Lecture 005

Book: Born Loser: A History of Failure in America.

Sandage: The interviewee argues that failure exists only with its context. What I mean is that the failure, or the feeling of failure, can not exist without comparison between the status, whether financial or physical, of people. Certainly, as the interviewee argues, technology does help spread the feeling of failure. Peer pressure is real among 21-century students. With technology, there are 1. more things to compare, 2. easier to compare, and 3. more meaningful to compare. Firstly, technology enables more communication between physical locations that are far apart. Therefore an individual can have comparisons across the globe, increasing the chance of finding someone who is "more successful" than himself/herself though digital technologies (phones, computers, TV, etc...). Secondly, there is statistical information that is designed for such a comparison. Students are evaluated with Grade-Point-Average, and you have to take certain units of class to graduate... And thirdly, with the credit system (the belief such that a person who succeeds will have a higher rate of success) and the stock market is designed to statistically quantify success in order to gain profit. With these, it is not hard to prove that there is more sense of failure than maybe a century ago.

Lincoln: This story reminds me of the global minimum. In machine learning (a statistical approach to generate good computer algorithms that do good jobs in many tasks such as classifying objects), failure is scary too. Machine Learning engineers need to find a way to balance between risk and failure in order to modify its learning algorithms in a such way that the algorithm is both fast and accurate. Like humans, computers need to take risks in trying different new approaches to solving a task. When it uses the greedy approach (ie. not taking any risk), the computer can learn nothing out of the dataset. When the risk is too high, the computer keeps exploring different new approaches without sticking to a solution that works. Neither of these two is a good strategy, so Machine Learning engineers need to adjust hyperparameters to find the sweet point between the two. We, humans, do the same. Without failure, there exists no success.

I want President: Here Response: Here Years Later: Here


He loves me not: Here Take care of yourself: Here


I am a volunteer investigative reporter, welcome to the Presidential Show for 2020 election. Any of your response will be broadcast to the whole United State of America. I am a student volunteer at the Presidential Show to interview you as a part of the program. You will be sitting in front of the camera to directly respond to this letter I am writing.

Mr. President Trump, I love your

Table of Content