Chrysophylax

Karma: 140

Chrysophylax 31 Jan 2013 17:30 UTC
32 points
in reply to: Robin_Hanson2’s comment on: Applause Lights
I tried this for my valedictoral speech and I gave up after about 15 seconds due to the laughter.

My preferred method is to use long sentences, to speak slowly and seriously, with great emphasis, and to wave my hands in small circles as I speak. If you don’t speak to this audience regularly, it is also a good idea to emphasise how grateful you are to be asked to speak on such an important occasion (and it is a very important occasion...). You get bonus points for using the phrase “just so chuffed”, especially if you use it repeatedly (a technique I learned from my old headmaster, who never expressed satisfaction in any other way while giving speeches).

I also recommend this technique, this way of speaking, to anyone who wishes to wind up, by which I mean annoy or irritate, a family member. It’s quite effective when used consistently, even if you only do it for a minute or two. Don’t you agree?

Chrysophylax 30 Jan 2013 23:18 UTC
16 points
on: The Futility of Emergence
As Eliezer requested, I offer my view on what emergence isn’t: emergence is not an explanation. When I say that a phenomenon is emergent, I am using a shorthand to say that I understand the basic rules, but I can’t form even a simple model of how they result in the phenomenon.

Take, for example, Langton’s Ant. The ant crawls around on an infinite grid of black and white squares, turning right at the centre of each white square and left ant the centre of each black square, and flipping the colour of the square it’s in each time it turns.

The first few hundred steps create simple patterns that are often symmetric, but after that the patterns Langton’s Ant produces become pseudorandom. If left to run for around 10000 steps, the Ant builds a highway—that is, it falls into a pattern of 104 of steps, and at the end of each cycle, it has moved diagonally and the cycle repeats. After millions of steps, the grid has a diagonal streak across it. As far as we know, the Ant always builds a highway.

Highways are emergent by the definition I use—that is, I know exactly how Langton’s Ant works, and therefore, in theory, know why it builds a highway, but I can’t form a model of its behaviour that I can actually use. I simply do not have a good enough brain to actually run Langton’s Ant. By this definition, conciousness is an emergent phenomenon (I know it’s caused by neurons, but I have no idea how) but the behaviour of gases is not (I know the ideal gas law and its predictions seem reasonable if I imagine a manageable number of molecules bumping about).

By my definition, emergent is much like blue. “It’s emergent!” and “It’s blue!” are both mysterious answers if I asked for an explanation, but useful answers if I asked for a description.

Chrysophylax 24 Mar 2014 9:43 UTC
12 points
in reply to: Jayson_Virissimo’s comment on: Stranger Than History

Are there any countries that allow gay marriage that don’t have a longish history of Christianity?

No. There are 17 countries that allow it and 2 that allow it in some jurisdictions. A list may be found here: http://www.pewforum.org/2013/12/19/gay-marriage-around-the-world-2013/

There have been plenty of cultures where homosexuality was accepted; classical Greece and Rome, for example. Cultures where marriage is predominantly a governmental matter rather than a religious one are all, as far as I am aware, heavily influenced by the cultures of western Europe. One might also observe that all of these countries are industrial or post-industrial, and have large populations of young people with vastly more economic and sexual freedom than occured before the middle of the 20th century. One might also observe that China, Japan and South Korea seem to be the only countries at this level of economic development that were not culturally dominated by colonial states.

The fact that a history of Christianity is positively correlated with approval for gay marriage does not imply that Christian memes directly influence stances on homosexuality. Christianity spread around the world alongside other memes (such as democracy and case law). Those countries where European colonies were culturally dominant also received the industrial revolution and the immense increases in personal rights that came as a consequence of the increased economic and political power of the working class. One might also point out that thinking black people are inferior is a meme that arose from the slave trade in Christian semi-democracies.

