- Math - moomers.org

Fwd: Ethicophysics I
by Eric Purdy 30 May '18

30 May '18

---------- Forwarded message ---------- From: Eric Purdy <epurdy(a)uchicago.edu> Date: Wed, May 30, 2018 at 12:26 PM Subject: Ethicophysics I To: readings <readings(a)moomers.org> Abstract: What are Good and Evil? How do we explain these concepts to a computer sufficiently well that we can be assured that the computer will understand them in the same sense as humans understand them? These are hard questions, and people have often despaired of finding any answers to the AI safety problem. In this paper, we lay out a theory of ethics modeled on the laws of physics. The theory has two key advantages: it squares nicely with most human moral intuitions, and it is amenable to rather straightforward com- putations that a computer could easily perform if told to. It therefore forms an ideal foundation for solving the AI safety problem. -- -Eric -- -Eric

1 0

truth and beauty bux
by Eric Purdy 26 May '18

26 May '18

Some notes I put together on how to make a cryptocurrency that isn't morally bankrupt. Comments welcome! Especially comments from people who are unfamiliar with cryptocurrency, as it's always easy to miss large amounts of context that would be necessary for such readers. Also, if any programmers are interested in helping me build something like this, I've started a github repo for the project here: https://github.com/advancedresearch/truth_beauty_bux . Will try to start pushing some stuff to it this week if people think the ideas make sense and are worth pursuing. -- -Eric

1 0

thunder token
by Eric Purdy 23 Apr '18

23 Apr '18

We've talked before on readings about how fixing the carbon footprint of the blockchain is a technico-moral imperative. Recently stumbled across this project, which seems to be at least using the right buzzwords. Curious to see if they have the math to back it up or not... they have no white paper on their website, but I think they can probably pull it off with some luck. Or at least, such is my hope, given my love of the blockchain and my love of a human-habitable planet. https://www.thundertoken.com/ -- -Eric

1 1

intersubjectivity and the free will / free won't paradox
by Eric Purdy 11 Apr '18

11 Apr '18

In this email I summarize some previous work on the species Homo sapiens and suggest future directions of study. The first work we note is the Ph.D. thesis of Dr. Raber. (Dr. Raber: would you mind posting a copy of your thesis and/or an Amazon link to your book if it's already out?) We posit that this is a tolerably complete description of the maturational process and mating habits of the human mind. (Also, Dr. Raber, if you feel moved to share any extracts from your second book, would love to see those! It would be super relevant to some threads we're having about economics, if I understand the thrust of it correctly. (My gloss: what are the proper uses of luxury in motivating human beings to excellence and public service?)) We posit that it is possible, via affiliation economics and choice architecture and computational linguistics and digital humanities, to assemble a tolerably complete picture of the pressures that any given human being is under at any given point in time. Presumably this fact is known to various governments of the world, and has been since the height of the Cold War. Furthermore, we posit that human beings do in fact have free will. It feels pretty obvious if you're driving one of the damn things, anyway. But, of course, biological materialism and physical determinism hold sway over the physical and biological universes. How do we reconcile these two facts? Basically, you never know for sure which way someone is going to jump, because you never have a complete copy of the inside of their mind. You can put people in arbitrarily scary and fucked-up and tempting situations, but they'll still surprise you. As Natalie Portman says in V for Vendetta, that last inch of one's will remains free. A person who is willing to die under torture can generally do whatever the fuck they want. (Ofc, dying under torture is a fairly likely outcome if you make the mistake of telling anyone this fact when they threaten you.) Finally, as glue to make all of this cohere in a way that will make sense to the academic humanities, we offer the novels of David Foster Wallace. I put it to you that they are about free will and mind control. The cryptogram for mind control in Broom of the System is, I believe, the Devil's wooden leg. The cryptogram for mind control in Infinite Jest is annular fusion. (Think about it for a second - imagine watching a television that then floats through the air and smashes you in the back of the head - it traces out an annulus.) The cryptogram for mind control in The Pale King is the Pale King, who is a DFW self-insertion character, as are most of the characters in that book. As further glue, we offer the play Hamlet, by William Shakespeare. I put it to you that it is a play about how to kill William Shakespeare with your words in a world where mind reading and mind control are widely distributed among the populace; Hamlet is Willie's self-insertion character, and all the other characters are various figments of Bill's imagination as he worked through the problem of trying to keep himself from dying or going nuts. Would love to hear any thoughts that this email triggered from anyone who I sent it to! -- -Eric

