juiceandjuice

Did anyone even bother checking the author's numbers or methods?

My biggest problem with this series of articles is that it's largely based on the "No commitment" answer to a questionnaire given to postdoctoral students. His "Nothing" section is actually "No definite commitment for employment or postdoctoral study" ( here http://www.nsf.gov/statistics/sed/2011/pdf/tab42.pdf )

which:

"Includes respondents who indicated "other" in all years and respondents who indicated "do not plan to work or study" in 2006 and 2011. Also includes 715 respondents for 2011 who indicated definite plans for "other full-time degree program."

That 715 is 1.5% total, or 25% of the 6% difference between 2006 and 2011.

I spent nearly all my time between 2006 and 2009 around grad students (mostly physics, but bio and engineering too). There was a fundamental shift in thinking in 2008, both for grad students and professors. Professors could keep grad students working after their PhD was conferred, in a position somewhere described as between a postdoc and student for a year or two, a longer holding pattern - because people were having a hard time finding jobs. This goes double for international students.

According to this ( https://chronicle.com/article/From-Graduate-School-to/131795... ), 23000 more doctoral students are receiving some sort of financial aid from the government than in 2007. Even if those numbers are increasing, the total percentage of people with a masters or higher is only 1.5% of the apparent population.

According to BLS, there's ~3.2 million people with a PhD. That means there was an increase of people, from ~.3% to ~1% of total people applying for aid. (I'm probably off by a little bit because I'm using numbers from different years and estimating, but not by a factor of 2)

So, we had an increase of 20,000 people applying for some sort of aid between 2007 and 2011 with a PhD. Cumulatively, if we were assuming that increase was purely from new grads and accurately portrayed the unemployment of a new grad, that would mean an increase of 5k year over year, or 11% of new graduates. But to say that was all new graduates would be doing quite a disservice: There has been major cuts in research at national labs, universities, and all sorts of other publicly funded institutions.

I'll admit, finding the right job is probably harder and takes longer today than in 2006/2007. I agree that we have much more foreign students who have a harder time figuring out what they will do after graduation, because of problems with visas and money. But at 2.5% unemployment and a median salary 50% higher than a bachelors degree, ( http://www.bls.gov/emp/ep_chart_001.htm ), I think it's hard to argue there's PhD bust.

If I had more time I'd delve into this deeper, but I got to get back to science...

This is the same concept.

AKA RPython for Javascript

I grew up next to Shiprock, the actual city, in New Mexico, in addition to other places around New Mexico.

The interesting thing is that the "state" of Shiprock would likely be the poorest and most nuclear capable state on this map, containing both Kirtland AFB and Nellis AFB in addition to Los Alamos National Labs and Sandia National labs.

They actually use mmap internally.

Hi, do you actually do the bookkeeping for the processing inside graph? How do you do this?

I work on a stream processing/workflow engine used by a few large physics experiments. It's declarative in nature too, but we use XML and let users write the glue. We also have the notion of persistent files and variables, although we don't compile and verify dependencies quite so much.

The male/female ratio of sexting is interesting.

A map is a hash table. I'm confused as to why you don't use either a custom hashing function or your own custom datatype that organizes your data according to the sequence order of your key. While a generic hashing function that can operate on any type of data could be implemented in the language, I believe the flexibility of defining the hashing function yourself is perfectly acceptable considering the performance implications of having a universal hash function. The reason why is this: The table lookup should fundamentally be roughly an O(1) operation after hashing your key value. A universal hashing function that operates on arbitrary-length immutable sequences would likely be an additional O(n) (where n is the length of the sequence) complexity built into the map structure. I don't believe that's acceptable from a language standpoint.

If you've ever traveled across the country, you'd realize people in Seattle are much different than people in Phoenix who are much different than people in St Louis who are much different than people in Boston who are much different than people in Atlanta who are much different than people in Honolulu. You could easily the population of Kansas is as different from the population of California as Colombia is different than the population of Argentina, so I don't think "foreign-born" is a good metric for heterogeneity in the context of this question.

I'd argue your brothers approach is just as valid as yours for either industry. You both hire for "cultural fit", just your ideas of what "culture" is differs.

So basically facebook is what facebook was 7 years ago when you could click on a word in a profile and find out everyone else in your (college) network that also had that listed.

I'm glad they brought back functionality they removed forever ago.

No, but it's reasonable to assume those details, while secret, were plausible in a system that has time and again proved itself irrational and unfair.

You are asking people to not jump to conclusions about people who jumped to conclusions.

Objectified is okay. The Eames documentary is much more interesting in terms of the design process, as is the Milton Glaser one.

Disclaimer - I'm out of my element as far as the science goes, but privy to relevant info.

Here are a few relevant publications from Fermi:

https://www.sciencemag.org/content/323/5922/1688.full

http://www.nature.com/nature/journal/v462/n7271/pdf/nature08...

