both survive.
All this has a parallel with your hypothetical.
I 've notice that problem, but I think it is a bit dramatic to call it rationality breaking. I think it's more of a problem of calling two things, the winning thing amd the truth seeking thing, by one name.
Well...
QM: Having strong positive beliefs on the subject would be not-even-wrong. Ruling out some is much less so. And that's what he did. Note, I came to the same conclusion long before.
MIRI: It's not uncritically accepted on LW more than you'd expect given who runs the joint.
Identity: If you're not letting it trap you by thinking it makes you right, if you're not letting it trap you by thinking it makes others wrong, then what dangers are you thinking of? People will get identities. This particular one seems well-suited to mitigating the dangers of identities.
Others: more clarification required
I think there's plenty of criticism voiced about that concept on LW and there are articles advocating to keep one's identity small.
And yet...
A universe where humans are running on brains with certain glitches that prevent them from coming to correct conclusions through reasoning about specific topics.