2 6

a unified theory of language acquisition and evolution
by Eric Purdy 07 Apr '18

07 Apr '18

Here are some thoughts on linguistics. I haven't checked all the details, but it all seems to hang together and I think it's a good solid foundation to build a new theory of linguistics on. Please help me finish the theory! Erica: I've added you to this thread so that you can contribute some knowledge of classical theoretical linguistics to the discussion. Feel free to rope in any other linguists who you think might be better able to kick the tires of the theory, so to speak. If you want to make sure you're current with all the latest stuff, maybe go ahead and subscribe to math(a)moomers.org, my official listhost for sharing my mathematical ideas with the world. -- -Eric

1 2

affiliation economics
by Eric Purdy 05 Apr '18

05 Apr '18

I wish to construct a new paradigm for the social sciences, that unifies economics and all the math that economics has with all of the deeply true stuff that the other social sciences know about. When this paradigm is full-grown, it should kill both neoliberalism and so-called "race realism". Every paradigm starts from what I call a loose thread, a thing that no one can explain but everyone agrees is true. In this case, the loose thread that I think is most helpful is, what is IQ? What does it measure? I assert that it is mostly an empty signifier; there is no such thing as in-born greatness. All greatness comes from life experience. People who seem stupid or crazy to you are just trying to accomplish something different with their lives than what you think they should be trying to accomplish. Dr. Bishop: you are one of the people I choose to create this paradigm. Your graduate work on social network formation is essential here, as is your vast experience with prediction markets. Ms. Laurenson: you are the other person I choose to create this paradigm. Your vast experience in the social justice world and your experience with the world of journalism are both essential here. Remember that conversation we had about the privilege-stigma axis being essentially a rich-poor axis? We also need someone with some real economics and finance training. Austin - maybe that's you? I know you don't have much training but you understand the market-making problem as well as anyone does, I think. Someone could also try to talk to my ex-wife about this, but we're not currently on speaking terms I don't think. You will need Purdy's theory of "social facts", which I enclose some random musings on. -- -Eric

2 3

hello everyone
by Eric Purdy 04 Apr '18

04 Apr '18

Probably going to start posting some stuff tomorrow. How is everyone? -- -Eric

4 4

possible worlds of complexity theory
by Eric Purdy 04 Apr '18

04 Apr '18

https://www.karlin.mff.cuni.cz/~krajicek/ri5svetu.pdf I'm pretty sure we live in Algorithmica. -- -Eric

1 0

p = np (?)
by Eric Purdy 04 Apr '18

04 Apr '18

Some thoughts I had, haven't checked them in too much detail but they feel pretty good to me. I invite the members of this list to complete the proof and write it up to a level sufficient to allow it to be published. You can ask anyone for help as long as you add their names to the paper. Probably the cleanest approach to finishing the proof is to tackle the specific case of 3-SAT. -- -Eric

1 0

paper: "Friendship networks and social status" (Brian Ball, M. E. J. Newman)
by Eric Purdy 17 Jan '13

