working on a meta-analysis of creatine & intelligence.
It's mostly worthless so far, because while creatine has been used by a fair number of psychologists, apparently psychologists have sworn an solemn oath "confusion to all gwerns!" - no two of these psychologists can agree on: what tests to run (30+ unique measures out of like 10 experiments!); how many grams to administer (total or per day); what kind of creatine to use (monohydrate vs CEE); whether to test the young or elderly, the omnivores vs vegetarian; or whether to publish the research as a paper, a patent, or not at all. (I know of at least one experiment with a null result which did not get published.) Oy vey.
So now I'm waiting on the unpublished experiments, waiting for a reply from the lead author on the German patent, and hoping Hammett will tell me more about his in-progress creatine experiments.
I have finished my transcription of Radiance.
Now I am working on annotating all the allusions, and waiting for a copy of New Legends with the original novella "Radiance" to arrive so I can read it and see if I need to transcribe that.
I have been doing a little experiment with Hacker News where I submit one gwern.net page a day. It's gone quite well so far (I believe each one has reached the main page); in reverse order:
I'm reading through the MIRI course list. I've been reviewing the books as I go. The primary goal is to become capable of working on open problems in FAI. A secondary goal is to provide useful data, both to other readers and to the maintainers of the course list.
With respect to the secondary goal, you are encouraged to chime in on relevant threads with personal reactions to the same books or suggestions for alternative books. On my own, I can only speak for the experience of a single person.
I have often found that teaching a subject improves both comprehension and retention. If you are reading any of the books that I have reviewed and you are having trouble understanding a concept (or if you would like to check your answers to exercises, or if you just want to chat), please PM me. I am by no means an expert on these subjects, but I appreciate any opportunity to articulate what I have read.
A number of people have noted that the course list does not contain the books most relevant to MIRI's current research. I will in fact be taking a number of detours. I've been collecting suggestions from people close to the research, further suggestions are welcome.
You can find the first four book reviews here:
I'm reading through the MIRI course list. I've been reviewing the books as I go. The primary goal is to become capable of working on open problems in FAI.
Why do you believe this particular book list is relevant to that goal (other than the list itself suggesting so)? It seems like a set of initial directions that might be useful for someone who knows little on these topics, and otherwise too sparse/arbitrary to be helpful "out of the box", without significant modification. Based on your reviews, you are not in that category.
(Edited to address your clarifications)
I assume that the book list is somewhat relevant, due to the fact that MIRI hosts FAI research groups and cultivates the book list with the stated purpose of outlining relevant subjects. That said, I am not treating the list as sacred.
For one thing, I'm jumping around to the subjects that seem relevant to the current research. (Category Theory and Set Theory were warm-ups for Model Theory, with which I am not familiar.)
For another, as I said in the OP
I've been collecting suggestions from people close to the research, further suggestions are welcome.
Luke has already pointed me towards some Provability Theory texts that I plan to read through after finishing Model Theory. The MIRI course list seemed like a good starting place, but I have no qualms about deviating from it.
1) I have translated 100 articles from LessWrong to Slovak language. The number is cool, but I am afraid I am already too deep in the diminishing returns territory; possibly the translation becoming a lost purpose. I get almost no feedback, except that when I once in a while put a list of new translations on Facebook, some of my friends "like" them. But Facebook "likes" are cheap and don't necessarily mean anything beyond "you have amused me for a millisecond".
I guess I should ask explicitly about feedback, possibly announcing that if I don't get any, I will stop making more translations.
2) I made a flyer about LessWrong: 1 paper A4 folded to three parts. The inner sides introduce the topics of Intelligence Explosion, Rationality, and Effective Altruism (with hyperlinks to Slovak translation of Facing the Singularity, and English pages of CFAR, GiveWell, and 80000 Hours). On the outside the front page is a short introduction about LessWrong, the back page is list of recommended books (Thinking Fast and Slow, Good and Real) and websites (LessWrong, Overcoming Bias, HP MoR), and the remaining part will be customized for the intended audience.
Then I made a version for new members of Mensa, because the Mensa testing was a week before our LW meetup. The customized part was something about how intelligence should be joined with rationality to achieve best results, and an invitation to the local LW meetup. I gave the flyer to 25 people who participated in the Mensa tests (right after the test, when the results were not known yet), and the result was... zero.
