TECHNOLOGY: NSA infiltrates links to Yahoo, Google data centers worldwide, Snowden documents say

Via the Washington Post:

The National Security Agency has secretly broken into the main communications links that connect Yahoo and Google data centers around the world, according to documents obtained from former NSA contractor Edward Snowden and interviews with knowledgeable officials. 

By tapping those links, the agency has positioned itself to collect at will from hundreds of millions of user accounts, many of them belonging to Americans. The NSA does not keep everything it collects, but it keeps a lot.

INNOVATION: Our non-profit Digital Green has won the Google Impact Challenge Award (yay!)

Via IBNLive.com:

Search giant Google announced the four winners of its 'Impact Challenge in India' competition, which asked Indian non-profit organisations to show how technology can positively impact India and the world. The winners - Agastya, Digital Green, Janaagraha Centre for Citizenship & Democracy and Sana - will each receive a Rs 3 crore Global Impact Award, 10 Nexus tablets and support from Google to make their project a reality, Google Senior Vice President Nikesh Arora said announcing the winners here.

INNOVATION: United States government teams up with Coursera

Via the New York Times:

Coursera, a California-based venture that has enrolled five million students in its free online courses, announced on Thursday a partnership with the United States government to create “learning hubs” around the world where students can go to get Internet access to free courses supplemented by weekly in-person class discussions with local teachers or facilitators. 

The learning hubs represent a new stage in the evolution of “massive open online courses,” or MOOCs, and address two issues: the lack of reliable Internet access in some countries, and the growing conviction that students do better if they can discuss course materials, and meet at least occasionally with a teacher or facilitator.

MATHEMATICS: Gathering for Gardner - "How to win at Wimbledon" - additional thoughts

A couple more points : (1) the first related to  Bayesian updating; and (2) the second related to framing of problems.

-+-

I believe that the choice you should make depends on how good you are at tennis (at base level). If your only chance of getting even one point against Federer is if he makes a double fault, then you should choose 6-0, 6-0, 6-6 (6 - 0). If you are good enough to perhaps win a game against Federer with non-vanishing odds, then your optimal strategy may be different. The reason for this is Bayesian updating.

- Supposing you choose 6-0, 6-0, 5- 0 (40 - love). If you are really bad, then Federer will adjust his game once he realizes how bad you are. He may choose to serve differently with an intention to avoid double faults as much as possible at the risk of making a bad serve. He has essentially adjusted his per-point win probability. If you, dear reader, are really bad, then one must adjust one's model to account for this Bayesian updating. Thus, 6-0, 6-0, 6-6 (6-0) might be the more realistic solution. Once FedEx realizes you are terrible, you will be toast.

-+-

Another thing to consider - this one very tricky judgement bias and psychological effect, namely, the "Overconfidence effect". Consider the following two examples :
(1) A survey of 1 million high school students showed that 70 percent think they are above-average leaders (only 2 percent rated themselves below average). [1]
(2) In another study 94 percent of college professors claimed that their research was above average. (See Seed magazine reference below : [1])

People may think that they may actually have a chance of taking even one point off Federer whereas in reality, that probability p may be much, much closer to zero. The "6-0, 6-0, 5- 0 (40 - love)" solution may suffer from this bias for people who are somewhere between the good to great spectrum.

-+-
Another point, this one related to the framing of problems.

A puzzle is not the same as a mathematical problem. Whereas a mathematical problem is in the abstract, a puzzle may add additional constraints because it is set in the real world. In my solution, the chances of Federer actually dying may seem like a humorous suggestion or an aside but such a probability must, in fact, be taken into account in the real world. Another thing to consider is that Federer may be tired (this point, which is Amit Chakrabarti's, is different from the situation of Federer actually dying).

Again, a puzzle is not the same as a mathematical problem. If constrained optimization is what you want, then the puzzle setter would want to state the problem more carefully. Settings in outer space are an oft-used techhnique for framing such problems so as to ensure strict mathematical correspondence. Use them. Robots are another useful device. Use them too.
-+-
References

[1] http://seedmagazine.com/content/article/on_overconfidence/

-+-
Update (July 8): fixed typos. updated the post a bit.

ECONOMICS: Raj Chetty is wrong

Raj Chetty writes in the NYT that economics is a science. He is wrong.

My email to Greg Mankiw.

-+-

As a computer scientist with an interest in the Philosophy of science, I am sad to see this title for an article in the NYT. 

For economics to be a science, it would need to have two capabilities, which it does not possess unlike the core sciences: experimentation and verification.

<stuff deleted>

-+-

Postscript: I love Krugman's elegant response to Raj Chetty's article. Elegant is the word for it. Note the use of the word 'maybe'. The reason why it is 'maybe' a science is that it is based on 'hard data'. The reason why it is 'maybe' not a science is that it is not based on reproducible experiments and verification.

-+-


Maybe Economics Is A Science, But Many Economists Are Not Scientists

Raj Chetty stands up valiantly for the honor of his and my profession, arguing that economics is too a science in which careful research is used to falsify some hypotheses and lend credibility to others. And in many ways I agree: there is a lot of good research in economics, maybe more than ever as the focus has shifted somewhat from theoretical models loosely inspired by observation — which, as he suggests, was my forte — to nitty-gritty empirical work.

MATHEMATICS: Gathering for Gardner - "How to win at Wimbledon" - solution

My email to Peter Winkler.

-+-

This is in reference to the "How to Win at Wimbledon" puzzle at :

http://www.math.dartmouth.edu/~pw/solutions.pdf

I have a better solution to your problem. In fact, you can win at Wimbledon with probability 1.

My belief is that the theory advanced by Amit Chakrabarti (who I happen to know actually :) ) is correct. I can see that he is a man after my own heart. The only thing about his theory is that it does not go far enough. The number N that was chosen may not be large enough - assuming that the players can come back for the match if it goes beyond a certain limit, the players may not even be all that tired.

The ideal scores, then, are :

(i) 6-6 (LargeNumberN - LargeNumberNMinus2), 6-6 (LargeNumberN' - LargeNumberN'Minus2), 6-6 (6-0)

(ii) 6-0 , 6-6 (LargeNumberN - LargeNumberNMinus2), 6-6 (6-0)

(iii) 6-6 (LargeNumberN - LargeNumberNMinus2), 6-0, 6-6 (6-0)

Under (ii) and(iii), one needs to choose a sufficiently large N. In particular, one must choose an N that is so large that Roger Federer is at the brink of death. In fact, N is so large and the man is so close to death that if he serves even once more, he will die. Furthermore, it is his turn to serve. That is the situation to find oneself in. 

I believe that I and Amit, *we* are the Ones we have been waiting for. :)

- Anand

-+-

MATHEMATICS: Gathering for Gardner - Winning at Wimbledon

As part of the Gathering For Gardner (G4G) celebrations worldwide, we will be pondering the following puzzle:

-+-

As a result of temporary magical powers, you have made it to the Wimbledon finals and are playing Roger Federe for all the marbles. However, your powers cannot last the whole match. What scoare do you want it to be when they disappear, to maximize your chances of hanging on for a win?

-+-