Book review on “Strike by Lightening: the Curious World of Probabilities” by Jeffrey Rosenthall
Life is a chance encounter, full of randomness. I came across this book when I was discussing the famous “Monty Hall Problem” with one of my friends. The Wiki link of the problem (http://en.wikipedia.org/wiki/Monty_Hall_problem) led me to the author of the book.
I finished reading this book on a train journey to the North of England. Although it is a book on popular science, I am very much interested in many insights of the author on randomness. The book demystifies many random phenomena and it’s a very interesting book to read. Personally, it brings back many of my fond memories on randomness. My first patent, granted worldwide, is about a random number generator design. It was done when I landed my first job in bonnie Scotland in the mid of ‘90s.
For many people, randomness is a vague concept. When asked what randomness is, few can give an accurate description. It is actually quite normal because randomness is not only mathematical but also philosophical, as said “If you can define it, it is not random.” Nevertheless, many efforts have been made to generate random sequences consisting of only 0’s and 1’s. A 0-1 random sequence simulates the result of tossing a coin multiple times. That is, if we toss a coin many times, we’ll get a sequence of heads and tails. By noting 0 for head (H) and 1 for tail (T), we’ll get a 0-1 random sequence, such as 010010100 for HTHHTHTHH.
A random number generator is a device to generate such sequences for many applications, including secure communication, payment, identification, electronic locks (e.g. car locks), real-world simulation, gaming, telephony systems, Monte Carlo method, and so on. The business unit that I worked for designed, manufactured and sold products containing a random number generator. A challenge that they were facing was to demonstrate their products would generate sequences random enough. One particular product was being questioned by the biggest customer, Visa, when I just joined the company, but my business unit was not able to provide a convincing answer. My boss, Ross, was concerned – his team's top priority was to solve all the application issues from the field to help the sales.
It was just before my first Christmas in Scotland and my arrival appeared perfectly timed. Randomness was my turf and I could apply my mathematics strength to this problem.
Like many big companies, my company enforced a long compulsory Christmas break to have the machinery serviced. It started from Christmas Eve and ended the day after Hogmanay. I took chance of this ten-day break to read many up-to-date research papers and work out a comprehensive test plan.
At that time, I was facing three formidable challenges. The first was that I needed to write a randomness test suite myself before the sales window was closed. It was long before the NIST SP 800-22 being published and adopted; and there was no off-the-shelf software package readily available. The second was about what I would have to do if our product should fail. It was speculative but I had to be prepared. The third was to compile and present the test results in a concise but convincing way. I reckoned I could get help from the company.
The test suite development went smooth. I made a test suite immediately after Christmas break. However, by applying the test suite to the product, the test result wasn't the desired. The product failed and the sales were panic. I felt somewhat guilty of “shooting down” the promising product. It was a dilemma – re-designing the product would miss the sale opportunity and lose all the investment, but a product with poor random number generator was not acceptable by the customer.
A meeting was called with the design team to see whether there was any remedy. During the design team’s presentation, I noticed a feature of the product that might compensate the drawbacks shown in the randomness test. An innovative patch was made to take advantage of the feature to have the product behave randomly in another dimension. The synergy of the two random operations (one by the product itself and the other by the feature) made the random number generator exceeded the customer’s expectation.
The finding of the feature and its usage were further developed and it became my first patent awarded worldwide. Thanks to this innovation, the whole product family could share the same design and manufacturing cost was significantly reduced. In addition to the award from the company’s patent committee, Ross handed me a $5,000 bonus notice in my first performance review after I joined his team for just four months. It was a significant lump sum in the mid 90’s and it was nice.
