As another undergrad at MIT (Course 2, mechE) I've got to agree with pretty much everything you said. I can really only speak of the people I know but the impression I get is that not many people here put much effort into making themselves a good admissions candidate. They were all more interested in ROM hacking, playing sports, going to science competitions, making Debian packages, building cars, blowing shit up, doing research, and playing video games. High SAT scores will get you looked at. Along with 7,000 other applicants. Don't make it a game, would you want to accept someone whose primary goal is 'getting into a top tier college'?
I mean, my goal is building a giant nuclear powered robot suit with which I can oppress the people of world. Evidently that's good enough for an early admission.
That's a joke, the admissions department never actually found out about my global oppression robot plans.
Seriously, I can't believe the people discounting this. I do my work as a mechanical engineer and find myself in the situation of trying to measure pictures quite often. Just last week for example one of the people I work with took a snapshot of a very wild engine camshaft (almost rectangular lobes) and I wanted to find the lift, slope, and phasing. The shot was taken with severe perspective, looking down the length from one end. It's easiest to see the profile from this angle but it's a huge amount of work to scale the image by hand. $99 would be a great deal for something so useful.
All computers on that use DHCP need registered MACs, if you've got a static IP there's no need at least as far as my experience has been.
IST does a damn fine job, the stakes on having the network running smoothly are quite high and they get it done, but more importantly is the amount of freedom they allow. We've got the most heterogeneous environment I can think of with hundreds of Course 6ers looking for new ways to bend the network and Course 15s finding new ways to try to break it. There's everything from half broken 486s to Playstation 3s running SVN repositories to completely custom embedded devices sitting all over the network (not that they support these devices) running like a well oiled machine.
I couldn't agree more, seriously get yourself a copy of Horowitz and Hill. The language is accessible and examples are excellent with a focus on useful design work rather than theory.
If you'd like to get your hands a little dirty with theory I also highly recommend Agarwal and Lang's Foundations of Analog and Digital Electronics. It's not nearly as fun to read as Horowitz and Hill but everything you need to know is in there and well written.
I completely agree with you, conversion is easy and it's easy to make imperial parts with metric tools and the reverse. The problem is that as long as my Bridgeport moves 0.001" for every tick on the dial I'm not going to design my parts in millimeters unless I've got a good reason. If I want to work efficiently I can't multiply or divide every number I read in the shop by 2.54. When I measure my part after a roughing cut I need to be able to make a quick subtraction with the dimension on the print and then turn my dial that many ticks. This means to make a metric part in a sane manner I need to convert the prints to imperial before anything. If I need to convert all the parts I designed in metric to imperial to actually make them why exactly did I write them in metric? Once I convert the prints I get a mess of dimensions like 0.394" (10mm), not the easiest numbers to work with.
As someone paying a machine shop to do a job why should I design parts that will cause increased confusion between me and person making the parts and a higher likelihood of error? Doesn't is make practical sense as an engineer to do my theoretical work in metric where it really shines and do my mechanical design in imperial units so it's easier on the machinist?
To follow up on this one of the other big difficulties with the switch is the need for metric tools. I'm not talking about a set of wrenches but the seriously expensive machine tools and metrology equipment that prototyping shops across the county rely on. I personally have several thousand dollars of just measuring equipment like micrometers, dial indicators, and gage blocks before even looking at the milling machine and lathe which have inch threaded screws. In a well equipped shop this could add up to several hundred thousand dollars of equipment that simply doesn't work in metric. To a shop that is perfectly happy using inch measurements there is no incentive to switch. On top of this, almost all machined parts are done in decimal inches. The whole power of ten advantage means nothing to a machinist because all the work is already done in decimal.
As more modern NC equipment trickles down to the smaller shops that form of the base of American manufacturing the problem is getting less severe because it's as easy to programming a few lines to switch to metric or pressing a button on digital measuring equipment. I wouldn't hold my breath though.