I guess I will try it one or two more times (the flyer is already made, the only work is going to the tests and giving it away), and if the result is still zero, I will give up my hopes of Mensa members as potential rationalists. I should also try the flyer with some other target group, specifically students of informatics and students of psychology (each group will have its own customized part of the flyer).
3) I started again doing my Udacity lessons. Not sure about the value of doing so; it's more like I am sometimes too tired in the evening to do anything more complicated than this, so I would prefer to have a habit of doing a few Udacity lessons than just browsing random web pages.
EDIT: I completed my first Udacity course. This serves mostly as a proof that I can finish the course I started. The information in the course was less important, because it was Computer Science 101 and I already knew programming; but I had to write dozen exercises in Python which I never used before.
I get almost no feedback, except that when I once in a while put a list of new translations on Facebook, some of my friends "like" them
You don't have any analytics set up to tell you how much traffic you're getting?
Hah, I completely forgot about that! Thanks!
So... when I published 6 new translated articles on September 25th and published a list of links on my Facebook page, google analytics shows that the typical noise of cca 30 "visits" on my website per day increased to 56, 94, 84 during the next three days, and then dropped to the usual noise level again. The values for "unique visitors" are almost the same, and the values for "pageviews" are cca 30% higher.
I am not sure I interpret this correctly, but seems to me that this suggests over 140 people, most of them reading only one of the articles. Assuming it's not the same person each three days, which would make the worst case of 64 people, reading two and half articles on average. So, let's just round it to 100 people.
Actually, that doesn't seem so bad...
(Note: My complaint about "no feedback" meant that when I started publishing the first translations, some people spontaneously sent me messages on facebook, asked questions, and two of them came to the local meetup. That was when I started doing this. During the following half of the year, there was no other activity of this kind. I kind of hoped that doing this would translate to at least one new person per month coming to a meetup.)
One year and one month ago I decided to post a technical article / blog post once per week. Still going strong. Some of them are weak, but mostly I'm happy with the results. Apparently I enjoy writing about technical stuff.
For example, recently I told someone about quantum pseudo-telepathy, and promptly realized I didn't understand the details when I tried to explain it. I decided to spend the time to figure it out, down to at least the level of a quantum logic circuit. The Wikipedia article was zero help, just confusing me even more. So I started reading sources until I understood it well enough to write about how to simulate it. It takes me up to ten seconds per written word (counting every revision) to write a post so it took quite a lot of time to get those ~4.5k words down the way I wanted.
Another post I enjoyed writing, and more in tune with Less Wrong's subject matter, was Solomonoff's Mad Scientist. It just talks about a hypothetical machine that picks actions that maximize the prediction entropy of a Solomonoff Inductor (it is definitely not friendly).
In the near future I'm hoping to come up with a nice simple game to demonstrate making something "un-map-hackable" via secure multi party computation. It's simplest where there's three players, and I've already got the math and the code for the mechanism worked out for that case, but I haven't come up with something trivial enough to demonstrate it.
Having an end goal of explaining what I did has pushed me to finish more projects, instead of just starting them.
I left my family, my job, and my girlfriend to move across the country, to Austin, to explore the world, to work for myself, and to become a Complete Social Creature, aka a Normal Person, aka an Adult.
Progress. Being so far from home has freed me to act like who I want to be, rather than who people expect me to be. This is coming at the same time that social interactions are making much more sense than they did when I was younger. I understand what I have to offer and what others have to offer me, and how to frame our interactions like that. Dating is easier and less stressful, which is amazing considering how inscrutable I found it even a year ago. I've been working full-time on the side business, and we filled over 650 orders in September.
Setbacks. The day I moved down here, an unexpected expense wiped out a good chunk of my savings, at the same time my buddy quit the stake (poker), and the consulting job I thought I'd lined up never happened. So I'm playing with less of a safety net than I'm accustomed to, and I have to live on minimal income which is a stressor but a minor one. I haven't been here long enough to build the deeper bonds of friendship so I need to tread carefully. But overall it's an adventure and a great challenge and I'm very excited to keep moving forward.
Working on getting the OpenOffice.org article on Wikipedia to featured status. It's a long and tedious road of polishing. A peer review helped. I nominated it for good article status, but frankly nobody cares about GA. I started getting it into shape over the Christmas 2012 holidays (in between sipping Christmas pudding mead) and have since completely reorganised it and read every suitable tech press reference I can find in several languages (here's to Google Translate and the tech press everywhere apparently writing entirely in cliches - I'd certainly never chance machine-translated Japanese in a subject area I didn't know the cliches in). Your opinions as fresh readers would be most helpful :-)
'Cos I FELT LIKE IT. It's also pretty historically important. Documents are boring, but our large and complicated civilisation runs on them. With OpenDocument (being both an open format and machine-processable), OOo changed the game.
edit: The article also used to cover Apache OpenOffice, but that's now a separate article (it got separated out just after AOO 4.0).
