Saturday, January 30, 2016

January 30, 2016 at 10:59PM

Today I Learned: 1) One of the developers of Titanfall spent his two-week vacation porting the game to Oculus rift, and the dev team tried it out. The conclusion? Titanfall was a terrible game for Oculus rift. Why? In Titanfall, you spent a lot of time moving very quickly and flying through the air, which is *terrifying* in Oculus rift. 2) Aphaenogaster cockerelli (the species of my new ants!) apparently like insects and seeds, but not sweet stuff like honey, fruit, or sugar water. 3) A few facts about genome sequencing: market price for a full-exome sequencing run, with analysis and interpretation, for a medical institution, is around $4,000; the most common way to get DNA for sequencing (again, in a medical context) is to extract it from blood. This seems a little crazy to me, since the vast majority of cells in blood have no DNA in them...; when DNA is purified and shipped before sequencing, rather than shipped in blood form, it is much more difficult to predict copy number variations, for reasons that are not entirely understood. Thanks Jeanne Morin-Leisk!

Friday, January 29, 2016

January 30, 2016 at 02:34AM

Today I Learned: 1) Don't empty bottles of LB-agar into a sink. You'll end up with lots of annoying chunks that are hard to clean up. Just dump it out in the trash can to begin with. It'll smell really funky, but it's not actually that messy and it ought to be totally sterile. 2) The Plank length is the fundamental unit of length in nature, and is thought to be effectively the smallest size a thing can be. ...or that's what I thought. Apparently that's a common misconception. Today I learned that there is no such thing, and it makes no real sense to talk about a "smallest size of thing" (the reason is relativity -- if you made something of the "smallest possible size", there would always be a reference frame you could come up with where that thing was smaller due to length contraction). The Plank length IS a thing, just not THAT thing. It's roughly the distance at which you have to stop ignoring quantum gravity, which is a concept I don't really understand. According to Wikipedia, which I may or may not trust on this matter, it's also roughly the smallest *measurable* distance. That's not an experimental limitation -- it's just that below that scale, traditional notions of distance aren't very useful because stuff is smeared out too much. Also according to Wiki (and I trust this fact more), a 0.1 millimeter dot (about the smallest speck you can see) is as much bigger than the Plank length as the universe is bigger than a 0.1 millimeter dot. That's a really small unit of measurement! The other cool thing about the Plank length is that it's defined entirely in terms of fundamental physical constants, so it doesn't depend on your choice of units (just your choice of universe). 3) 2 AA batteries (3 volts) is sufficient to split water with a graphite cathode and a graphite electrode. 1 AA battery (1.5 volts) is not. Slightly relatedly, volt meters apparently tend to use a special 12-volt battery that looks like a AA battery squashed down to the length of a AAA battery.

January 29, 2016 at 05:09AM

Today I Learned: 1) There is a website called "willyoupressthebutton.com" which gives you a hypothetical button to press, or not press. If you press the hypothetical button, you get some really good thing, stated by the site, but it is accompanied by something really bad, also listed. For example, one button I ran across would let you visit any place in the world for a week, but the whole time you would be there, a large spider would be nesting in your hair. Do you press the button? Another one: You will become immortal and live forever, but everyone will hate you. Do you press the button? Warning -- the site is lightweight but add-heavy. If you find this sort of site annoying, do not go there. Warning -- this site has an extremely fast action loop and high addiction potential. If you do not have time to spare, do not go to there. 2) The Hanon exercises, a staple of piano education, were written in the mid 19th century by, well, Hanon, who was FRENCH. For more than a decade, I have assumed that Hanon was Italian, or perhaps Austrian or German. Not so, not so. 3) One of the biggest rate-limiting steps on game developers deploying patches is vendor quality assurance, or QA. Any developer has to put new code through rigorous tests before deploying it, or it is pretty much guaranteed to have bugs, which in the case of games, can be, well, game-breaking. Testing takes time and resources, which means it can take a while for patches to get from developers' computers to players' machines. Today I learned that even after a patch has been tested in-house, there is an additional layer of testing performed by the QA departments of the *platform* on which the software was released (i.e., Microsoft for an Xbox game, Apple for an iTunes app, etc). The vendor QA team makes sure that the new patch isn't malicious, doesn't break anything obviously, and otherwise meets the quality standards of the platform. This additional QA testing takes between 7 and 14 days for a patch for a typical AAA console video game, which can be a big chunk of the time required to churn out patches. Having a layer of vendor QA also incentivizes developers to spend longer testing and verifying their patch, because if anything goes wrong, it will remain live and unfixed for a MINIMUM of a week before any counterpatch can possibly be deployed. This and lots of other interesting tidbits on engineering the guts of video games from this tech talk at Rackspace by Jon Shiring, one of the engineers from Respawn Entertainment: https://www.youtube.com/watch?v=ayF8e8q_aA8

Thursday, January 28, 2016

January 28, 2016 at 04:00AM

Today I Learned: 1) Polymerases have EXTREMELY high affinity for DNA. As a consequence, the vast majority of polymerases in any given cell are bound to DNA at any one time. This applies both to DNA polymerases, which replicate your genome, and RNA polymerases, which transcribe your genes into RNA. 2) High molecular weight DNA (i.e., genomes) is quite viscous. Check out any youtube video of at-home DNA extraction to see for yourself... ...or better yet, try out a DNA extraction yourself! You basically just need some fruit, high-percentage ethanol (>=70%), salt, and detergent. Try it out here: http://ift.tt/1QuhdLv. Note that this procedure isn't sensitive to much except, to some degree, the makeup of the lysis buffer. If you don't have 50 mL conicals at home (and who does?), just use a glass. If you don't have a coffee filter, use a paper towel or colander or anything else that will strain out the big chunks. 3) ...what the (valid English) sentence "Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo" means. It uses three versions of the word "buffalo": the animal otherwise known as a bison; the city in New York; and the archaic use meaning roughly "to bully" (or possibly to bully roughly?). The 7x buffalo sentence translates to, it's "New York bison that New York bison bully, bully New York bison." Thanks to Robert Johnson for explaining this one to me, and also for pointing out that you can take it to 11x buffalo quite easily -- "New York bison that New York bison bully, bully New York bison that New York bison bully".

