I’ve been reading more books during this break. This week’s main read is one of those books, recommended by a former colleague at Disney. The rest is the usual mix of randomness.

  • Competing in the Age of AI: I finally read this book after a colleague at Disney recommended it to me. This person had built an impressive ML team that did some groundbreaking and unique work analyzing Disney content to generate additional metadata for personalization, improve QC of Disney+ artwork, and a number of other things (their patent collection is pretty impressive). I’m about halfway through, but can recommend this one wholeheartedly. The first half summarizes the core data strategy I’ve espoused to my teams for a long time. The book also describes the challenges with instituting this sort of strategy at a company as big as Disney. Silos are nasty things when it comes to data, and inertia is a major force, even at a place as aggressive about reinventing itself as modern Disney. Some really good ideas in this book so far, gives me hope that the strategy we kicked off before I left may bear some fruit.
  • Deep Learning Is Hitting a Wall: This was a really interesting read, for a couple of reasons. First, I now definitely need to read more about a possible division in the AI space between symbols and deep learning camps - I’m amazed that’s actually a fight, and the author is not a neutral observer. Yet, the history he presents is plausible and compelling. Second, the site publishing this essay seems pretty interesting. Nautilus is a science-focused outlet (with a subscription program) that seems to go pretty deep into the topics they cover. Loved this article about new hardware for Neural Networks, for example. Never thought about the power efficiency of AI models.
  • New best-in-class driver for Waveshare 5.65in (F) 7-color e-paper display: First, wow, there’s a 7-color e-paper display… hadn’t seen that! Second, this library is pretty impressive. If you need a low power periodic display, e-paper is a really compelling option. I am impressed at how far people are pushing them. This library looks good. Curious to see it more in the wild.
  • Plain Text Sports: Two months ago, maybe I don’t share this, but honestly, I don’t see how this thing will make money, so it probably doesn’t matter to ESPN in the grand scheme of things. Also, if I were going to build a new sports product in 2022, while I’d definitely work on loading speed and privacy first advertising, I’d focus on other gaps that are, IMO, a much bigger opportunity. Lots of ways to get scores quickly out there. Yet, all the scores pages are structured and work the same, even this one. There are big, unsolved problems in that UX.