Probably very late to the thread, oh well. Some weeks ago I started vividly imagining a certain style of 3D pixel art, and the imagination was so insistent that I wrote an editor program for that kind of art using JavaScript and WebGL (it will be on GitHub once I clean it up), and before even finishing the editor I started drawing stuff with it and couldn't stop. Today I set up a tumblr for the things I draw: eachothersbreath.tumblr.com.
In the home stretch of transcribing every essay from every issue of Ideological Commentary as a major addition to the George Walford International Essay Prize. This will aid students applying for the Prize.
Cataloging my collection of papers by and about R. Buckminster Fuller. This is going to readily double my bibliography of Fuller (which is already the most extensive bibliography of his work that has been published). I enjoy it and I hope it is useful to others.
Perhaps 1/3rd done preparing a collection of egoist essays by an author who has been out of print for decades and who has never been collected. Oh my, it's good. Currently negotiating with two institutions to get copies of the last two essays I need. I will gladly pay a LW reader in or near Amsterdam who can go get some scans for me - the institutional rates for scans are high, the DIY costs are not. It pleases me to put this book into print.
Still promoting the work of egoist Dora Marsden. I am the first to collect and publish her work, with the hope more will read it. Please do!
Gave away 300+ plans on making model rockets at recent Portland (Oregon) Mini Maker Faire, now making rockets so that when it's rocket season again I'll have a fleet ready. Fun.
Learning to fly an R/C helicopter. Fun.
an author who has been out of print for decades and who has never been collected
Dora Marsden
Are these the same person, or two different people? If they are two different people, who is the first author?
You have noticed i did not write the name of the first of these two authors. There's always Dora Marsden for now, with the other author appearing as soon as possible.
I'm curious: why is it important not to let on who the other one is? (Perhaps you can't answer that either without giving away things you don't want to.)
Learning the Rust programming language. It looks like a surprisingly solid candidate to push into the C++ niche of very high-performance large-scale programming. The big difference to most other candidates is that the developers figured a basic coding idiom that doesn't use garbage collection, while still guaranteeing that you won't have memory leaks, accessing invalid memory or null pointer dereferences crashing your program unless you go out of your way to do unsafe things at designated-unsafe blocks of code, at the cost of a somewhat complicated type system. It's backed by the Mozilla Project, so it also has a bit more resources to go with than average single-programmer language projects.
I'm working on creating a good Anki deck for learning human physiology that I'm going to release publically.
I want to run a study with it to see whether certain options in Anki help users. That means I also need to write a plugin to extract Anki review data in a way that allow the users of the deck to send it back to myself.
I made Sodna, a lightweight virtual terminal library for developing text mode games like NetHack or Dwarf Fortress. It's meant to be an alternative to the currently widely used libtcod, which comes with a large amount of game logic routines for C programmers as well as the virtual terminal implementation. Sodna opens the graphical window, lets you draw colorful letters and read keyboard input, and does nothing else, so it's a good alternative for people who want to roll their own game logic code.
Mostly made this to scratch my own itch, I found libtcod annoyingly bloated for making quick prototype games when using a new programming language, since I'd be making the game logic code in the new language anyway and only wanted the library for quickly bringing up a simple text terminal.
The other thing I'm working on: the loved one quit her horrible job in August 'cos she couldn't stand it any more, and is now MUCH happier. Unfortunately, this means she needs to somehow produce an income as an artist. This is of course difficult. Though the T-shirts are selling a few (US store, UK store). And the gap from 0 sales to 1 sale is much bigger than the gap from 1 sale to 100 sales. She's also writing a children's book with dinosaurs, with the eager assistance of the 6yo.
We've also got about 70 litres of mead brewing at home, and it's coming out shockingly nice. Brewing for sale is of course highly regulated, and if we wanted to start it as a separate business we'd be starting in the realm of £10k starting capital needed; but partnerships with already-licenced brewing premises are feasible, so she's talking to a local indie brewery about doing a nice mead starting in mid-October (when they have a spare vat) for sale and hence money.