There seems to be abundant evidence that the Abrahamic religions have strongly influenced societal views worldwide with regard to sexual morals; indeed, I cannot imagine a remotely plausible argument for this being untrue. I also wish to observe that Eastern Orthodox Christianity survived the USSR and still affects cultural values in Russia; it seems highly improbable that it did not influence Russian culture in the 1930s.

Chrysophylax 14 Jan 2014 21:45 UTC
10 points
in reply to: pnrjulius’s comment on: Serious Stories
Because a child who doesn’t find pain unpleasant is really, really handicapped, even in the modern world. The people who founded A Gift of Pain had a daughter with pain asymbolia who is now mostly blind, amongst other disabilities, through self-inflicted damage. I’m not sure whether leprosy sufferers have the no-pain or no-suffering version of pain insensitivity (I think the former) but apparently it’s the reason they suffer such damage.

This book seems to be a useful source for people considering the question of whether pain could be improved.

Chrysophylax 14 Jan 2014 12:06 UTC
9 points
on: What rationality material should I teach in my game theory course
Newcombe-style problems, including the Prisoner’s Dilemma, and the difference between rationality-as-winning and rationality-as-rituals-of-cognition.

Chrysophylax 13 Jan 2014 13:32 UTC
9 points
in reply to: Tenoke’s comment on: Karma query
Thank you. I was not aware that there is an Open Thread; that is clearly a superior option. My apologies.

Chrysophylax 29 Dec 2013 19:16 UTC
9 points
in reply to: DonGeddis’s comment on: Undiscriminating Skepticism

This doesn’t argue that infants have zero value, but instead that they should be treated more like property or perhaps like pets (rather than like adult citizens).

You haven’t taken account of discounted future value. A child is worth more than a chimpanzee of equal intelligence because a child can become an adult human. I agree that a newborn baby is not substantially more valuable than a close-to-term one and that there is no strong reason for caring about a euthanised baby over one that is never born, but I’m not convinced that assigning much lower value to young children is a net benefit for a society not composed of rationalists (which is not to say that it is not an net benefit, merely that I don’t properly understand where people’s actions and professed beliefs come from in this area and don’t feel confident in my guesses about what would happen if they wised up on this issue alone).

The proper question to ask is “If these resources are not spent on this child, what will they be spent on instead and what are the expected values deriving from each option?” Thus contraception has been a huge benefit to society: it costs lots and lots of lives that never happen, but it’s hugely boosted the quality of the lives that do.

I do agree that willingness to consider infanticide and debate precisely how much babies and foetuses are worth is a strong indicator of rationality.

Chrysophylax 29 Dec 2013 11:24 UTC
8 points
on: How the Grinch Ought to Have Stolen Christmas
Actually, causing poverty is a poor way to stop gift-giving. Even in subsistence economies, most farm households are net purchasers of the staple food; even very poor households support poorer ones in most years. (I have citations for this but one is my own working paper, which I don’t currently have access to, and the other is cited in that, so you’ll have to go without.) Moreover, needless gift-giving to the point of causing financial difficulties is fairly common in China (see http://www.economist.com/news/china/21590914-gift-giving-rural-areas-has-got-out-hand-further-impoverishing-chinas-poor-two-weddings-two).

The universe is always, eternally trying to freeze everyone to (heat) death, and will eventually win.

Chrysophylax 9 Jan 2014 15:07 UTC
7 points
in reply to: XiXiDu’s comment on: The genie knows, but doesn’t care
XiXiDu, I get the impression you’ve never coded anything. Is that accurate?
1. Present-day software is better than previous software generations at understanding and doing what humans mean.
Increasing the intelligence of Google Maps will enable it to satisfy human intentions by parsing less specific commands.
Present-day everyday software (e.g. Google Maps, Siri) is better at doing what humans mean. It is not better at understanding humans. Learning programs like the one that runs PARO appear to be good at understanding humans, but are actually following a very simple utility function (in the decision sense, not the experiental sense); they change their behaviour in response to programmed cues, generally by doing more/less of actions associated with those cues (example: PARO “likes” being stroked and will do more of things that tend to preceed stroking). In each case of a program that improves itself, it has a simple thing it “wants” to optimise and makes changes according to how well it seems to be doing.