There's a difference in method between the Fermi and Nemiroff methods. The Nemiroff method is novel, but with lower statistics. Nemiroff himself has stated the results are not yet statistically significant. More short, bright GRBs need to occur.

There seems to be a power substation here: 39.580234, 76.058357

And one large tower connecting it to a grid extending south here: 39.579052, 76.060278

The buildings/structures out there are in interesting ratios. One of the U buildings is ~64m on 3 sides. Another one is 125% bigger on the smaller size, and then in a ratio of 1:2:1 per side.

A couple more buildings are about 108m, then there's more at about 86. Again, the bigger one is 125% larger. The big building is about 144m on the left and 154m on the right. To the right of that, there's another complex that's about 133m to the right of that (125% of 108m) Maybe it's on the inside of the build black building?

Or, you know, you could try not reinventing the wheel and use emacs.

Sort of off topic, but there are physical examples of "flat" design applied to objects. One notable example could be the Quad 33 preamp:

http://en.wikipedia.org/wiki/File:Quad_33_Front.jpg

I like how they used twitter to complain about a social media application having ads.

The first real library I had to learn to use was GSL. The first framework I ever learned to use was ROOT. I also had to learn how to use FFTW and all sorts of other math libraries. I've had to hack fortran a bit, but mostly just know how to read it and compile it with g77 and stuff because that's what my professors would program in. In my only real course in programming at a university I learned how to link C to fortran libraries. I used Maple way more than matlab, and all of this was the majority of my programming training before I graduated from college.

If there is a divide between fortran and LISP hackers, I'd have to say that Python is going to be the adopted parent/mentor, and it's the only programming language truly situated to bring them together. This is why projects like PyPy get me excited. Other projects like Julia are really interesting as well.

I have to agree with the premise of the article though. There is a divide between mathematical hackers and non-mathematical hackers. I don't think it's anti-intellectualism per se, it's possibly more of a result of the increasing stratification of backgrounds in an industry which has increasingly focused on products, services, and most importantly consumers and "social".

I went to a talk yesterday by an veteran applied mathematician who does research at HP labs. He mentioned something that really resonated with this notion: The days of academic style research, of Bell Labs and the like, has disappeared. This isn't to say that research and math has disappeared from the industries, but as a researcher you have to justify the research by explaining how it can contribute to a bottom line for the company. There's a few companies with looser restrictions on this (Microsoft, Google), but typically this kind of research is best done at companies that are monopolies with high profit margins. In that case, a company like Apple would be best situated to conduct research on the level of the Bell Labs of the past, but it's evident that they don't. The mathematical hacker started falling out of vogue sometime in the 70s. If the mathematical hacker ever becomes vogue again, it won't be because of Apple. It will be because of big data, autonomous cars, medicine and bioinformatics.

Except that somewhere isn't implicitly in the vicinity of the new development. That somewhere could be in the suburbs, another city, a dorm room, a room in a parent's house, or any other options which doesn't have any real impact in the availability of housing in an area.

The categories are a little confusing. If you go into scientific research, say at a publicly funded institution, where does that put you? College Education? Technology? Engineering? Government? or just Other?

Hah, yes. I don't know if the comment before yours was intended to be advice to me, but I'm one of those python astrophysics people (professionally). I deal much more on the data and distributed systems side much more than the analysis side, or really... I write software and frameworks that the analysis guys (from several very large experiments) use.

You can't just say something is an antiquated institution, and reference a wikipedia article just because a very similar system, successful or not, has existed for a very long time. By that logic:

Democracy is an antiquated institution [1] that makes about as much sense in today's context as ...

[1] http://en.wikipedia.org/wiki/Athenian_democracy

I'm not a huge fan of patents either, but let's try to pretend to be a bit more objective and sane about the arguments we are proposing and use some relevant references when making a claim like that.

As someone who has taken engineering courses and holds a BS in applied physics and a minor in math, while working as a developer (for physics experiments), I'd have to say that physics, more than any other major, teaches you how to learn. In reality, the main reason a BS in Physics even exists is to teach you how to learn so you're prepared when you get to graduate school to work on new problems people haven't had to face before.

The unfortunate thing is that corporate american tends to highly undervalue the BS in Physics. While I graduated at the worst time possible time (2008/2009), it took me over a year and a half to find a "real" job. No, I didn't have two + years experience in developing with .Net. I did have 4 years of experience in everything from LabView to Verilog, photomultiplier tube calibration and particle detector design to advanced electrodynamics to CADD and machining, and all the calculus and linear algebra necessary to get there.

As someone who was perpetually frustrated by a lack of consideration when applying to jobs, I'd just like to recommend that anyone reading this comment to always consider a candidate with a background in the hard sciences, even if they don't have the relevant experience.

Yeah, I don't mind doing QFP and SMD resistors (actually QFP is pretty quick when you get used to it), but individual resistors take forever.

