HN user
Its not that bad, I took public bus to school as a thirteen year old. It was as crowded as an MTA bus. Except there is no Air Conditioning, otherwise it is not that bad. Trains however tend to be much worse, since the doors never close.
While I wouldn't go as far to call it BS. I agree with your argument.
I believe we need better laws that will ensure the freedom rather than modifying the system. In all other aspects of human life, we have relied on legal system to achieve economies of scale, while preserving our freedom. We dont grow our own food, or manufacture our own drugs, rather we rely on oversight of FDA etc. Sure the system isn't perfect but still its better than having a farm in your backyard and a chemical plant in your basement.
We need a legal revolution to address issues posed by rapidly evolving technology.
I have tried using both Hadoop (55 node cluster at Cornell) and a single AWS High Memory double extra large instance with 32GB memory.
I have found that since the Twitters social graph is small enough to fit in the memory, a single instance with huge amount of RAM is much more efficient, especially when your algorithm iterates over nodes in the network.
You can read about it here:
Hadoop based results: www.akshaybhat.com/LPMR/
Results using a single High Memory instance AWS instance www.akshaybhat.com/LPMR/GRAPHLAB
Even the startup hunch has taken a similar approach and use a single machine with large amount of memory rather than a hadoop cluster.
I believe that the benefits of a social network such as Facebook outweigh the risks. The problem is that, it is very hard to quantify the positive effect which arises from small interactions. Sure there is huge scope for improvement, but one could have made similar arguments against telephony when it was invented.
I believe we are still to see rise of real social network based applications. e.g. something that allows us to estimate trust for a person, given his and your social network.
its load shedding (which means that electricity is turned off for few hours everyday) which is quite common in the subcontinent.
another guy https://twitter.com/#!/smedica
Rather than using Google scholar, I would suggest looking at papers in ICML, NIPS and Journal of Machine Learning Research.
For vision based research I would suggest CVPR and ICCV conference and IEEE Pattern Analysis and Machine Intelligence journal.
link to the paper: http://aje.oxfordjournals.org/content/early/2011/04/19/aje.k...
I guess that a paper won't get accepted in Am. J. Epidemiol without proper consideration of such correlation.
The studies generally try to control for homophilly, contagion and confounding etc.
I completely agree with you!
As someone who has dysgraphia (http://en.wikipedia.org/wiki/Dysgraphia), I have faced this problem so many time in school. Luckily in India esp. in my state, if you have Dysgraphia, you could use a scribe and dictate your answers at least for major examinations.
I am happy that even though I am in a grad school, I now rarely have to write anything.
I tried it and found it really useful.
It is especially useful for people like me who aren't very well versed in CSS. I guess you should market it as WYSIWYG CSS template builder.
You are comparing Garage Band to A High End Medical device?
Technically, they had the feature but the feature didn't help the doctors be better doctors. Features don't automatically translate to outcomes.
In fact your whole argument is based on incorrect assumptions, well if you are building something like a GarageBand or one of the Me Too To Do list/Collaboration apps, go ahead sure do whatever you want. But when you are creating an EMR software or an MRI machine which costs Millions of dollars you better make sure that it has all types of features and customization capability.
The problem is that you are confident in your skills as a developer to predict the needs, While your hunch might be correct with software that is used in daily life, it might be totally wrong while making something that is not
Also "What people can do with your software" is a subset of "What software can do". Thus its the software that is the limiting case.
Interesting article!
A lot of commenter here are scared of using Magik code.
However note that Search/Information Retrieval is a hard problem. Unlike other problems, developing a generalized full text search engine is difficult.
Testing search algorithms is even more difficult e.g. NIST organizes TREC conference http://trec.nist.gov/pubs/call2011.html in which a major emphasis is on evaluation of search algorithms.
In fact Search is as what my advisor calls it, an AI-Complete problem, i.e. creating a perfect search engine would amount to creating a Human like artificial intelligence capable of understanding your query and the corpus.
Think of it this way, it is easier for anyone to write code for solving simple problem. however this does not makes you stand out from the crowd, also at the same time there is no guarantee that the same person wrote the code, i.e. it wasn't plagiarized.
A good metric for measuring ability of git hub user is to look for followers of his repos. Github is useful only when someone is making large amount of contribution or significant contribution, such as solving a hard problem.
A college can give someone A- in introductory algorithms and that gives you some confidence about his skills in algorithms. However github cannot provide this sort of credential, with same amount of confidence. Github can spot "exceptional" programmers, since large or significant contributions are more difficult to fabricate.
Following are the counterpoints to their argument:
1. decreasing ROI on a college degree: They are confusing between supply and demand. While it is possible in future that a college degree will no longer provide an edge over High-School diploma, yet this might lead to even more intense competition for top schools. They cite Law schools as an example, however there isn't a uniform decrease in enrollment. In fact a bad job market means that you need to go to even better/reputed university than before. Also it does not means that alternate signaling mechanism is going to alleviate this problem.
The issue is the demand, assuming that in future the number of jobs available for college graduates decrease. This wont help alternative credential providing mechanisms. The reason people are hiring via git hub and stack overflow, is because there is an excess demand and low supply. If anything a bad job market will only lead to more competition for top universities.
2. MIT open course ware/ ITunes University etc. : While these websites make it easier for you to get access to the knowledge, yet they dont provide you other things which are needed, such as access to labs, examination etc.