Well if you had RTFA'd you notice the Creative Commons License 2.5 button on the bottom of the page. Many professors and assistants go to great lengths to get permission to post materials they use in courses (I get notices from time to time for permission to post diagrams from our FSAE team), find alternative Free sources, and strip out non-Free material. It usually works out very well since a lot of classes rely almost exclusively on the professor's own material.
Same here, thanks to a generous internet connection at school many of the people I live with (including myself) maintain personal servers in the dormitory. My server is a normal windows box (I'm a mechE so Linux is a non-option) and I just Remote Desktop (or rdesktop) or VNC in from wherever I am, whether it's the lounge twenty feet away or at home a couple hundred miles away. The connection is nearly seamless even over an 801.11g link over cable internet. I can dial in from my underpowered laptop and load insanely sized files in Solidworks, read Slashdot or whatever else I feel like. I can also start long jobs on the server and shut my laptop off whenever I want. For Linux work I just SSH into a Debian based server.
I've already got everything an online desktop could provide, except I'm using real full-featured software packages.
Checking in on the hall IRC server I see twelve users, ten of whom are logged in from boxes within a hundred yards of the IRC server but none of the users are within fifty miles of each other.
It is like suggesting that Formula 1 engine design will give better engines for everyday use. This is just not so. The goals of F1 engine design are to give max power for a couple of hours; it does not matter if the engine blows up after 4 hours of use or if it has to be stripped and rebuilt after every race. That mindset is completely hopeless for everyday vehicles.
The idea that bleeding edge research leads to direct spin-offs is one of the most misdirected arguments I've seen used by either side of the debate on this. Pull apart any car engine of somewhat recent design and you'll find it borrows liberally from both current and past high performance design. You may not be running at 15,000 RPM but you're using electronic controls, materials science, mechanical design, and analysis developed for bleeding edge both from previous generations and today. Where would we be without Fracture Mechanics originally developed because of bleeding edge applications? How about CNC machines that were originally commissioned by the US Air Force? The definition of research IS bleeding edge. It's not application specific products that fall down down to household use, it's bits and pieces of theory and process that come together to make the things that keep your Honda running for 300,000 miles and put 777s in the air.
I worked at a base this summer and had to get one for physical and computer access and was pretty impressed with how they were handled. They were taken very seriously from a security standpoint and had few technical problems.
I'd check out what SIPB (Student Information Processing Board) has done for the MIT community. They've been around almost forever and have done a lot of great of things over the years.
I guess that clears that up then. I'm surprised we don't have better tools to confirm underground tests what with the whole Cold War and all. Hopefully we'll get some rock solid results tomorrow.
With the seismic data showing that something happened, my question now is which event happened first. Did NK perform a sucessful test which caused the disturbance, or did NK detect a natural event and declare it a sucessful nuclear test? The claim that "no radiation was released" makes me suspicious of the latter, but I don't think any of us have enough data to form much of an opinion on the matter.
Even more impressive than that is the new XM982 GPS guided arillery round. The engineering required to make a modern guidence computer and control mechanisms survive several thousand Gs is absolutely mind-blowing.
I certainly didn't think my post should have come off advocating a reliance on machines; my point was that proper engineering should have a reliance on real world testing. I firmly believe that anyone running FEA software should at very least have a working knowledge of both traditional methods of analysis and be familiar with the actual mathematical process of FEA. As many others have said, you'll never know when the software is wrong without a sense of what is right, but I'd like to add that if someone is relying on software they should know how the algorithm works and how it fails.
You're right that an alternative method of analysis is a good way to check a problem, but it's still no excuse for proper testing of a critical part. Structural analysis is based on so many assumptions and approximations (let's put our hands over our ears and yell 'linear elastic!') that it can never be 100%. Something can always come out of a blind spot and kill someone. No matter how good your analysis is it'll never tell you that the manufacturer's process leaves cracks in the material. I could say something that Tacoma Narrows, but it's far too abused.
I don't know the specifics of the problem you worked out but your mention of NASTRAN told me it was originally done with FEA. I assumed that if the original guy broke out FEA to solve the problem it wouldn't have been as trivial as you made it out to be. Of course, I just showed a coworker how to solve a beam bending problem when he wanted a copy of ANSYS to do it (it DID take several pages of my notebook to figure out good path to the solution, but still, doing FEA on the thing was ridiculous) so there goes my theory.
