aksbhat

vor 14 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

its load shedding (which means that electricity is turned off for few hours everyday) which is quite common in the subcontinent.

aksbhat

vor 15 Jahrendiscuss

another guy https://twitter.com/#!/smedica

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

How can you determine utility of a feature before it is built? Well the answer is you cant! And surely not for a High End Medical device!

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

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.

aksbhat

vor 15 Jahrendiscuss

"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.

aksbhat

vor 15 Jahrendiscuss

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... ?

aksbhat

vor 15 Jahrendiscuss

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.

Submissions

This Guy Live-Tweeted The Bin Laden Raid Yesterday Without Knowing It

Mining Twitter follower network of 36 million users

comments