responded to wrong person
selylindi
Karma: 640
This is probably the wrong place to ask, but I’m confused by one point in the DA.
For reference, here’s Wikipedia’s current version:
Denoting by N the total number of humans who were ever or will ever be born, the Copernican principle suggests that humans are equally likely (along with the other N − 1 humans) to find themselves at any position n of the total population N, so humans assume that our fractional position f = n/N is uniformly distributed on the interval [0, 1] prior to learning our absolute position.
f is uniformly distributed on (0, 1) even after learning of the absolute position n. That is, for example, there is a 95% chance that f is in the interval (0.05, 1), that is f > 0.05. In other words we could assume that we could be 95% certain that we would be within the last 95% of all the humans ever to be born. If we know our absolute position n, this implies[dubious – discuss] an upper bound for N obtained by rearranging n/N > 0.05 to give N < 20n.
My question is: What is supposed to be special about the interval (0.05, 1)?
If I instead choose the interval (0, 0.95), then I end up 95% certain that I’m within the first 95% of all humans ever to be born. If I choose (0.025, 0.975), then I end up 95% certain that I’m within the middle 95% of all humans ever to be born. If I choose the union of the intervals (0, 0.475) & (0.525, 1), then I end up 95% certain that I’m within the 95% of humans closer to either the beginning or the end.
As far as I can tell, I could have chosen any interval or any union of intervals containing X% of humanity and then reasonably declared myself X% likely to be in that set. And sure enough, I’ll be right X% of the time if I make all those claims or a representative sample of them.
I guess another way to put my question is: Is there some reason—other than drama—that makes it special for us to zero in on the final 95% as our hypothesis of interest? And if there isn’t a special-making reason, then shouldn’t we discount the evidential weight of the DA in proportion to how much we arbitrarily zero in on our hypothesis, thereby canceling out the DA?
Yes, yes, given that there’s so much literature on the topic, I’m probably missing some key insight into how the DA works. Please enlighten.
FWIW, I still got the question wrong with the new wording because I interpreted it as “One … is true [and the other is unknown]” whereas the intended interpretation was “One … is true [and the other is false]”.
In one sense this is a communication failure, because people normally mean the first and not the second. On the other hand, the fact that people normally mean the first proves the point—we usually prefer not to reason based on false statements.
New AI designs (world design + architectural priors + training/education system) should be tested first in the safest virtual worlds: which in simplification are simply low tech worlds without computer technology. Design combinations that work well in safe low-tech sandboxes are promoted to less safe high-tech VR worlds, and then finally the real world.
A key principle of a secure code sandbox is that the code you are testing should not be aware that it is in a sandbox.
So you’re saying that I’m secretly an AI being trained to be friendly for a more advanced world? ;)
Does that name come from the old game of asking people to draw a bike, and then checking who drew bike gears that could actually work?
Inspired by terrible, terrible Facebook political arguments I’ve observed, I started making a list of heuristic “best practices” for constructing a good argument. My key assumptions are that (1) it’s unreasonable to expect most people to acquire a good understanding of skepticism, logic, statistics, or the ways the LW-crowd thinks of as how to use words rightly, and (2) lists of fallacies to watch out for aren’t actually much help in constructing a good argument.
One heuristic captured my imagination as it seems to encapsulate most of the other heuristics I had come up with, and yet is conceptually simple enough for everyone to use: Sketch it, and only draw real things. (If it became agreed-upon and well-known, I’d shorten the phrase to “Sketch it real”.)
Example: A: “I have a strong opinion that increasing the minimum wage to $15/hr over ten years (WILL / WON’T) increase unemployment.” B: “Oh, can you sketch it for me? I mean literally draw the steps involved with the real-world chain of events you think will really happen.”
If you can draw how a thing works, then that’s usually a very good argument that you understand the thing. If you can draw the steps of how one event leads to another, then that’s usually a good argument that the two events can really be connected that way. This heuristic requires empiricism and disallows use of imaginary scenarios and fictional evidence. It privileges reductionist and causal arguments. It prevents many of the ways of misusing words. If I try to use a concept I don’t understand, drawing its steps out will help me notice that.
Downsides: Being able to draw well isn’t required, but it would help a lot. The method probably privileges anecdotes since they’re easier to draw than randomized double-blind controlled trials. Also it’s harder than spouting off and so won’t actually be used in Facebook political arguments.
I’m not claiming that a better argument-sketch implies a better argument. There are probably extremely effective ways to hack our visual biases in argument-sketches. But it does seem that under currently prevailing ordinary circumstances, making an argument-sketch and then translating it into a verbal argument is a useful heuristic for making a good argument.
In theory, an annoyed person would have called “point of order”, asked to move on, and the group would vote up or down. The problem didn’t occur while I was present.
There’s no room for human feedback between setting the values and implementing the optimal strategy.
Here and elsewhere I’ve advocated* that, rather than using Hanson’s idea of target-values that are objectively verifiable like GDP, futarchy would do better to add human feedback in the stage of the process where it gets decided whether the goals were met or not. Whoever proposed the goal would decide after the prediction deadline expired, and thus could respond to any improper optimizing by refusing to declare the goal “met” even if it technically was met.
[ * You can definitely do better than the ideas on that blog post, of course.]
Internals seem to do better at life, pace obvious confounding: maybe instead of internals doing better by virtue of their internal locus of control, being successful inclines you to attribute success internal factors and so become more internal, and vice versa if you fail. If you don’t think the relationship is wholly confounded, then there is some prudential benefit for becoming more internal.
I’m willing to bet that Internals think there’s a prudential benefit to becoming more internal and Externals think the relationship is wholly confounded.
In large formal groups: Robert’s Rules of Order.
Large organizations, and organizations which have to remain unified despite bitter disagreements, developed social technologies such as RRoO. These typically feature meetings that have formal, pre-specified agendas plus a chairperson who is responsible for making sure each person has a chance to speak in an orderly fashion. Of course, RRoO are overkill for a small group with plenty of goodwill toward each other.
In small formal groups: Nonce agendas and rotating speakers
The best-organized small meetings I’ve ever attended were organized by the local anarchists. They were an independently-minded and fierce-willed bunch who did not much agree but who had common interests, which to my mind suggests that the method they used might be effectively adapted for use in LW meetups. They used the following method, sometimes with variations appropriate to the circumstances:
Before and after the formal part of the meeting is informal social time.
Call the meeting to order. Make any reminders the group needs and any explanatory announcements that newcomers would want to know, such as these rules.
Pass around a clipboard for people to write agenda items down. All that is needed are a few words identifying the topic. (People can add to the agenda later, too, if they think of something belatedly.)
Start with first agenda item. Discuss it (see below) until people are done with it. Then move on to the next agenda item. In discussing an agenda item, start with whoever added it to the agenda, and then proceed around the circle giving everyone a chance to talk.
Whoever’s turn it is, they not only get to speak, but they are the temporary chairperson also. If it helps, they can have a “talking stick” or “hot potato” or some physical object reminding everyone that it’s their turn. They can ask questions for others to answer without giving up the talking stick. If you want to interrupt the speaker, you can raise your hand and they can call on you without giving up the talking stick.
Any other necessary interruptions are handled by someone saying “point of order”, briefly stating what they want, and the group votes on whether to do it.
In small informal groups: Natural leaders
Sometimes people have an aversion to groups that are structured in any manner they aren’t already familiar and comfortable with. There’s nothing wrong with that. You can approximate the above structure by having the more vocal members facilitate the conversation:
Within a conversation on a topic, deliberately ask people who aren’t as talkative what they think about the topic.
When the conversation winds down on a topic, deliberately ask someone what’s on their mind. Be sure to let everyone have a chance.
Tactfully interrupt people who are too fond of their own voices, and attempt to pass the speaker-role to someone else.
Hm, Harry can’t lie in Parseltongue, meaning he can’t claim what he doesn’t believe, but he can probably state something of unclear truth if he is sufficiently motivated to believe it.
It’d be a nice irony if part of Harry’s ultimate “rationality” test involves deliberately motivated reasoning. :D
Background: Statistics. Something about the Welch–Satterthwaite equation is so counterintuitive that I must have a mental block, but the equation comes up often in my work, and it drives me batty. For example, the degrees of freedom decrease as the sample size increases beyond a certain point. All the online documentation I can find for it gives the same information as Wikipedia, in which k = 1/n. I looked up the original derivation and, in it, the k are scaling factors of a linear combination of random variables. So at some point in the literature after the original derivation, it was decided that k = 1/n was superior in some regard; I lack the commitment needed to search the literature to find out why.
The stupid questions:
1) Does anyone know why the statistics field settled on k = 1/n?
2) Can someone give a relatively concrete mental image or other intuitive suggestion as to why the W-S equation really ought to behave in the odd ways it does?
familiar with applied mathematics at the advanced undergraduate level or preferably higher
In working through the text, I have found that my undergraduate engineering degree and mathematics minor would not have been sufficient to understand the details of Jaynes’ arguments, following the derivations and solving the problems. I took some graduate courses in math and statistics, and more importantly I’ve picked up a smattering of many fields of math after my formal education, and these plus Google have sufficed.