I am quite familiar with Fermi problems and back of the envelope calculations (my notebook is filled with them). Knowing how to break a problem down into something that can be checked for feasibility is something that too many engineers I know are useless at. They'll never know what ideas are worth pursuit without preliminary analysis. However it is exactly that, preliminary. In the end real testing is required.
You're definitely correct about students being able to choose the subculture more easily, but I disagree with the ten year number. It's too hard to live this "geek dream" and be taken seriously. While we may play that game at school when we can, very few have any problem putting it away to interact with the real world. If it takes taking shower (I'm looking at you, course 6 Stallman wannabes) and putting on a suit to get something done it'll happen.
Those of us who work for the Army need our caps locks! A quarter of all the official documents I write are in all caps. We even have special rules of punctuation for dealing with all caps (special cases for appostrophes and such). Just think of how badly it would hurt to write a document that starts with DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE without my caps lock!
You say there is absolutely no correlation between technical ability and "geek-hood" but my experience, while limited, is completely contrary. I live with what many have deemed among the most technically able students in the country. While this distinction can be argued my personal experience has indicated it to be true.
These are exactly the people you have described. Comic books, anime, explosions in the courtyard, jokes using regular expressions and set theory, a very peculiar sense of fashion, it's all here. There is definitely some relationship between the top end of technical ability and these traits. My home being in the area of the old Bell Labs I've alse heard the same stories from people that used to work with the researchers there. I'm convinced that it's not a case of confirmation bias.
Being host to many geeky events (things like the Star Wars musical and the Penny Arcade lecture) we also get to see plenty of the dreadfully technically and socially incompetent nerds come in from the surrounding area. I can't say any more than my experience has led me to believe, but I think technical ability definitely correlates (no causality stated or implied) with the quirky traits mentioned. There are plenty of geeks with no ability, but the majority of the people with the greatest ability that I know are definitely geeks.
Umm, that engineer was right, you don't check computation with more computation, whether it be by hand or computer. You check ALL computation with testing. Are you telling me that you'll make any progress checking even remotely complex structures by hand? Any analysis should be assumed faulty in critical cases and on low FOS. All your analysis is worthless until you apply the actual loads with actual instrumentation.
# And now comes this OSS issue. Once again, these Europeans appear to be a bit ahead.
And here I thought that my school was firmly planted in the United States. I mean it's not like we've had a whole lot to do with this OSS thing, but we've been at it for a pretty long time.
I really wish some of these projects would do a better job of compiling for different architectures. My friends and I would love to be able to help some of the more interesting projects with the extra cycles on our servers but very few of the projects are compiled for Alpha and fewer still will let us do the compiling. We have quite a bit of CPU power to give but the only projects we can find to support us are SETI and Distributed.net, neither of which rank very high in what we consider useful. Does anyone know of some other projects we can run on our EV67s?
I mean, we'd write something cool ourselves, but what do we look like, guys who aren't lazy?
Thank you for making this point so well. As the situation currently stands there is no incentive to take any risk in the shuttle program (and many other government programs). This isn't entirely bad, it works well for things that aren't difficult or inherently risky, but when pushing envolope it just doesn't work. I don't think the people working on the program are actively thinking about job security in the macro perspective of prolonging the shuttle program. My personal experience leads me to believe that it's mostly a major case of short-sighted risk aversion on the micro scale (like "the evidence indicates this pretty strongly but we really need to do more several more test cases to be completely sure")
I don't know whether it's the fault of the slides (I never used them) but I think you must have really missed something. I remember a huge part of his class is spent on electronic structure and how it leads to other physical properties. The class covers up to MO theory and focuses on crystal structure and formation with a heavy emphasis on the band gap structure of semiconductors. There is no way anyone could have gotten close to passing the class without knowing everything you just wrote very solidly.
Slashdot Mirror
LSC..
As another undergrad at MIT (Course 2, mechE) I've got to agree with pretty much everything you said. I can really only speak of the people I know but the impression I get is that not many people here put much effort into making themselves a good admissions candidate. They were all more interested in ROM hacking, playing sports, going to science competitions, making Debian packages, building cars, blowing shit up, doing research, and playing video games. High SAT scores will get you looked at. Along with 7,000 other applicants. Don't make it a game, would you want to accept someone whose primary goal is 'getting into a top tier college'?
I mean, my goal is building a giant nuclear powered robot suit with which I can oppress the people of world. Evidently that's good enough for an early admission.
That's a joke, the admissions department never actually found out about my global oppression robot plans.
Seriously, I can't believe the people discounting this. I do my work as a mechanical engineer and find myself in the situation of trying to measure pictures quite often. Just last week for example one of the people I work with took a snapshot of a very wild engine camshaft (almost rectangular lobes) and I wanted to find the lift, slope, and phasing. The shot was taken with severe perspective, looking down the length from one end. It's easiest to see the profile from this angle but it's a huge amount of work to scale the image by hand. $99 would be a great deal for something so useful.
All computers on that use DHCP need registered MACs, if you've got a static IP there's no need at least as far as my experience has been.
IST does a damn fine job, the stakes on having the network running smoothly are quite high and they get it done, but more importantly is the amount of freedom they allow. We've got the most heterogeneous environment I can think of with hundreds of Course 6ers looking for new ways to bend the network and Course 15s finding new ways to try to break it. There's everything from half broken 486s to Playstation 3s running SVN repositories to completely custom embedded devices sitting all over the network (not that they support these devices) running like a well oiled machine.
I couldn't agree more, seriously get yourself a copy of Horowitz and Hill. The language is accessible and examples are excellent with a focus on useful design work rather than theory.
If you'd like to get your hands a little dirty with theory I also highly recommend Agarwal and Lang's Foundations of Analog and Digital Electronics. It's not nearly as fun to read as Horowitz and Hill but everything you need to know is in there and well written.
I completely agree with you, conversion is easy and it's easy to make imperial parts with metric tools and the reverse. The problem is that as long as my Bridgeport moves 0.001" for every tick on the dial I'm not going to design my parts in millimeters unless I've got a good reason. If I want to work efficiently I can't multiply or divide every number I read in the shop by 2.54. When I measure my part after a roughing cut I need to be able to make a quick subtraction with the dimension on the print and then turn my dial that many ticks. This means to make a metric part in a sane manner I need to convert the prints to imperial before anything. If I need to convert all the parts I designed in metric to imperial to actually make them why exactly did I write them in metric? Once I convert the prints I get a mess of dimensions like 0.394" (10mm), not the easiest numbers to work with.
As someone paying a machine shop to do a job why should I design parts that will cause increased confusion between me and person making the parts and a higher likelihood of error? Doesn't is make practical sense as an engineer to do my theoretical work in metric where it really shines and do my mechanical design in imperial units so it's easier on the machinist?
To follow up on this one of the other big difficulties with the switch is the need for metric tools. I'm not talking about a set of wrenches but the seriously expensive machine tools and metrology equipment that prototyping shops across the county rely on. I personally have several thousand dollars of just measuring equipment like micrometers, dial indicators, and gage blocks before even looking at the milling machine and lathe which have inch threaded screws. In a well equipped shop this could add up to several hundred thousand dollars of equipment that simply doesn't work in metric. To a shop that is perfectly happy using inch measurements there is no incentive to switch. On top of this, almost all machined parts are done in decimal inches. The whole power of ten advantage means nothing to a machinist because all the work is already done in decimal.
As more modern NC equipment trickles down to the smaller shops that form of the base of American manufacturing the problem is getting less severe because it's as easy to programming a few lines to switch to metric or pressing a button on digital measuring equipment. I wouldn't hold my breath though.
Well if you had RTFA'd you notice the Creative Commons License 2.5 button on the bottom of the page. Many professors and assistants go to great lengths to get permission to post materials they use in courses (I get notices from time to time for permission to post diagrams from our FSAE team), find alternative Free sources, and strip out non-Free material. It usually works out very well since a lot of classes rely almost exclusively on the professor's own material.