Making software that understands humans at all is beyond our current capabilities. Theory of mind, the ability to recognise agents and see them as having desires of their own, is something we have no idea how to produce; we don’t even know how humans have it. General intelligence is an enormous step beyond programming something like Siri. Siri is “just” interpreting vocal commands as text (which requires no general intelligence), matching that to a list of question structures (which requires no general intelligence; Siri does not have to understand what the word “where” means to know that Google Maps may be useful for that type of question) and delegating to Web services, with a layer of learning code to produce more of the results you liked (i.e., that made you stop asking related questions) in the past. Siri is using a very small built-in amount of knowledge and an even smaller amount of learned knowledge to fake understanding, but it’s just pattern-matching. While the second step is the root of general intelligence, it’s almost all provided by humans who understood that “where” means a question is probably to do with geography; Siri’s ability to improve this step is virtually nonexistent.

catastrophically worse than all previous generations at doing what humans mean

The more powerful something is, the more dangerous it is. A very stupid adult is much more dangerous than a very intelligent child because adults are allowed to drive cars. Driving a car requires very little intelligence and no general intelligence whatsoever (we already have robots that can do a pretty good job), but can go catastrophically wrong very easily. Holding an intelligent conversation requires huge amounts of specialised intelligence and often requires general intelligence, but nothing a four-year-old says is likely to kill people.

It’s much easier to make a program that does a good job at task-completion, and is therefore given considerable power and autonomy (Siri, for example), than it is to make sure that the program never does stupid things with its power. Developing software we already have could easily lead to programs being assigned large amounts of power (e.g., “Siri 2, buy me a ticket to New York”, which would almost always produce the appropriate kind of ticket), but I certainly wouldn’t trust such programs to never make colossal screw-ups. (Siri 2 will only tell you that you can’t afford a ticket if a human programmer thought that might be important, because Siri 2 does not care that you need to buy groceries, because it does not understand that you exist.)

I hope I have convinced you that present software only fakes understanding and that developing it will not produce software that can do better than an intelligent human with the same resources. Siri 2 will not be more than a very useful tool, and neither will Siri 5. Software does not stop caring because it has never cared.

It is very easy (relatively speaking) to produce code that can fake understanding and act like it cares about your objectives, because this merely requires a good outline of the sort of things the code is likely to be wanted for. (This is the second stage of Siri outlined above, where Siri refers to a list saying that “where” means that Google Maps is probably the best service to outsource to.) Making code that does more of the things that get good results is also very easy.

Making code that actually cares requires outlining exactly what the code is really and truly wanted to do. You can’t delegate this step by saying “Learn what I care about and then satisfy me” because that’s just changing what you want the code to do. It might or might not be easier than saying “This is what I care about, satisfy me”, but at some stage you have to say what you want done exactly right or the code will do something else. (Currently getting it wrong is pretty safe because computers have little autonomy and very little general intelligence, so they mostly do nothing much; getting it wrong with a UFAI is dangerous because the AI will succeed at doing the wrong thing, probably on a big scale.) This is the only kind of code you can trust to program itself and to have significant power, because it’s the only kind that will modify itself right.

You can’t progress Siri into an FAI, no matter how much you know about producing general intelligence. You need to know either Meaning-in-General, Preferences-in-General or exactly what Human Prefernces are, or you won’t get what you hoped for.