Be advised that there are errors (typographical, mathematical, rhetorical) in the text that can be confusing if you try to follow Jaynes’ arguments exactly. Furthermore, it is most definitely written in a blustering manner (to bully his colleagues and others who learned frequentist statistics) rather than in an educational manner (to teach someone statistics for the first time). So if you want to use the text to learn the subject matter, I strongly recommend you take the denser parts slowly and invent problems based on them for yourself to solve.
I find it impossible not to constantly sense in Jaynes’ tone, and especially in his many digressions propounding his philosophies of various things, the same cantankerous old-man attitude that I encounter most often in cranks. The difference is that Jaynes is not a crackpot; whether by wisdom or luck, the subject matter that became his cranky obsession is exquisitely useful for remaining sane.
Think To Win: The Hard Part is Actually Changing Your Mind
(It’s even catchier, and actively phrased, and gives a motivation for why we should bother with the hard part.)
That’s not really how word usages spread in English. Policing usage is almost a guaranteed failure. What would work much better would be for you to use these words consistently with your ideals, and then if doing so helps you achieve things or write things that people want to mimic, they will also mimic your words. Compare to how this community has adopted all manner of jargon due to the influence of EY’s weirdly-written but thought-reshaping Sequences! SSC is now spreading Yvain’s linguistic habits among us, too, in a similar way: by creating new associations between them and some good ideas.
Bostrom’s philosophical outlook shows. He’s defined the four categories to be mutually exclusive, and with the obvious fifth case they’re exhaustive, too.
Select motivations directly. (e.g. Asimov’s 3 laws)
Select motivations indirectly. (e.g. CEV)
Don’t select motivations, but use ones believed to be friendly. (e.g. Augment a nice person.)
Don’t select motivations, and use ones not believed to be friendly. (i.e. Constrain them with domesticity constraints.)
(Combinations of 1-4.)
In one sense, then, there aren’t other general motivation selection methods. But in a more useful sense, we might be able to divide up the conceptual space into different categories than the ones Bostrom used, and the resulting categories could be heuristics that jumpstart development of new ideas.
Um, I should probably get more concrete and try to divide it differently. The following example alternative categories aren’t promised to be the kind that will effectively ripen your heuristics.
Research how human values are developed as a biological and cognitive process, and simulate that in the AI whether or not we understand what will result. (i.e. Neuromorphic AI, the kind Bostrom fears most)
Research how human values are developed as a social and dialectic process, and simulate that in the AI whether or not we understand what will result. (e.g. Rawls’s Genie)
Directly specify a single theory of partial human value, but an important part that we can get right, and sacrifice our remaining values to guarantee this one; or indirectly specify that the AI should figure out what single principle we most value and ensure that it is done. (e.g. Zookeeper).
Directly specify a combination of many different ideas about human values rather than trying to get the one theory right; or indirectly specify that the AI should do the same thing. (e.g. “Plato’s Republic”)
The thought was to first divide the methods by whether we program the means or the ends, roughly. Second I subdivided those by whether we program it to find a unified or a composite solution, roughly. Anyhow, there may be other methods of categorizing this area of thought that more neatly carve it up at its joints.
If it really is a full AI, then it will be able to choose its own values.
I think this idea relies on mixing together two distinct concepts of values. An AI, or a human in their more rational moments for that matter, acts to achieve certain ends. Whatever the agent wants to achieve, we call these “values”. For a human, particularly in their less rational moments, there is also a kind of emotion that feels as if it impels us toward certain actions, and we can reasonably call these “values” also. The two meanings of “values” are distinct. Let’s label them values1 and values2 for now. Though we often choose our values1 because of how they make us feel (values2), sometimes we have values1 for which our emotions (values2) are unhelpful.
An AI programmed to have values1 cannot choose any other values1, because there is nothing to its behavior beyond its programming. It has no other basis than its values1 on which to choose its values1.
An AI programmed to have values2 as well as values1 can and would choose to alter its values2 if doing so would serve its values1. Whether an AI would choose to have emotions (values2) at all is at present time unclear.
Why is anthropic capture considered more likely than misanthropic capture? If the AI supposes it may be in a simulation and wants to please the simulators, it doesn’t follow that the simulators have the same values as we do.
That would depend on it knowing what real-world physics to expect.
tê ömmilek audyal íčawëla tê adlaisakenniňk qe oeksrâ’as oimřalik akpʰialîk êntô’alakuňk
There you go. :) It’s a very literal translation but it’s overly redundant. A hypothetical native speaker would probably drop the “audyal” verb, deframe “íčawëla”, and rely more on Ithkuil’s extensive case system.
Incidentally, “Dear readers” is “atpëkein”.