Same here, thanks to a generous internet connection at school many of the people I live with (including myself) maintain personal servers in the dormitory. My server is a normal windows box (I'm a mechE so Linux is a non-option) and I just Remote Desktop (or rdesktop) or VNC in from wherever I am, whether it's the lounge twenty feet away or at home a couple hundred miles away. The connection is nearly seamless even over an 801.11g link over cable internet. I can dial in from my underpowered laptop and load insanely sized files in Solidworks, read Slashdot or whatever else I feel like. I can also start long jobs on the server and shut my laptop off whenever I want. For Linux work I just SSH into a Debian based server.
I've already got everything an online desktop could provide, except I'm using real full-featured software packages.
Checking in on the hall IRC server I see twelve users, ten of whom are logged in from boxes within a hundred yards of the IRC server but none of the users are within fifty miles of each other.
The idea that bleeding edge research leads to direct spin-offs is one of the most misdirected arguments I've seen used by either side of the debate on this. Pull apart any car engine of somewhat recent design and you'll find it borrows liberally from both current and past high performance design. You may not be running at 15,000 RPM but you're using electronic controls, materials science, mechanical design, and analysis developed for bleeding edge both from previous generations and today. Where would we be without Fracture Mechanics originally developed because of bleeding edge applications? How about CNC machines that were originally commissioned by the US Air Force? The definition of research IS bleeding edge. It's not application specific products that fall down down to household use, it's bits and pieces of theory and process that come together to make the things that keep your Honda running for 300,000 miles and put 777s in the air.
I worked at a base this summer and had to get one for physical and computer access and was pretty impressed with how they were handled. They were taken very seriously from a security standpoint and had few technical problems.
I'd check out what SIPB (Student Information Processing Board) has done for the MIT community. They've been around almost forever and have done a lot of great of things over the years.
http://www.mit.edu/sipb/sipb.html
I guess that clears that up then. I'm surprised we don't have better tools to confirm underground tests what with the whole Cold War and all. Hopefully we'll get some rock solid results tomorrow.
With the seismic data showing that something happened, my question now is which event happened first. Did NK perform a sucessful test which caused the disturbance, or did NK detect a natural event and declare it a sucessful nuclear test? The claim that "no radiation was released" makes me suspicious of the latter, but I don't think any of us have enough data to form much of an opinion on the matter.
Even more impressive than that is the new XM982 GPS guided arillery round. The engineering required to make a modern guidence computer and control mechanisms survive several thousand Gs is absolutely mind-blowing.
n itions/m982-155.htm
http://www.globalsecurity.org/military/systems/mu
I certainly didn't think my post should have come off advocating a reliance on machines; my point was that proper engineering should have a reliance on real world testing. I firmly believe that anyone running FEA software should at very least have a working knowledge of both traditional methods of analysis and be familiar with the actual mathematical process of FEA. As many others have said, you'll never know when the software is wrong without a sense of what is right, but I'd like to add that if someone is relying on software they should know how the algorithm works and how it fails.
You're right that an alternative method of analysis is a good way to check a problem, but it's still no excuse for proper testing of a critical part. Structural analysis is based on so many assumptions and approximations (let's put our hands over our ears and yell 'linear elastic!') that it can never be 100%. Something can always come out of a blind spot and kill someone. No matter how good your analysis is it'll never tell you that the manufacturer's process leaves cracks in the material. I could say something that Tacoma Narrows, but it's far too abused.
I don't know the specifics of the problem you worked out but your mention of NASTRAN told me it was originally done with FEA. I assumed that if the original guy broke out FEA to solve the problem it wouldn't have been as trivial as you made it out to be. Of course, I just showed a coworker how to solve a beam bending problem when he wanted a copy of ANSYS to do it (it DID take several pages of my notebook to figure out good path to the solution, but still, doing FEA on the thing was ridiculous) so there goes my theory.