So, her job is to be the artist and mine is to be the manager. Not much money yet, but we're not starving (here's to high-paying day jobs in IT) and it's lotsa fun.
I'm working on helping effective altruists make more money and be more successful in their careers. 80,000 hours has done a great job helping people make high-level strategic decisions such as "should I work in Law or Finance?" My focus is on helping people who've chosen a career path be more effective and make more money. For example, I've helped several of my friends negotiate an increase in their salaries and hope to spread these skills amongst the EA community at large.
One example of this has been individual conversations regarding career advice. I've also been talking to people who have thought about this issue, such as Alexei or the folks at CFAR. By working with several people, I'm hoping to get a set of coherent career advice and turn this into a course that EAs could benefit from.
One candidate so far seems to be a course on negotiating your salary and promotions. Other common questions have included:
As part of a larger project to learn javascript, HTML/CSS, and other web dev skills, I've been working through the http://nodeschool.io/ exercises. I'm currently through 11/13 of the first series, "Learn you Node". I've had 2 books sitting around on javascript/node.js, but I've barely made a dent in them before I found this series. There's something about compressing the actual learning into bite-size "challenges" that nerd-snipes my brain into wanting to learn, work through the problem, and then branch out into more. It seems to be mostly unrelated to ease of acquiring the knowledge, as most of the reference material that I've been using to learn for the nodeschool exercises is just the online API documentation, which is well-written but less straightforward than the O'reilly books.
Of course, I was making steady progress on learning Lua, and that's mostly fallen by the wayside, so there are still no free lunches.
I founded a company, Lambda, a talent agency for exceptional developers, designers, technology professionals.
Here's the pitch:
It's currently very difficult for many nontechnical people to find high-quality software developers and designers to do contract work. Because it's often difficult to judge technical ability if you're not a technical person yourself, people often select the lowest bid for the project they want to get done, resulting in poor work or missed deadlines. People want to pay for high-quality talent, they just can't tell the difference.
By screening all candidates ourselves, prospective clients can come to us and get an assurance that their project will get done well. We're tapping into a dark pool of high-quality talent, by signing up entrepreneurs and other skilled folks who aren't on the market otherwise, who wouldn't even consider participating in a marketplace filled with low-quality talent, because they value their time.
More importantly, we provide a service to developers by doing the legwork of finding clients and screening them for quality. There's a world of difference between clients who understand what's involved in a complex software project, and those who don't.
I've been doing small bitcoin transactions by hand using bitcoind raw transactions, and storing the information in leveldb (which is dead simple, and apparently scales well). My near-term goal is to create a colored coin client that tracks inputs/outputs to make sure they have a common origin, so they can be distinguished from regular bitcoins and acquire additional value. I've been using python and getting promising results for interacting with leveldb and bitcoind. The same API can apparently be used with electrum servers, so the client does not need to download the whole blockchain.
My ambition with this is to create a type of coin that generally represents stock in utilitarian ethical causes. Anyone who confused about how they should best donate their money could buy these "Utility coins" and hang onto them, thus driving up the value for other prospective donors and ensuring that those who are less confused will have more to spend.
It's a munchkin idea that might or might not work. For now I'm telling myself the main reason to do it is because it is fun for me and as a way to gather information about what kinds of things can be done in this area (high-value research) rather than because it will necessarily be a good way to fundraise for high-utility causes.
The most likely source would be for me to donate them to the genesis transaction myself. For example, I could use 1.0 BTC that I haven't otherwise encumbered. Or a collective of EAs could donate small parts of it so that no one feels they own a particular stake in it. (Maybe 10.0 BTC would make a better starting point, as there would be a billion possible sub-units (satoshi) that way.)
After that, there would still be the question of how to best distribute them. I could e.g. feed them through a mixing service and thus ensure that they are owned by random (anonymity-seeking) individuals. However it would probably make more sense to donate them somewhere that provides good evidence of utility, instead of probable drug-users.
A place like GiveWell might make sense. However I'm not sure direct donation a nonprofit institution is best for the context, as a priority would be ensuring that they go up in value over time, which would be partly contingent on them being held and prized by people. A nonprofit with an immediate use for the money might just sell it on the market for USD, which would mean that a lot of the value from early-adopter effect goes to speculators (which isn't really what I'm going for).
