All of dyokomizo's Comments + Replies

FWIW we implemented the FDT, CDT, and EDT in Haskell a while ago.

https://github.com/DecisionTheory/DecisionTheory

1justinpombrio
Oh, excellent! It's a little hard to tell from the lack of docs, but you're modelling dilemmas with Bayesian networks? I considered that, but wasn't sure how to express Sleeping Beauty nicely, whereas it's easy to express (and gives the right answers) in my tree-shaped dilemmas. Have you tried to express Sleeping Beauty? And have you tried to express a dilemma like smoking lesion where the action that an agent takes is not the action their decision theory tells them to take? My guess is that this would be as easy as having a chain of two probabilistic events, where the first one is what the decision theory says to do and the second one is what the agent actually does, but I don't see any of this kind of dilemma in your test cases.

The post doesn't talk about nor imply a traditional war with congress approval. For example, placing a battleship in international waters but close enough China's maritime space is enough to trigger another Arkhipov situation. This is just a specific scenario, some don't lead to disastrous outcomes and some do. The post is intended to spark discussion, not set policy without debate, and point to the risks. As stated we entered 2020 in the worst level of risk ever, and then a pandemic happened, with a bunch of unthinkable events happening.

In my understanding, ban in sequences-inspired rationality, particularly for politically-charged topics, is a reminder that Politics is the Mind-Killer. I made it explicit in the text.

Wikipedia has an article for Considered Harmful. "Goto Considered Harmful" was the title an editor gave to Dijkstra's paper originally titled "A Case Against the Goto Statement". It's an informal tradition in computer science to write papers with this title pattern, including ""Considered Harmful" Essays Considered Harmful".

It was not intended to be misleading, only a reference to a crowd that I, perhaps erroneously, assumed would be familiar with this pattern.

Now, on the core of the argument. First the ep... (read more)

2jimmy
That article is just a list of a bunch of opinions people have and it is nothing more than a gossip piece. Literally all it does is repeat things like: and It does nothing to even begin commenting on why these ideas keep spreading, just that they are and who is spreading them. Likewise, exactly nothing in that article responds to anything I've said. It's no wonder that linking to trash like this doesn't convince anyone. To even get started you need to be able to link to things like this. Then you need to have people who can understand why that is credible explain it to their social circle who respect them and wouldn't understand it on their own. And that means you need an army of people who are capable of empathizing with the very real concerns that these "conspiracy theorists" have instead of falling into the trap of arrogance to hide from their own difficulties in being persuasive and credible. Yes, it's hard. Let's get to work.
Answer by dyokomizo50

miniKanren is a logic/relational language. It's been used to solve questions related to programs. For example, once you give miniKanren a description of the untyped λ-calculus extended with integers you can ask it "give me programs that result in 2" and it'll enumerate programs from the constant "2" to "1 + 1" to more complicated versions using λ-expressions. It can even find quines (if the described language supports it).

http://minikanren.org/

Answer by dyokomizo30

The Nanopass Framework is built for that:

"The nanopass framework provides a tool for writing compilers composed of several simple passes that operate over well-defined intermediate languages. The goal of this organization is both to simplify the understanding of each pass, because it is responsible for a single task, and to simplify the addition of new passes anywhere in the compiler."

https://nanopass.org/

https://docs.racket-lang.org/nanopass/index.html

I'm going again, it was too fun/interesting to miss.

Around São Paulo, yes. Around LW, not much anymore, I mostly read it via feed reader.

dyokomizo-20

This model seems to be reducible to "people will eat what they prefer".

A good model would be able to reduce the number of bits to describe a behavior, if the model requires to keep a log (e.g. what particular humans prefer to eat) to predict something, it's not much less complex (i.e. bit encoding) than the behavior.

-2newerspeak
No, because preferences are revealed by behavior. Using revealed preferences is a good heuristic generally, but it's required if you're right that explanations for behavior are mostly post-hoc rationalizations. So: People eat what they prefer. What they prefer is what they wind up having eaten. Ergo, people eat what they eat.
6AdeleneDawner
Maybe I've misunderstood. It seems to me that your original prediction has to refer either to humans as a group, in which case Luke's counterexample is a good one, or humans as individuals, in which case my counterexample is a good one. It also seems to me that either counterexample can be refined into a useful prediction: Humans in general don't eat petroleum products. I don't eat spicy food. Corvi doesn't eat meat. All of those classes of things can be described more efficiently than making lists of the members of the sets.

I agree vague is not a good word choice. Irrelevant (using relevancy as it's used to describe search results) is a better word.

I would classify such kinds of predictions as vague, after all they match equally well for every human being in almost any condition.

7Douglas_Knight
I think "vague" is a poor word choice for that concept. "(not) informative" is a technical term with this meaning. There are probably words which are clearer to the layman.