Also, I forgot to mention most processors are BGA packages AFAIK, so impossible to actually solder manually.

Actually it's likely that most(all) of the resistors and LEDs are SMDs soldered using reflow or wave soldering.

Your electronics from before the 80s (and often the 90s) was soldered by a human though.

I've looked into that a bit, and it wouldn't work precisely because the transactional nature and reprocessing of the tree we'd need to do in order to make it worthwhile. We do get an idea of the tree structure because each stream correlates directly to a task, but we have no idea what the fanout will be until everything is done executing, because a jython scriptlet could create 1 stream or 1000, and even that can be parameterized and changed on subsequent executions.

I think what would work best in our case is, in our stream table, an extra column for what I'd call "ancestor", or the top most stream and execution, then using that (with an additional column denoting an execution "isLatest") to quickly grab a subset of likely-nodes in an execution branch, and then process them to construct the tree of actual nodes. It's all kind of complicated because a third level stream could be the latest execution from a second level stream that's not the latest execution. A full closure table would be better and quicker, but that would require an extra table and a whole lotta entries which would just be bigger headache when it comes to locking a node for a transaction, because now you have lock one row in the main table, and a lot more rows in the closure table for an update.

I didn't design the system. It makes sense when it comes to science processing, especially when you have things where people want to reprocess data processed a few years ago with some new calibration or something. I know how I'd redesign it if I was starting from scratch though, but even then people often think of weird ways to use a system this flexible that can be a pain to work around.

I develop software almost similar to this, we even call it Pipeline. It's mainly a dispatcher/supervisor for computing grids and heterogeneous batch systems.

For us, a task is a predefined set of processes. An instance of a task is a stream, an instance of a process is a process instance. Typically, a stream will have a few process instances that are typically batch jobs sent to the batch farm. We also allow jython scriptlets as process instances to do house keeping, namely to instantiate multiple sub streams and define variables and the environment that might be required for a job. The root level task can have a version and a revision number.

For instance, a top level stream will have a process instance that instantiates 1000 substreams, which have batch jobs that can be sent out to the batch farm. Those complete, and the substream can send do more or exit based on the status and preconditions. The top level stream will end in success, failure, termination, or cancelation depending on conditions. Any stream or process instance can be rolled back, but only when the parent streams are all in a final state because typically the preconditions (onSuccess, onFailure, onDone, etc...) applies to all streams and process instances to avoid race conditions and data corruption.

We use Oracle to do all our bookkeeping. Jython scriptlets are ran on the main server (typically only used to control flow). We use email for batch system messaging. Typically our jobs are very long jobs and take multiple hours complete, especially monte carlo jobs, and because of this, email has scaled fine, because the amount of emails rarely exceeds 100/second, even if we are concurrently running 2000+ batch jobs. I'd like to move to 0mq for the slight benefit in performance, but really email has been fine.

Most messages are logged directly to Oracle. Batch instances will typically have a log file, but that usually ends up on an nfs server or somewhere a user defines. Previous executions are archived. A user will define a log root, and the log file would typically be in the form of something like:

$LOGROOT/TaskName/TaskVersion.TaskRevision/StreamID/StreamID/ProcessInstanceID/log.txt $LOGROOT/TaskName/TaskVersion.TaskRevision/StreamID/StreamID/ProcessInstanceID/archive/executionID/log.txt $LOGROOT/TaskName/TaskVersion.TaskRevision/StreamID/StreamID/archive/executionID/ProcessInstanceID/executionID/log.txt

This can sometimes lead to directories with a very large amount of files,

We abstract out the batch farm and have custom daemons so we can connect dispatch jobs to various batch farms around the world.

So far, it's scaled pretty good. We have over 30 million streams and over 60 million process instances logged in our database. Some of the tasks are pretty complex tree structures do to the parent-child relationship and nesting we allow. This is one reason we've stuck with oracle, due to the fact that it has a CONNECT BY statement. In practice, that's been okay, and would continue to be okay if we had a good partition plan. We have plans to move away from Oracle and implement a database agnostic solution based on temporary tables. Due to legacy requirements and uptime necessity (this software has been developed for the last 5 years), it's been hard to implement these fixes in production. We plan to overhaul some of this when we get a new database server.

Storage is typically up to users writing a task, who often use nfs or, more typically, xrootd (http://xrootd.slac.stanford.edu/) depending on requirements. Many will register the data with other software I develop, called datacatalog, which is mainly for record keeping and metadata maintenance for the data. Access for this is typically done through jython scriptlets.

A phase/state is usually tied to structure and/or energy of matter. Dark matter is the blanket term we give to particles that interact (observably) through gravitational forces.

In general, cosmology is concerned with the quantification and characteristics of dark matter through observation, so a cosmologist is likely to mean the congregation of dark matter in the universe. Particle physicists are generally concerned with direct observation and characteristics through experiment.

I don't know who messed with the wikipedia article on state of matter, but they're wrong.