They could instead be distributed directly to sufficiently tech-savvy homeless people, or perhaps (as with GiveDirectly) to poor villagers in economies where money is scarce. (The client would have to run on a smartphone.) The idea is that people with almost no money would be the most motivated to try and sell them at above previous market rates and/or trade them directly for the high-utility goods and services they need (metal roofs, etc).
So I'm still thinking about the details on that end, as well as whether/how hard to try to make a profit.
It's worth noting that Bitcoin has attained its value primarily by being hard to come by within the context of a game-like setting, i.e. mining where the person with the most effective hardware wins. I could end up doing that with some or all of the coins. If I were to use a proof-of-work distribution system, I would probably use something less parallelizable such as scrypt hashes. (Litecoin uses this, but is otherwise an alternative blockchain of bitcoin -- it is not a colored bitcoin like this would be.)
There are ways to make alternative schemes for proof-of-work function within the existing bitcoin protocol, for example attach some coins to a private key that is the result of a greatly expanded version of a smaller seed, and publish only part of the seed. The missing bits would have to be guessed repeatedly, and the expansion from the original seed would have to take place each time, before someone could claim the coins. Difficulty would be determined by the number of missing bits.
Restart doing sport regularly. I started going to the gym one year ago, with the goal of 3 times per week. I reached an average of about 1 time per week, but with a very uneven distribution - i.e. 1 week 3 times, then two weeks nothing. Since July, I didn't have any regular exercise. At the last less wrong meetup, I set the goal to go to the gym 1/week and do the 7 minute workout twice per week. That was on the 22nd. In the first week after setting the goal, I didn't succeed. This week, I did the 7 minute workout once.
Till the next meetup (27th October), read chapter one of "Good and Real" by Gary L. Drescher. I haven't started yet, but am a fast reader and hope to do it this week-end.
Currently working on a Django app to create directed acyclic graphs, intended to be used as dependency graphs. It should be accessible enough to regular consumers, and I plan to extend it to support to-do lists and curriculum mapping.
I need to work on my JavaScript skills. The back-end structure is easy enough, but organising how the graphs are displayed and such is proving more challenging, as well as trying to make a responsive interface for editing graphs.
My personal area of interest is to work out a framework which could map human 'ideas' to AI goals, so that we may be able to monitor friendliness.
It is a lot more complex than I understand, so I I believe the best long term benefit would be for me to acquire maths skills and knowledge.
So, I have been taking a fairly intensive load of online courses to learn maths and other AI topics, to acquire as much practical knowledge as possible. I don't think it will cover all the gaps in a formal maths eduction, but being a motivated self learner I am interested in seeing how close I can get.
I am calculating the first several terms of a combinatorial function that is useful in the counting of certain elements of a polytope I'm studying. The combinatorial function has three integer parameters, so it forms a tetrahedral array. It's not in OEIS.
I have a recursive means of calculating the function. Next, I'm going to figure out the function as a rational expression in integers i, j, k. Then, I'll post it on Math Stackexchange. Then, I'll submit it to OEIS.
I've worked it out, and now I'm not sure that this function is OEIS-worthy (although it's at least as worthy as Jenny's constant). I will definitely post a question on Math StackExchange, and not answer it (if even necessary) for a month or two, in honor of my namesake.
Here is a link to the question.
Here is a link to a related question that is more fun.
I have committed to a food log with social back up, I am testing the hypothesis that to the first degree of approximation calories out > calories in = weight loss.
I have started to hard code a personal website using team tree-house (style sheets and two pages complete). I figure that the last comparative advantage we have before the machines take over is coding so why not test if I can do it.
I have committed to a food log with social back up, I am testing the hypothesis that to the first degree of approximation calories out > calories in = weight loss.
I have started to hard code a personal website using team tree-house (style sheets and two pages complete). I figure that the last comparative advantage we have before the machines take over is coding so why not test if I can do it.
so un retracting is not possible.
My personal area of interest is to work out a framework which could map human 'ideas' to AI goals, so that we may be able to monitor friendliness.
It is a lot more complex than I understand, so I I believe the best long term benefit would be for me to acquire maths skills and knowledge.
So, I have been taking a fairly intensive load of online courses to learn maths and other AI topics, to acquire as much practical knowledge as possible. I don't think it will cover all the gaps in a formal maths eduction, but being a motivated self learner I am interested in seeing how close I can get.
This is the supposedly-bimonthly-but-we-keep-skipping 'What are you working On?' thread. Previous threads are here. So here's the question:
What are you working on?
Here are some guidelines: