Raw thoughts from Alex Dong

The Feudal cloud, Einstein's secret of learning anything, Random forest in Python and dirt cheap to save all US phone calls

Schneier on Security: The relationship between us and the “cloud service providers” * is inherently a feudal relationship. We cede control of our data and computing platforms to these companies and trust that they will treat us well and protect us from harm. … In the longer term, we all need to work to reduce the power imbalance. Medieval feudalism evolved into a more balanced relationship in which lords had responsibilities as well as rights. Today’s Internet feudalism is both ad hoc and one-sided. We have no choice but to trust the lords, but we receive very few assurances in return.*

The Secret to Learning Anything: Albert Einstein’s Advice to His Son: I am very pleased that you find joy with the piano. This and carpentry are in my opinion for your age the best pursuits, better even than school. Because those are things which fit a young person such as you very well. Mainly play the things on the piano which please you, even if the teacher does not assign those. That is the way to learn the most, that when you are doing something with such enjoyment that you don’t notice that the time passes. I am sometimes so wrapped up in my work that I forget about the noon meal… .

Random Forests in Python: random forest is an aggregation of other models. The algorithm to induce a random forest will create a bunch of random decision trees automatically. Since the trees are generated at random, most won’t be all that meaningful to learning your classification/regression problem (maybe 99.9% of trees). Simply, the 99.9% of trees that are irrelevant make predictions that are all over the map and cancel each another out. The predictions of the minority of trees that are good top that noise and yield a good prediction.

What would it cost to store all of America’s phone calls?: Intuitively we all know that storage is cheap but this one still surprised me. Considering Facebook’s 57B market cap, no wonder why privacy is so cheap. $27M in capital cost, and $2M in electricity and take less than 5,000 square feet of space to store and process all US phone calls made in a year. The NSA seems to be spending $1.7 billion on a 100k square foot datacenter that could easily handle this and much much more.