Sunday Snapshots (27th October, 2019)

Invisible Scripts, Michael Bloomberg, Natural disasters and neural networks, drones, and demonstrated excellence

Hey everyone,

Greetings from Chicago!

We all have a set of “invisible scripts” that guide our thinking at the most subconscious levels. Going off script doesn’t seem possible – if you did, would you even be yourself anymore?

I’ve been challenging myself to think about the invisible scripts that I follow blindly. These scripts are often good heuristics, but they can be dangerous as well. It’s easy to go through life on autopilot and end up with a life where you didn’t fulfill your potential.

My challenge to you this week is to think about an invisible script that you follow. Send it to me by replying to this email or on Twitter at @sidharthajha.

Here’s a photo of the Chicago skyline I took on Friday:

In this issue of Snapshots, I want to talk about:

  • Michael Bloomberg and blessings in disguise

  • Predicting damage caused by natural disasters

  • Drones and beachhead markets

  • Demonstrated work ethic and persistence

  • And more!

Book of the week

This week, I read The Many Lives of Michael Bloomberg by Eleanor Randolph. I first read about Bloomberg in Bradley Tusk’s The Fixer but that mostly focuses on his time as mayor of New York City. I wanted to dive into how Mike Bloomberg became one of the best leaders across different arenas. Here were my three main takeaways from the book:

  1. Bloomberg as a student: He often quips when speaking to student groups that he “was the kind of student who made the top half of the class possible.” Bloomberg was a C student with the occasional D. But what he lacked in academics, he made up with his leadership engagements. He charmed cooks and classmates alike. He was conscious about being perceived as a hard worker because he wanted to make things look effortless. His efforts paid off – a longtime friend says that “Most of us were just college kids living in the moment, Mike was living in the future.”

  2. Blessings in disguise: Bloomberg started his career at Solomon Brothers in 1966. By 1973, he became a General Partner at the firm. But when the firm was sold, he was pushed out. The partnership he loved so much did not love him back. He was devastated. Ultimately, this turned out to be a blessing in disguise. With the equity piece of the company he was owed as a partner, Bloomberg started his eponymous company which made him a member of the “three-comma club.” He also learned a key skill – how to fire people. At Solomon, he was let go without a proper sendoff and etiquette. But when an executive at Bloomberg LP is laid off, the first phone call is from Mike.

  3. Riding a rising wave: In investing, it’s almost always better to ride a rising wave than to go against the current. When Bloomberg started his beta tests of his terminal in the 1982 with Merrill Lynch as their first customer, different sources of financial data were just coming on the computer. Bloomberg’s terminal was an aggregator of all this information which allowed traders to make faster and better trades. By riding the wave of information explosion, Bloomberg started an empire.

The book is not particularly well-written, but it’s a broad introduction to Bloomberg. I would like to explore some of the other works on his life, including his autobiography.

Long read of the week

Building Damage Detection in Satellite Imagery Using Convolutional Neural Networks

When natural disasters strike, lots of resources need to be deployed as soon as possible. The details of how much and where those resources go to depend on the location and magnitude of damage. It’s often difficult to estimate this damage quickly as on-the-ground communications and manpower is scarce in the hours after a disaster. In this paper, Google researchers collaborate with the United Nations World Food Programme Innovation Accelerator to develop an image detection technique based on neural networks to assess the damage done to buildings.

The tricky part of this technique is that different locations and disaster types cause different types of damage. Therefore, it’s difficult to predict how future damage will look like based on past damage. The researchers develop a novel way to deal with this – they train their algorithm on a dataset with images from a mixture of different locations and a very small set of data of the actual disaster that you’re trying to assess the damage for. This data could be obtained through one or two rescue workers on the ground fairly easily.

Another important part of the paper is how they evaluate different architectures of their neural network model. They use something called AUC (Area under the curve) of the ROC curve. The ROC curve looks at the relationship between false positives (predicting something is true when it’s not true) and true positives (predicting something is true when it is true) for different classification thresholds – as you decrease how difficult it is to be classified as true, both false and true positives increase. But you want to minimize false positives while keeping true positives high.

A way to think about how well a model does at managing this tradeoff is aggregating true and false positives across all classification thresholds. This can be done using AUC. You think about it the probability that the model ranks a random positive example more highly than a random negative example. So the higher the AUC value, the better your model is at balancing the true/false positive tradeoff.

I really like this paper because it’s well-written and relatively easy to follow if you know some of the fundamentals.

Business move of the week

The drones are coming (WSJ)

The most expensive part of logistics is last-mile delivery. This is typically from the retailer or a small warehouse to the end customer. The routes become customized and as a result, optimal routes become more difficult to find or infeasible. But that might be about to change.

Drones offer another dimension of travel that offers direct routes. You don’t need to wait for a full load. You simply fly off when an order comes in. While Amazon is obviously the pioneer when it comes to this technology, an interesting use case is medical supplies that need to be delivered just in time. Zipline was the original player in the space, but UPS is also trying to get into this market. I think it’s the perfect beachhead. Organs and supplies need to be transported on short notice and cost are a secondary consideration. This allows the company to build a fleet across the country and iron out the kinks in the system. The life and death nature of the deliveries also means that reckless adventures with the “move fast and break things” motto will be minimized.

Gesture of the week

Demonstrated work ethic and persistence in any field is transferable. With this principles in mind, Shopify CEO Tobi Lütke hired one of the best Starcraft players as an intern. It’s an incredible gesture not just for the player, but also for kids who have strong interests in non-traditional fields. I wish we saw more of this.

Meal of the week

This week, I went back to Politan Row in West Loop but tried a different food stall – Piko. The Singaporean Chicken Rice was good and it brought me back to when I lived in Manila when everything was laced with a cocktail of soy sauce and MSG. Definitely worth a visit.

That wraps up this week’s Sunday Snapshots. If you want to discuss any of the ideas mentioned above or have any books/papers/links you think would be interesting to share on a future edition of Sunday Snapshots, please reach out to me by replying to this email or sending me a direct message on Twitter at @sidharthajha.

Until next Sunday,