Tuesday, January 26, 2016

January 27, 2016 at 01:00AM

Today I Learned: 1) Neil DeGrasse Tyson has a nephew, Stephen Tyson, who raps. Look him up and you'll run across a rather amusing, if slightly alarming, recent story. 2) The E. coli-based cell-free extract system we use in lab, TX-TL, is MUCH better at producing GFP than similar commercial kits. TX-TL is essentially cell innards minus the membrane, while commercial kits are built from the ground up with known components. Also, TX-TL is an order of magnitude or two cheaper than commercial systems, which is nice. 3) "Tort reform" is any change to the legal system that limits the ability of people to suit. In the US, tort reform almost always refers to the practice of capping payouts for medical malpractice payouts. The goal of tort reform (in the context of health care) is to reduce the amount of suits brought against doctors for malpractice and to reduce the payouts associated with those suits. This, in turn, should lower medical malpractice insurance, lowering the cost of healthcare. Does it work? Almost unambiguously no. Malpractice insurance only accounts for a couple of percent of US health care spending anyway, so even a massive reduction in malpractice insurance costs wouldn't do much to bring down overall costs. There's also no evidence that tort reform has lowered healthcare costs *at all* in the states where it has been implemented. Florida is an interesting case study -- when tort reform was enacted in Florida, the primary effect was that insurance companies didn't have to pay out as much money on lost litigation, and they didn't lower their prices, so their profits went up by a factor of 40 (not a typo).

Monday, January 25, 2016

January 26, 2016 at 02:24AM

Today I Learned: 1) ...how to reserve meeting rooms at Caltech. 2) Marvin Minsky, one of the founders of AI research, was still alive until yesterday! ... He just died from a cerebral hemorrhage. I also learned, from Minsky's wikipedia, of a nice little hacker koan from the Jargon File featuring Minsky, which I will reproduce here. (Incidentally, if you haven't checked out the Jargon File, you should) In the days when Sussman was a novice, Minsky once came to him as he sat hacking at the PDP-6. "What are you doing?" asked Minsky. "I am training a randomly wired neural net to play Tic-tac-toe," Sussman replied. "Why is the net wired randomly?" asked Minsky. "I do not want it to have any preconceptions of how to play," Sussman said. Minsky then shut his eyes. "Why do you close your eyes?" Sussman asked his teacher. "So that the room will be empty." At that moment, Sussman was enlightened. 3) There are a lot of ways in which people believe they can tell the sex of an unborn baby. They are all wrong except for a) ultrasound or other imaging techniques that can actually look, c) genetic tests, or d) only using embryos pre-screened for one sex. There is also a very slight but statistically significant correlation between extreme, debilitating morning sickness and having a daughter rather than a son. Many, many other techniques have been systematically tested and shown to be no more accurate than chance, including measuring the belly of the mother, measuring heartate of the baby, intensity of morning sickness (again, only slightly and only for the strongest morning sickness!), and a weird test involving Drano.

January 25, 2016 at 03:50AM

Today I Learned: 1) ...how to steam-sterilize soil! I wouldn't use this as a substitute for autoclaving for truly sensitive tasks, but apparently steaming (preferred) or baking soil to about 180 F sterilizes it pretty well. It's also supposedly important to not heat it much over 180 F, or it will "create toxins". 2) I learned a few things about my handwriting today. Erik Jue and I compared how we write our letters today, and we discovered that I've heavily optimized my handwriting for efficiency of movement, with a couple of exceptions. I also have a strong tendancy to start letters from the bottom and pushing up rather than from the top pulling down. These two observations probably explain a lot of why my handwriting is as unreadable as it is. The two of us each came up with a list of stroke orders to change. For instance, I have an egregiously wasteful 'r', and my 'k' is a little less efficient and innaccurate than the more standard protocol. I've also decided to re-work my 'g's for better control, at the expense of some speed. We also discovered a slightly better way to write 'j' than either of us were using -- we both start at the top of the main part of the letter, stroke down, then go back up to dot the j. It seems just better to dot first, then add the main part underneath, saving yourself potentially a full character-height of travel time. Thanks, obviously, to Erik Jue. 3) I've looked up what tensors are many, many times, and never quite figured out what they are. There's something a bit sideways about how they're defined that somehow makes them slippery to my mind. For instance, almost every source on tensors begins by stating that they are generalizations of scalars, vectors, and matrices. This is only slightly helpful to me. There's more than one way one might imagine generalizing a matrix, for instance. I think I understand at least in part what a tensor is now. In programming terms, a tensor of rank N is a function that takes N index arguments and returns a number. For example, a matrix is a rank 2 tensor, which takes two indexes (a row number and a column number) and returns a number (the value of that row and column). A scalar (otherwise generally known as a number) is a rank 0 tensor, which takes zero indexes and returns a number. Now that I write it down, it seems so simple, but it took me a long time to figure out that function signature. Now, there's a bit more to tensors, as they also have to satisfy certain properties about invariance under changes of coordinate bases, which leads somehow to a more intrinsic definition, but I think this is a pretty good first-order approximation. Thanks also to Erik Jue!