Another perspective: the number of humans in history who were friendly is very, very small. The number of humans who are something resembling capital-F Friendly is virtually nil. Why should “an AI created by humans to care” be Friendly, or even friendly? Unless friendliness or Friendliness is your specific goal, you’ll probably produce software that is friendly-to-the-maker (or maybe Friendly-to-the-maker, if making Friendly code really is as easy as you seem to think). Who would you trust with a superintelligence that did exactly what they said? Who would you trust with a superintelligence that did exactly what they really wanted, not what they said? I wouldn’t trust my mother with either, and she’s certainly highly intelligent and has my best interests at heart. I’d need a fair amount of convincing to trust me with either. Most humans couldn’t program AIs that care because most humans don’t care themselves, let alone know how to express it.

Chrysophylax 28 Dec 2013 22:27 UTC
6 points
in reply to: coredumperror’s comment on: How the Grinch Ought to Have Stolen Christmas
While walking through the town shopping centre shortly before Christmas, my mother overheard a conversation between two middle-aged women, in which one complained of the scandalous way in which the Church is taking over Christmas. She does not appear to have been joking.

This occured in Leatherhead, a largish town a little south of London in the UK. It is fairly wealthy, with no slummy areas and a homeless population of approximately zero. It is not a regional shopping hub; if they came specifically to shop, they almost certainly came from villages. Of the local schools, only the main high school is not officially Christian. We have at least three churches in town, one of which rings its bells every hour two streets from the shopping centre, but no mosque and no synagogue.

I think it is safe to say that someone has stolen Christmas, but I suspect they were intending to sell it, not destroy it.

Chrysophylax 28 Jan 2014 19:12 UTC
5 points
in reply to: VincentYu’s comment on: [Error communicating with LW2 server]
Upvoted.

To clarify: VNM-utility is a decision utility, while utilitarianism-utility is an experiential utility. The former describes how a rational agent behaves (a rational agent always maximises VNM-utility) and is therefore ordinal, as it doesn’t matter what values we assign to different outcomes as long as the preference order does not change. The latter describes what values should be ascribed to different experiences and is therefore cardinal, as changing the numbers matters even when decisions don’t change.

Chrysophylax 13 Jan 2014 14:13 UTC
5 points
on: Open Thread for January 8 − 16 2014
A query about threads:

I posted a query in discussion because I didn’t know this thread exists. I got my answer and was told that I should have used the Open Thread, so I deleted the main post, which the FAQ seems to be saying will remove it from the list of viewable posts. Is this sufficient?

I also didn’t see my post appear under discussion/new before I deleted it. Where did it appear so that other people could look at it?

Chrysophylax 30 Jan 2013 18:51 UTC
5 points
in reply to: V_V’s comment on: Fake Explanations
Students who do not care about education do get away with not knowing anything. Detention is not much of a punishment when you don’t show up.

It is difficult to prevent a student who cares deeply about eduction from admitting ignorance, since admitting ignorance is necessary in asking for explanations. The difficult task is persuading students who care about doing well to seek knowledge, rather than good marks. These students are not motivated enough to learn of their own accord—they never volunteer answers or ask questions openly, because they care more about not being thought ignorant (or, of course, keen) than about not being ignorant.

The point is not to allow students to “get away with” admitting ignorance. There is a vast difference between not knowing the answer and not wanting to know. Personally, I have never found it hard to tell the difference between students who don’t want to know and students who don’t want to be judged by their peers.

It has to teach you how to behave in the world, where you often have to make choices based on incomplete information.

It is very rarely a bad idea to publicly admit that you might be wrong, especially when you are guessing. A school that does not teach the importance of separating your beliefs and your ego has failed miserably. Whatever else it has taught, it has not taught its students how to learn.

Chrysophylax 27 Mar 2014 0:08 UTC
4 points
in reply to: Eugine_Nier’s comment on: Stranger Than History
I’ve read it. Views about black people in the Islamic Golden Age were not the cause of views about black people in the nations participating in the transatlantic slave trade; a quick check of Wikipedia confirms that slavery as a formal institution had to redevelop in the English colonies, as chattel slavery had virtually disappeared after the Norman Conquest and villeinage was largely gone by the beginning of the 17th century. One might as well argue that the ethic of recipricocity in modern Europe owes its origin to Confucian ren.