3. Github / Stackoverflow :
Systems such as Github and Stackoverflow tend to have Pareto distribution i.e. top 20% have 80% contribution/reputation. Even if you are building a credentialing system, it needs to have a distribution similar to the grade curve used in colleges. Also at the same time, you must make sure that there is "some" protection of academic integrity.
what is even more interesting is that books are also much cheaper e.g. Algorithms by Cormen, et. al. cost only 6$. It would be interesting how eBooks are priced in India, as compared to US.
If you are a good dev/student and some bank is willing to lend you 40,000$, it is far easier to join a good US university for a masters degree and then get a job in Silicon valley.
So all you are left with are people who cannot take above option.
Not true, "most" of the research gets funded by NSF, NIH, DARPA, ARPA, Naval Research. Of course there are companies such as IBM research, Microsoft research but they are few.
By money and for money generally gives you High Frequency trading bots.
"Hammerbacher looked around Silicon Valley at companies like his own, Google (GOOG), and Twitter, and saw his peers wasting their talents. "The best minds of my generation are thinking about how to make people click ads," he says. 'That sucks.'"
I loved this quote, I also had a similar experience. In past I had an internship in a similar role and pursued few projects on mining communities in large social networks. However luckily I got an amazing opportunity of working at med school/hospital affiliated with my university. I now apply similar algorithms, but now I help radiologists and physicians.
Lets hope that in future the methods that are developed for optimizing ad clicks could probably be useful in some other field. E.g. how IBM is now planning to use Watson in healthcare. My guess is that a lot of algorithm used in developing Watson, were developed for ranking Ads. Closer home, I use a 55 node Hadoop cluster for processing 19 Million annotated PubMed abstracts.
https://www.amia.org/files/workforthesecondaryuseofhealthdat... provides a good discussion about the subject.
Then what are your thoughts about http://www.practicefusion.com/pages/pr/practice-fusion-teams... ?
Do you imply that services that sell anonymyzed patient data are unprofessional?
Medicine has progressed because of sharing of information, and I dont see any harm in anonymyzed information being shared, as long as its a fair and open market.
Cant you use it on Xoom considering it has a Web interface.
It's Stanford.
You can have a look at stats of Cornell CS grads: http://www.engineering.cornell.edu/resources/career_services...
The median was ~ 75K in 2009
Complex numbers are a powerful tool for studying the two-dimensional plane: each point corresponds to a unique complex number . The beauty of this correspondence is that it allows you to add, subtract, and multiply points in the plane
Harvard affiliates? There are no Harvard affiliates but there are higher education affiliates, which are called as universities. Harvard itself is one such affiliate in some sense.
Universities are a result of Theory of The Firm, since a university is generally restricted to small geographic region, it can decrease the transaction costs between the various components of higher education. However if you franchise a university across multiple locations, the net transaction costs go up.
E.g. University of California system can be considered as an affiliate system. However you would still find one UC e.g. Berkeley more reputed than other e.g. Santa Barbara.
Peter Thiel incorrectly assumes, that a University system can share a lot of resources between distant geographic locations. However most of the resources such as faculty, labs cannot be shared. Thus you are better off increase capacity of students rather than branching out.
First, your assertion
"college can be valuable but is not necessarily so"
You can always claim that "X can be valuable but is not necessarily so". Where X can be anything in this universe.
Second, you are trying to turn a personal experience into a generalization. A good question to ask would be what would happen If the college enrollment dropped by 10% or 20% or 50% ? How would this affect the workforce and the industrial output?
Another extremely important question is that how do you decide who exactly should drop out?? Should people drop out on basis of financial resources or some limited test of intelligence. The problem is that it is hard to tell who should enroll and who shouldn't.
While certain subjects can be learnt online, a large number of subjects cannot be learnt. A university provides Labs, Wide range of subjects, access to a social network of like minded people, Faculty, Assessment and most importantly some guarantee of Academic Integrity. Universities are an example of the Theory of the firm.[1] You can surely separate different components of college education which I mentioned above, however there will be significant transaction costs.
The underlying theme of this discussion is not about education or elitism. It is about allocation of capital. Peter Thiel being a Venture Capitalist is interested in allocation of capital, thus the whole 20 under 20. Dont confuse his efforts with education or elitism. If anything his program might be more elitist than Harvard.
I find the discussion here as well as the article myopic in the scope. Programming/Web Startups are just a small part of the world. You still need Doctors, Lawyers, Engineers and Scientists. Sure one can question the utility of spending four years to get a degree in Liberal Arts, but most other subjects require close supervision that one receives in a college education.
I agree they are not very high confidence classifications. Due to random nature of the algorithm, It is very hard to correctly label a community. Also since for most users explicit permission to follow is not required which leads to dubious/spam users. Since the data was collected in June 2009, it is possible that a lot of users might have blocked spam profiles from following them.
I guess I need to clarify this point.
I made a mistake in typing the url.
EMR's are just one part of healthcare, What we need is for all EMR's to follow standards which would make extracting/reusing information from them straightforward.
on a related note: http://challenge.gov/HHS/134-smart-apps-for-health
Sorry but I side with Facebook, a freely available public graph of millions of users could have been used for re-identification attacks.
Frankly you should never share your friends list publicly.