17 Jan '13

First, some general advice about reading technical papers! Read the introduction, then the discussion/conclusion, then the experimental results, and THEN the technical details (in this case, the section entitled "Inference of rank from network structure"). This is out of order, obviously, but it's a good triage strategy: you read things in increasing level of difficulty, so that you always know whether you care enough to read the next, more difficult part. It's also good because you have more context to interpret the difficult parts if you've read all the easier parts of the paper. I've put a list of wikipedia articles at the bottom of this email, under the heading "Study Guide". They are a hopefully comprehensible explanation of the math needed to understand the most technical parts of the paper. But, if you don't feel like wading through the most technical parts, you don't have to read the study guide articles in order to participate in the discussion. (Also, I'm omitting some background for technical parts that I don't know much about, like ANOVA and statistical tests of model fit. If someone feels like adding anything to the study guide, go for it.) == Paper == PDF link: http://arxiv.org/pdf/1205.6822v1.pdf Arxiv page: http://arxiv.org/abs/1205.6822 == Abstract == In empirical studies of friendship networks participants are typically asked, in interviews or questionnaires, to identify some or all of their close friends, resulting in a directed network in which friendships can, and often do, run in only one direction between a pair of individuals. Here we analyze a large collection of such networks representing friendships among students at US high and junior-high schools and show that the pattern of unreciprocated friendships is far from random. In every network, without exception, we find that there exists a ranking of participants, from low to high, such that almost all unreciprocated friendships consist of a lower-ranked individual claiming friendship with a higher-ranked one. We present a maximum-likelihood method for deducing such rankings from observed network data and conjecture that the rankings produced reflect a measure of social status. We note in particular that reciprocated and unreciprocated friendships obey different statistics, suggesting different formation processes, and that rankings are correlated with other characteristics of the participants that are traditionally associated with status, such as age and overall popularity as measured by total number of friends. == Code == Mike Bishop has some C++ code that (I think?) the authors of the paper are releasing publicly. Mike, can you send that out? It would probably be a good base to start from if we want to run simulations. == Study guide == These are just to be helpful, don't feel like you have to read any of them to participate. And feel free to suggest additions, or say that one of these articles wasn't very helpful. Adjacency matrix: (http://en.wikipedia.org/wiki/Adjacency_matrix) This is useful because it gives you a dictionary between graph theory and linear algebra. It only comes up a little bit, but it's often used without any explanation. Elementary Bayesian stuff: (http://en.wikipedia.org/wiki/Bayesian_statistics) This is a very high-level article that gives links to more in-depth articles. Less elementary Bayesian stuff: (http://en.wikipedia.org/wiki/Bayesian_inference) This is the main piece of background for the technical section of this paper. If you really want to know this stuff, Bishop's "Pattern Recognition and Machine Learning" is the best textbook I know, although it's not accessible without a lot of mathematical background or a lot of free time. Bayesian games: (http://en.wikipedia.org/wiki/Bayesian_game) I've never seen this before, but it seems like the appropriate math to use when thinking about social science stuff with Bayesian methods. EM: (http://en.wikipedia.org/wiki/Expectation%E2%80%93maximization_algorithm) The workhorse of mathematically rigorous Bayesian statistics. It's a flaky piece of shit, but it's pretty much all we have. MCMC: (http://en.wikipedia.org/wiki/Markov_chain_Monte_Carlo) The workhorse of mathematically rigorous frequentist statistics. Gibbs sampling: (http://en.wikipedia.org/wiki/Gibbs_sampling) This is the most mature mathematics that specifies how emergent properties emerge from local interactions. It was first created to model thermodynamics, which is basically all about emergent properties of molecule interactions, like temperature, pressure, etc. Ising model: (http://en.wikipedia.org/wiki/Ising_model) This is the most straightforward example of a Gibbs distribution, it's what you should think about when you try to understand them. On Wed, Oct 3, 2012 at 10:37 AM, Eric Purdy <epurdy(a)uchicago.edu> wrote: > I am starting a new listhost, math(a)moomers.org. I am hoping that it > will function much like readings, but that we can have long and > intricate technical discussions without annoying anyone. > > Some proposed ground rules: > > - There are no stupid questions, and we're interested in everyone's > opinion. In particular, you are encouraged to subscribe to > math(a)moomers.org even if you do not have much mathematical training. > If you're interested in watching math being done, that's all that > matters. > > - There will be no status games, and no arguments from authority. > Nobody should sign up to math(a)moomers.org because they want to impress > anyone, and nobody can cite their professional status or mathematical > training to make someone else shut up. > > - If any publishable research is done on math(a)moomers.org, we will > decide by consensus whose names should be on any papers that result. > > The first proposed paper to read is "Friendship networks and social > status", by Ball and Newman. It is available online at > http://arxiv.org/abs/1205.6822 . I am interested in it because it > gives a nice model of social stratification as an emergent property of > social networks. I think it would be cool to try to extend it, and I > think it should be possible to give a scientific account of the > phenomenon of discursive privilege. > > Here is the abstract: > > In empirical studies of friendship networks participants are typically > asked, in interviews or questionnaires, to identify some or all of > their close friends, resulting in a directed network in which > friendships can, and often do, run in only one direction between a > pair of individuals. Here we analyze a large collection of such > networks representing friendships among students at US high and > junior-high schools and show that the pattern of unreciprocated > friendships is far from random. In every network, without exception, > we find that there exists a ranking of participants, from low to high, > such that almost all unreciprocated friendships consist of a > lower-ranked individual claiming friendship with a higher-ranked one. > We present a maximum-likelihood method for deducing such rankings from > observed network data and conjecture that the rankings produced > reflect a measure of social status. We note in particular that > reciprocated and unreciprocated friendships obey different statistics, > suggesting different formation processes, and that rankings are > correlated with other characteristics of the participants that are > traditionally associated with status, such as age and overall > popularity as measured by total number of friends. > > -- > -Eric -- -Eric

4 30