Chrysophylax 24 Mar 2014 0:08 UTC
4 points
in reply to: Jayson_Virissimo’s comment on: Stranger Than History
I agree with the statements of fact but not with the inference drawn from them. While Jiro’s argument is poorly expressed, I think it is reasonable to say that opposition to homosexuality would not have been the default stance of the cultures of or derived from Europe if not for Christianity being the dominant religion in previous years. While the Communists rejected religion, they did not fully update on this rejection, but rather continued in many of the beliefs that religion had caused to be part of their culture.

I am not sure that “the atheists actually thought gay marriage was a sane idea but didn’t say so for fear of how they’d look to their religious neighbors” was Jiro’s position, but I think that it is a straw man.

Chrysophylax 13 Jan 2014 14:59 UTC
4 points
in reply to: Lalartu’s comment on: Stupid Questions Thread—January 2014

4) Subscribing for cryonics is generally a good idea. Result if widespread: these costs significantly contribute to worldwide economic collapse.

Under the assumption that cryonics patients will never be unfrozen, cryonics has two effects. Firstly, resources are spent on freezing people, keeping them frozen and researching how to improve cryonics. There may be fringe benefits to this (for example, researching how to freeze people more efficiently might lead to improvements in cold chains, which would be pretty snazzy). There would certainly be real resource wastage.

The second effect is in increasing the rate of circulation of the currency; freezing corpses that will never be revived is pretty close to burying money, as Keynes suggested. Widespread, sustained cryonic freezing would certainly have stimulatory, and thus inflationary, effects; I would anticipate a slightly higher inflation rate and an ambiguous effect on economic growth. The effects would be very small, however, as cryonics is relatively cheap and would presumably grow cheaper. The average US household wastes far more money and real resources by not recycling or closing curtains and by allowing food to spoil.

Chrysophylax 29 Jan 2013 22:58 UTC
4 points
in reply to: royf’s comment on: Fake Explanations

But in this case it’s not the rational thing to say when you use decision theory.

I disagree. The proper response to not knowing the answer is to admit to not knowing and then give your best guess, not to try to hide your ignorance, because if you succeed then the teacher doesn’t know you need help. A student who is more concerned with not displaying ignorance than with not being ignorant is not trying to learn, which is not rational. That which can be destroyed by the truth should be, and it probably won’t be if you try to avoid finding out what the truth is.

You can easily do worse than maximum entropy if you guess at random.

If you base your guess correctly on an incomplete model of reality, which you’ve constructed correctly from past observations, you can never do worse, on average, than maximum entropy. More evidence can never lead to less information (as per the Data Processing Inequality).

The key phrase here is “on average”. If you guess at random from all possible explanations of a given phenomenon, you will, on average, die before guessing the correct answer. There is a reason the monkeys with typewriters are given infinite time to reproduce Hamlet.

Moreover, as the set of answers considered increases in size, the expected utility from giving any one answer tends towards the expected utility of a wrong answer. As long as giving the wrong answer gives less utility than admitting ignorance, admitting ignorance is almost always the utility maximising option if you don’t know.

Furthermore, “getting it right” [...] does not necessarily mean that you possess any anticipation-controllers.

On the contrary, it mean exactly that. Being rewarded for predictive powers improves your model of the world, whereas “I don’t know” is an excuse for not knowing.

If I write down a number and then take a number from a table of random numbers, and the numbers are the same, does this mean that I’m psychic? Because if getting the correct answer means that I have useful anticipation controllers then I must be.

“I don’t know” is not an excuse for not knowing. That makes no sense whatsoever. “I don’t know” is a statement about whether I know something or not, not a statement about whether I ought to know. If you can’t admit fallibility then you will never learn anything.