How about a prediction that a particular human will eat bacon instead of jalapeno peppers? (I'm particularly thinking of myself, for whom that's true, and a vegetarian friend, for whom the opposite is true.)

dyokomizo520

There's no way to create a non-vague, predictive, model of human behavior, because most human behavior is (mostly) random reaction to stimuli.

Corollary 1: most models explain after the fact and require both the subject to be aware of the model's predictions and the predictions to be vague and underspecified enough to make astrology seems like spacecraft engineering.

Corollary 2: we'll spend most of our time in drama trying to understand the real reasons or the truth about our/other's behavior even when presented with evidence pointing to the randomness of our actions. After the fact we'll fabricate an elaborate theory to explain everything, including the evidence, but this theory will have no predictive power.

This (modulo the chance it was made up) is pretty strong evidence that you're wrong. I wish it was professionally ethical for psychologists to do this kind of thing intentionally.

8[anonymous]
How detailed of a model are you thinking of? It seems like there are at least easy and somewhat trivial predictions we could make e.g. that a human will eat chocolate instead of motor oil.
5Perplexed
Downvoted in agreement. But I think that the randomness comes from what programmers call "race conditions" in the timing of external stimuli vs internal stimuli. Still, these race conditions make prediction impossible as a practical matter.

It doesn't seem to me that you have an accurate description of what a super-smart person would do/say other than match your beliefs and providing insightful thought. For example, do you expect super-smart people to be proficient in most areas of knowledge or even able to quickly grasp the foundations of different areas through super-abstraction? Would you expect them to be mostly unbiased? Your definition needs to be more objective and predictive, instead of descriptive.

1taw
I don't know what's the correct super-smartness cluster, so I cannot make objective predictive definition, at least yet. There's no need to suffer from physics envy here - a lot of useful knowledge has this kind of vagueness. Nobody managed to define "pornography" yet, and it's far easier concept than "super-smartness". This kind of speculation might end up with something useful with some luck (or not). Even defining by example would be difficult. My canonical examples would be Feynman and Einstein - they seem far smarter than the "normally smart" people. Let's say I collected a sufficiently large sample of "people who seem super-smart", got as accurate information about them as possible, and did a proper comparison between them and background of normally smart people (it's pretty easy to get good data on those, even by generic proxies like education - so I'm least worried about that) in a way that would be robust against even large number of data errors. That's about the best I can think of. Unfortunately it will be of no use as my sample will be not random super-smart people but those super-smart people who are also sufficiently famous for me to know about them and be aware of their super-smartness. This isn't what I want to measure at all. And I cannot think of any reasonable way to separate these. So the project is most likely doomed. It was interesting to think about this anyway.

How would you describe the writing patterns of super-smart people? Similarly, how would meeting/talking/debating them would feel like?

4taw
I think my comment was rather vague, and people aren't sure what I meant. This is all my impressions, as far as I can tell evidence of all that is rather underwhelming; I'm writing this more to explain my thought than to "prove" anything. It seems to me that people come in different level of smartness. There are some people with all sort of problems that make them incapable of even human normal, but let's ignore them entirely here. Then, there are normal people who are pretty much incapable of original highly insightful thought, critical thinking, rationality etc. They can usually do OK in normal life, and can even be quite capable in their narrow area of expertise and that's about it. They often make the most basic logic mistakes etc. Then there are "smart" people who are capable of original insight, and don't get too stupid too often. They're not measuring example the same thing, but IQ tests are capable of distinguishing between those and the normal people reasonably well. With smart people both their top performance and their average performance is a lot better than with average people. In spite of that, all of them very often fail basic rationality for some particular domains they feel too strongly about. Now I'm conflicted if people who are so much above "smart" as "smart" is above normal really exists. A canonical example of such person would be Feynman - from my limited information he seems to be just so ridiculously smart. Eliezer seems to believe Einstein is like that, but I have even less information about him. You can probably think of a few such other people. Unfortunately there's a second observation - there's no reason to believe such people existed only in the past, or would have aversion to blogging - so if super-smart people exist, it's fairly certain that some blogs of such people exist. And if such blogs existed, I would expect to have found a few by now. And yet, every time it seemed to me that someone might just be that smart and I start

Hi, I'm Daniel. I've read OB for a long time and followed on LW right in the beginning, but work /time issues in the last year made my RSS reading queue really long (I had all LW posts in the queue). I'm a Brazilian programmer, long time rationalist and atheist.

Hi, I'm a lurker mostly because I was reading these off my RSS queue (I accumulated thousands of entries in my RSS reader in the last year due to work/time issues),

0Paul Crowley
Hi, welcome to Less Wrong, thanks for delurking!

Sao Paulo, Brazil

0Gust
Same! Are you still around?