Hi all,
Welcome to the second weekly edition of MBW. I spent much of the week worrying about finishing a “post” that I could share this week, before realizing I’d rather just share my thoughts as they sprout. Perfect is the enemy of good, as some would say.
Anyway, going forward there will be less polish but more consistency. Just an organic crop of brain waves for y’all.
Hope you enjoy. Let me know what interests you!
Article of the Week
“Inside the complex world of illegal sports streaming” (Yahoo Sports – my former classmate at Northwestern actually wrote this)
It’s rare that a title is so true to its writing. Illegal sports streaming is a market that has complex ramifications for multiple (global) stakeholders, and is a world of its own.
How many people watch these streams? How do they even work? How much money are professional sports leagues missing out on? Why can’t they (or the government) stop the streams? What about companies like ESPN?
The future of sports entertainment is at a crucial inflection point.
A Look At Machine Learning
There’s two parts.
A. (Beginner): A great understandable example of machine learning with obvious benefit.
B. (Advanced): A theory-heavy blog post about the evolving and overlapping roles of man and machine.
Cool? Read either, neither, or both.
A. Stitch Fix
You may be familiar with the subscription “clothing in a box” company. Stitch Fix only charges customers for clothes that they keep, which means it’s in their best interest to send clothes that customers will want. (I love when business and customer incentives are clearly aligned.)
Introducing: Style Shuffle. To learn more about customers’ style preferences, Stitch Fix introduced the swipe-based game that lets people judge clothing options – when presented with a shirt, pair of shoes, or hat, they can say only “yes, I would wear that” or “no, I wouldn’t wear that”. From this information, the company uses machine learning to determine what other pieces of clothing you may like, even if you haven’t explicitly told them.
Or, explained by CTO Cathy Polinsky:
“Let’s say we sent you that turquoise chunky necklace and you hated it,” Polinsky said. “We can see similar people who hated that chunky turquoise necklace also hated the purple chunky necklace. And so we can make sure that we don’t send you that item. Even though there’s nothing about your style profile that had explicitly told us that you wouldn’t like that, we can implicitly gain that based on similarities of other people who liked or disliked certain items that we sent you that you kept or returned.”
Machine learning can be used for a wide variety of objectives, but this application (identifying clothing people may like, then sharing with them) seems like one of the most welcome and digestible.
The article has more detail on how Stitch Fix “maps” each customer’s style and the role of machine learning in increasing satisfaction from customers.
(Disclosure: I own shares of Stitch Fix, although I am not a customer. In the near future, I hope to write about my investment philosophy! As I mentioned, I love how the incentives of the company and customer are aligned, which is important to me.)
B. “Finding the Point of Human Leverage”
The paradigm of "what humans are for" vs. "what computers are for" has changed drastically as the internet and new technologies have emerged. Machine learning in particular appears to be a force with significant potential for further shaking up this relationship.
Here are some excerpts (of the less abstract points).
In the past, when we thought about abuse of computer systems, we thought about technical exploits of various kinds... We thought about ‘hackers’ finding gaps in the software engineering. But if YouTube or Facebook are distributed computer systems where the routers are old-fashioned software but the CPUs are people, then a bad actor thinks of finding exploits in the the people as well as the software. Common cognitive biases become as important as common programming errors.
A crucial and relevant question about the social media moderation issue (emphasis mine):
The original Yahoo internet directory was an attempt at the ‘pay people to do all of it’ approach - Yahoo paid people to catalogue the whole of the web. To begin with this looked feasible, but as the web took off it quickly became an impossibly large problem… Conversely, Google Maps has humans (for now) driving cars with cameras along almost every street on earth and other humans looking at the pictures, and this is not an impossibly large problem - it’s just an expensive one… We’re exploring the same question now with human moderation of social content - how many tens of thousands of people do you need to look at every post, and how much can you automate that? Is this an impossibly large problem or just an expensive one?
We get to the titular point eventually:
“…a lot of the system design is around finding the right points of leverage to apply people to an automated system. Do you capture activity that’s already happening? Google began by using the links that already existed. Do you have to stimulate activity in order to capture the value within it? Facebook had to create behaviors before it could use them. Can you apply your own people to some point of extreme leverage?
“Points of leverage” can be thought of as “what humans can do that machines can’t”. Stitch Fix, for example, has personal stylists, but machine learning does a lot of the work of identifying a customer’s style. Instead of having a personal stylist for each customer, then, learning their style through trial and error alone, Stitch Fix can leverage stylists at scale. This is not only more cost effective, but allows stylists to focus on the uniquely human aspects of the job – building relationships, providing high-quality service.
As machines become capable of more and more, there is the potential to free ourselves to do the most inspiring work. Automation (of any sort) is a challenge for our economy, but it is also an opportunity.
Uncategorized Thoughts
I said of last week’s video (a chimpanzee using Instagram), “a reminder that humans are primates”. I love the idea of a “Humans are Primates” section in future editions – after all, my training as a behavioral scientist often encourages species-level analysis.
Rituals vs. Addictions: Both are regularly repeated behaviors, but rituals are intentional and have a positive purpose. Addictions are compulsive behaviors that we perform despite their drawbacks. Can we use rituals to cure our addictions?
The way we do food delivery right now is unsustainable. It’s a maze for restaurants, who give up meaningful chunks of revenue to participate in a gig economy manufactured by large tech companies (Grubhub, Uber, Postmates, etc.). Also, delivery as it is now generates an incredible amount of waste materials. What if we could deliver, return, and reuse high-quality, stylish, and functional containers and utensils? I’m sure there’s a business model lurking somewhere…
Jk Grubhub has a solution guys!!
If You Live in New York
Presidential candidate Andrew Yang is having a rally in Washington Square Park on Tuesday from 6-8pm. For those of you unfamiliar, he’s an entrepreneur that is running on a platform of “Humanity First”, specifically referencing the threat of technological automation and the need to address the shifting nature of work. Putting aside any political identities, I’m super curious to hear his ideas –and general rhetoric– on the relationship between humanity and technology (that’s kind of the theme of this newsletter, right?) Also, I don’t know if I’ve ever had the opportunity to see a presidential candidate speak in person, and Yang just passed the 65,000 donor threshold for participating in the Democratic primary debates. Let me know if you want to join! It should be interesting!
Video of the Week
Let me know if you want to go in on a group order!
See y’all next week!