The points you make about the benefits of testing students’ knowledge are true. Unfortunately, they miss the point—while it is important not to penalise guessing incorrectly, so as not to dissuade admitting ignorance, it is much better to actively reward admitting that you have tried and failed. If a confused student does not always seek an explanation, the reward for seeking explanations isn’t large enough yet. If students are content to remain ignorant, something is seriously wrong with your system for making students less ignorant.

Chrysophylax 24 Mar 2014 8:49 UTC
3 points
on: Open thread, 11-17 March 2014
Yet another karma query: yesterday my karma was 37. Today my karma is 12 and I am at −2 karma for the last 30 days. What’s going on here?

Chrysophylax 23 Mar 2014 23:57 UTC
3 points
in reply to: Lumifer’s comment on: Stranger Than History
An argument is valid if, given true premises, it always and exclusively produces true conclusions. A valid argument in this context might therefore be “given that we wish to maximise social welfare (A) and that allowing gay marriage increases social welfare (B), we should allow gay marriage (C)”. A and B really do imply C. Some people contend that the argument is not sound (that is, that its conclusion is false) because at least one of its premises is not true (reflecting reality); I am not aware of anyone who contends that it is invalid.

Jiro is contending that people who oppose gay marriage do not do so because they have valid arguments for doing so; if we were to refute their arguments they would not change their minds. Xe has argued above that people (as a group) did not stop being anti-homosexuality for rational reasons, i.e. because the state of the evidence changed in important ways or because new valid arguments were brought to bear, but rather for irrational reasons, such as old people dying.

The fact that Jiro considers it rational to believe that gay marriage is a good thing, and thus that people’s beliefs are now in better accord with an ideal reasoner’s beliefs (“are more rational”), does not contradict Jiro’s belief that popular opinion changed for reasons other than those that would affect a Bayesian. Eugine_Nier appears to be conflating two senses of “rational”.

As RichardKennaway observes, we ought to ask why Jiro believes that we should allow gay marriage. I suspect the answer will be close to “because it increases social welfare”, which seems to be a well-founded claim.

Chrysophylax 17 Feb 2013 14:29 UTC
3 points
on: Einstein’s Superpowers
There is woolly thinking going on here, I feel. I recommend a game of Rationalist’s Taboo. If we get rid of the word “Einstein”, we can more clearly see what we are talking about. I do not assign a high value to my probabilty of making Einstein-sized contributions to human knowledge, given that I have not made any yet and that ripe, important problems are harder to find than they used to be. Einstein’s intellectual accomplishments are formidable—according to my father’s assessment (and he has read far more of Einstein’s papers than I), Einstein deserved far more than one Nobel prize.

On the other hand, if we consider three strong claimants to the title of “highest-achieving thinker ever”, namely Einstein, Newton and Archimedes, we can see that their knowledge was very much less formidable. If the test was outside his area of expertise, I would consider a competition between Einstein and myself a reasonably fair fight—I can imagine either of us winning by a wide margin, given an appropriate subject. Newton would not be a fair fight, and I could completely crush Archimedes at pretty much anything. There are millions of people who could claim the same, millions who could claim more. Remember that there are no mysterious answers, and that most of the work is done in finding useful hypotheses—finding a new good idea is hard, learning someone else’s good idea is not. I do not need to claim to be cleverer than Newton to claim to understand pretty much everything better than he ever did, nor to consider it possible that I could make important contributions. If I had an important problem, useful ideas about it that had been simmering for years and was clearly well ahead of the field, I would consider it reasonably probable that I would make an important breakthrough—not highly probable, but not nearly as improbable as it might sound. It might clarify this point by saying that I would place high probability on an important breakthrough occuring—if there is anyone in such a position, I conclude that there are probably others (or there will be soon), and so the one will probably have at least met the people who end up making the breakthrough. It is useful to remember that for every hero who made a great scientific advance, there were probably several other people who were close to the same answer and who made significant contributions to finding it.