I am quite familiar with Fermi problems and back of the envelope calculations (my notebook is filled with them). Knowing how to break a problem down into something that can be checked for feasibility is something that too many engineers I know are useless at. They'll never know what ideas are worth pursuit without preliminary analysis. However it is exactly that, preliminary. In the end real testing is required.
You're definitely correct about students being able to choose the subculture more easily, but I disagree with the ten year number. It's too hard to live this "geek dream" and be taken seriously. While we may play that game at school when we can, very few have any problem putting it away to interact with the real world. If it takes taking shower (I'm looking at you, course 6 Stallman wannabes) and putting on a suit to get something done it'll happen.
Those of us who work for the Army need our caps locks! A quarter of all the official documents I write are in all caps. We even have special rules of punctuation for dealing with all caps (special cases for appostrophes and such). Just think of how badly it would hurt to write a document that starts with DISTRIBUTION STATEMENT A: APPROVED FOR PUBLIC RELEASE without my caps lock!
You say there is absolutely no correlation between technical ability and "geek-hood" but my experience, while limited, is completely contrary. I live with what many have deemed among the most technically able students in the country. While this distinction can be argued my personal experience has indicated it to be true.
These are exactly the people you have described. Comic books, anime, explosions in the courtyard, jokes using regular expressions and set theory, a very peculiar sense of fashion, it's all here. There is definitely some relationship between the top end of technical ability and these traits. My home being in the area of the old Bell Labs I've alse heard the same stories from people that used to work with the researchers there. I'm convinced that it's not a case of confirmation bias.
Being host to many geeky events (things like the Star Wars musical and the Penny Arcade lecture) we also get to see plenty of the dreadfully technically and socially incompetent nerds come in from the surrounding area. I can't say any more than my experience has led me to believe, but I think technical ability definitely correlates (no causality stated or implied) with the quirky traits mentioned. There are plenty of geeks with no ability, but the majority of the people with the greatest ability that I know are definitely geeks.
Umm, that engineer was right, you don't check computation with more computation, whether it be by hand or computer. You check ALL computation with testing. Are you telling me that you'll make any progress checking even remotely complex structures by hand? Any analysis should be assumed faulty in critical cases and on low FOS. All your analysis is worthless until you apply the actual loads with actual instrumentation.
And here I thought that my school was firmly planted in the United States. I mean it's not like we've had a whole lot to do with this OSS thing, but we've been at it for a pretty long time.
I really wish some of these projects would do a better job of compiling for different architectures. My friends and I would love to be able to help some of the more interesting projects with the extra cycles on our servers but very few of the projects are compiled for Alpha and fewer still will let us do the compiling. We have quite a bit of CPU power to give but the only projects we can find to support us are SETI and Distributed.net, neither of which rank very high in what we consider useful. Does anyone know of some other projects we can run on our EV67s?
I mean, we'd write something cool ourselves, but what do we look like, guys who aren't lazy?
Thank you for making this point so well. As the situation currently stands there is no incentive to take any risk in the shuttle program (and many other government programs). This isn't entirely bad, it works well for things that aren't difficult or inherently risky, but when pushing envolope it just doesn't work. I don't think the people working on the program are actively thinking about job security in the macro perspective of prolonging the shuttle program. My personal experience leads me to believe that it's mostly a major case of short-sighted risk aversion on the micro scale (like "the evidence indicates this pretty strongly but we really need to do more several more test cases to be completely sure")
I don't know whether it's the fault of the slides (I never used them) but I think you must have really missed something. I remember a huge part of his class is spent on electronic structure and how it leads to other physical properties. The class covers up to MO theory and focuses on crystal structure and formation with a heavy emphasis on the band gap structure of semiconductors. There is no way anyone could have gotten close to passing the class without knowing everything you just wrote very solidly.
Looks like Sadoway may just be on his way to that Nobel prize he's been obsessed with. :P
n gineering/3-091Fall-2004/LectureNotes/index.htm
For those that aren't familiar with MIT's most pimp chem prof you can enjoy a full semester of his lectures right here: http://ocw.mit.edu/OcwWeb/Materials-Science-and-E