Wednesday, October 31, 2007

Learning from United States of America

When I was in my undergrad in India sometime in 2001-2002, our college Principal Narasimha Reddy said "Go to America and learn the work culture". Recently in 2007 at a workshop on "Computer Vision applications for developing countries" that was in conjunction with ICCV in Rio de Janeiro Brazil, Ramesh Jain gave a talk in which he mentioned how the research in developing countries like India is aimed at working for USA since most researchers focus on publishing in conferences, journals run by USA which address problems pertinent to socio-economic situations of typical developed countries.

OK. The longer I stay in America the deeper my appreciation of America grows. What can we (developing countries) do to really learn from it? Ideally we want to learn the underlying process that generates the American success. We typically use observed data to learn the process: an inverse problem. We need data that can truly characterize American success. If we use wrong data no matter how strong our learning algorithms are, we are bound to learn wrong concepts. Typical problems that we run into in learning are that the data acquired is too noisy and even corrupted by unwanted signficant processes. So to be efficient in learning from USA we need data as little corrupted as possible. To acquire such data:
  • One needs to have a first hand experience of life in USA.

  • One needs to be open and courageous so that he actually collects the data.

  • Should interact with atleast a few elite Americans not necessarily personally but through different media like reading articles, blogs and working etc.

  • I say elite because they minimally corrupt the data. And if you encouter a few of them it's good enough to learn the ground truths and you can peel off layers when interacting with average or below average Americans.

  • Check objectively if you are able to learn. A good test is to check if you like America more than just for money and comforts. Understand that I am not preaching money is not important I am just saying it is not the causal variable for deep dynamics. It's how we use money that matters.

  • Periodically "smooth" your data by re-interpreting past experiences.

Just a side-note. Nobody is perfect. America has its faults but it is actually a worthy leader in doing what matters, that is struggling for ensuring longterm human survival. Doing nothing is trivial and doing things by being all powerful is also trivial. What is non-trivial is understanding the dynamics with constraints.

My expressions in last paragraph are influenced by my regular reading of Scott's blog. Especially the "trivial/non-trivial part" and "the doing what matters thing"s.

Note: I edited this post so many more times than any other posts after publishing.

No comments: